Expert Opinion on Drug Metabolism & Toxicology

ISSN: 1742-5255 (Print) 1744-7607 (Online) Journal homepage: www.tandfonline.com/journals/iemt20

Beyond the hype and toward application: liver

complex in vitro models in preclinical drug safety

Sushma Jadalannagari & Lorna Ewart

To cite this article: Sushma Jadalannagari & Lorna Ewart (2024) Beyond the hype and toward
application: liver complex in vitro models in preclinical drug safety, Expert Opinion on Drug
Metabolism & Toxicology, 20:7, 607-619, DOI: 10.1080/17425255.2024.2328794

To link to this article: https://doi.org/10.1080/17425255.2024.2328794

© 2024 Emulate Inc. Published by Informa

UK Limited, trading as Taylor & Francis
Group.

Published online: 13 Mar 2024.

Submit your article to this journal

Article views: 3895

View related articles

View Crossmark data

Citing articles: 5 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=iemt20
EXPERT OPINION ON DRUG METABOLISM & TOXICOLOGY
2024, VOL. 20, NO. 7, 607–619
https://doi.org/10.1080/17425255.2024.2328794

REVIEW

Beyond the hype and toward application: liver complex in vitro models in
preclinical drug safety
Sushma Jadalannagari and Lorna Ewart
Department of Bioinnovations, Emulate Inc, Boston, MA, USA

ABSTRACT ARTICLE HISTORY

Introduction: Drug induced Liver-Injury (DILI) is a leading cause of drug attrition and complex in vitro Received 15 January 2024
models (CIVMs), including three dimensional (3D) spheroids, 3D bio printed tissues and flow-based Accepted 6 March 2024
systems, could improve preclinical prediction. Although CIVMs have demonstrated good sensitivity and
KEYWORDS
specificity in DILI detection their adoption remains limited.
Microphysiological systems;
Areas covered: This article describes DILI, the challenges with its prediction and the current strategies organ-on-a-chip; preclinical
and models that are being used. It reviews data from industry-FDA collaborations and strategic risk assessment; complex in
partnerships and finishes with an outlook of CIVMs in preclinical toxicity testing. Literature searches vitro models; drug-induced
were performed using PubMed and Google Scholar while product information was collected from liver injury; predictive
manufacturer websites. toxicology; Liver
Expert Opinion: Liver CIVMs are promising models for predicting DILI although, a decade after their
introduction, routine use by the pharmaceutical industry is limited. To accelerate their adoption, several
industry-regulator-developer partnerships or consortia have been established to guide the develop­
ment and qualification. Beyond this, liver CIVMs should continue evolving to capture greater immuno­
logical mimicry while partnering with computational approaches to deliver systems that change the
paradigm of predicting DILI

1. Introduction
based on the severity, causality, and frequency of a drug’s poten­
New drug approval rates are keenly watched by the pharma­ tial to cause DILI. Although these drugs went through rigorous
ceutical industry, patients, payers, and shareholders. In 2022, preclinical testing and clinical trials, the lack of predictive animal
drug approvals by the US Food and Drug Administration (FDA) models and diagnostic accuracy as well as the noncompliance
were down by 25%, approaching 2016 levels [1], but figures just with drug treatments are leading to drug failures, patient safety
released revealed that approvals in 2023 were up by 46%—a concerns, and reduced R&D productivity.
20-year high [2]. Pharmaceutical companies continue to plow In today’s world, some might argue that, with a greater under­
billions of dollars into research and development (R&D), and in standing of the pharmacophores that drive hepatotoxicity and
2022, some companies spent 30% of their revenue on R&D [3]. enhanced earlier screening activities, DILI is less of a problem for
Thus, the R&D productivity crisis described as Eroom’s law [4] drug candidates that are under regulatory review. But an analysis
remains as relevant today as it was a decade ago. A key com­ of publicly disclosed reports over the last two years highlights at
ponent to reversing this trend has been attributed to the least eight examples where pharmaceutical companies have
development and use of preclinical models that have a high experienced clinical hold, clinical stops, and, in one case, patient
predictive validity in the R&D process [5,6]. fatalities due to liver toxicity (Figure 1). Thus, accurate and timely
Drug Induced Liver-Injury (DILI) is a rare, but serious, poten­ prediction of DILI is required for patient safety as well as R&D
tially life-threatening adverse drug reaction that is a significant productivity. Reasons for the inadequacies of the current testing
concern in preclinical drug development, clinical trials, and post- paradigm are many and varied [9], but failure to predict DILI is
market surveillance [7,8]. It is also the primary cause of candidate often attributed to its multifactorial nature. DILI can be categorized
drug attrition leading to regulatory activities such as safety com­ into three major categories: (1) intrinsic (predictable, reproducible,
munications, drug withdrawals, black-box warnings, and/or and dose-dependent), (2) idiosyncratic (unpredictable, not neces­
denial in approvals [8] (Figure 1). According to the 2023 sarily dose-dependent), and (3) indirect DILI. Indirect DILI is being
DILIrank dataset from the US Food and Drug Administration, of increasingly recognized as mechanistically associated with the
the 1036 FDA-approved drugs, 192 were classified as most DILI pharmacodynamic properties of the drug candidate, with cited
concerning, 278 as less DILI concerning, 312 with no DILI con­ examples involving the adaptive immune system [10]. Over 11
cern, and 254 as ambiguous DILI concern [7]. These data were potential mechanisms have been implicated in DILI, and clinical

CONTACT Sushma Jadalannagari [email protected] Department of Bioinnovations, Emulate Inc, 27 Drydock Avenue, Boston, MA
02210, USA
© 2024 Emulate Inc. Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/),
which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent.
608 S. JADALANNAGARI AND L. EWART

part of the regulatory requirements for first-in-human adminis­

Article highlights tration of a new drug candidate or within a weight of evidence
● Drug-induced liver injury is a potentially life-threatening, serious drug
approach supporting investigational new drug applications [16].
reaction, that is a leading cause of drug attrition resulting in clinical Species differences in metabolism and transport of, and sensitiv­
trial failure and put patient safety at risk. ity to, potential hepatotoxicants have also been identified, lead­
● Species differences from animal models and the lack of appropriate
human relevant, clinically translatable models in the pre-clinical test­
ing to further difficulties in translating data from animal models
ing phase of drug development plays a significant role in adequately used preclinically. An infamous example of this is the anti-viral
identifying drug candidates that carry a hepatotoxicity liability. drug Fialuridine. This drug led to the death of five out of 15
● Liver complex in vitro models are a group of assembled human tissue
constructs which include spheroids, bio printed tissues, microphysio­
clinical trial participants, and two of the remaining 10 required
logical systems to generate clinically translatable data and thus are liver transplantation [17]. Sadly, animals failed to alert the spon­
poised to predict DILI and its associated conditions more accurately. sor and regulators to the risk. Latterly, it was demonstrated that
● There are many complex in vitro models, but no single model can
efficiently emulate all the functions of the liver. Broader adoption by
species differences in the Equilibrative Nucleoside Transporter
end-users is limited by the small number of published papers show­ (ENT [18];) were responsible for the hepatotoxicity.
ing system reproducibility, reliability, and robustness. Whatever the classification, DILI is a complex process with
● Many industry-regulatory agency collaborations/strategic partner­
ships have been established to define guidelines and generate data
limited information, no uniform criteria for diagnosis, and
sets to build confidence in complex in vitro models. a lack of effective management processes. Hence, human-
relevant models can play a role in achieving a greater under­
standing of biological mechanisms, especially if they faithfully
recapitulate hepatotoxic pathways. Also, they may help
presentation has been associated with over 15 phenotypes [11]. develop protective or preventative therapies and eventually
Biochemically, DILI typically presents as three patterns; hepatocel­ aid in the clinical diagnosis of DILI.
lular, cholestatic, or mixed, yet clinicians only have four rudimental,
routine diagnostic biomarkers at their disposal: alanine transferase
2. The new device in the preclinical toolbox: liver
(ALT), aspartate transaminase (AST), alkaline phosphatase (ALK P)
complex in vitro models (CIVMs)
and bilirubin. These biomarkers may be compromised if the
patient also presents with a preexisting liver condition. Over the last 15 years, there has been a dramatic increase in the
Pharmaceutical companies today typically evaluate the safety use of in vitro models, particularly within the realm of drug dis­
and toxicity of drugs using low complexity in vitro two- covery and development. These models can be more predictive of
dimensional (2D) monoculture or co-culture models to balance human biology while aligning with 3Rs goals (replacement, reduc­
throughput and cost, enabling an early deployment in the drug tion, and refinement) of animal experimentation. The US FDA
discovery process [9]. However, hepatocytes in these models fail defines such models as microphysiological systems or MPS’— ‘A
to maintain long-term viability, lose function quickly, and lack microscale cell culture platform for in vitro modeling of functional
complex architecture as well as the associated zonation [12]. features of a specific tissue or organ of human or animal origin by
Recently, 3D spheroids employing human hepatocytes from exposing cells to a microenvironment that mimics the physiologi­
cadaver donors, human chimeric mice, human embryonic stem cal aspects important for their function or pathophysiological
cells, induced pluripotent stem cells (iPSC), or hepatocellular condition’ [19]. However, the term complex in vitro models
carcinoma cell lines have also been added to preclinical testing ‘(CIVM) (Figure 2) is rapidly gaining attention to differentiate MPS
strategies [13–15]. However, animal models such as rodents, from more conventional in vitro models. CIVMs have been defined
dogs, and non-human primates nevertheless predominantly as ‘systems having a multi-cellular environment within
serve as the essential tool in the study of liver toxicity, either as a biopolymer or tissue-derived matrix, a 3D structure potentially

Figure 1. Drugs in clinical hold over DILI concerns.

Since 2022, 9 drugs from pharmaceutical companies are in clinical hold due to DILI risk
 EXPERT OPINION ON DRUG METABOLISM & TOXICOLOGY 609

Induced Pluripotent CRISPR Emerging Technology Bioconvergence CIVMs Support

Stem Cell (Yamanaka) Described World Economic Forum Gains Recognition Regulatory Submission

1990 2010 2013 2018 2022

2006 2012 2016 2020 2023

Human Genome Lung-Chip Paper 3D Bioprinted Tissue IQ MPS Affiliate FDA Modernization
Project Initiated (Ingber) (Organovo) Formed Act Passed

Figure 2. A timeline of key milestones involved with shaping the development of CIVMs.

including two or more of the following: mechanical factors such as executed by specialized cells called hepatocytes. However,
stretch or perfusion (e.g. breathing, gut peristalsis, flow), incorpor­ these prominent roles also make the liver a target organ of
ating primary or stem cell-derived cells, and/or including immune toxicity, a phenomenon compounded by increased exposure to
system components’ [20]. Consequently, spheroids, organoids, toxicants delivered via the portal vein which drains from the
three-dimensional (3D) bio printed tissue, and Organ-Chips (single gastrointestinal tract. This is evidenced by hepatotoxicity remain­
or multi-organ) can all be considered a CIVM (Table 1) [21]. Either ing a leading cause of drug attrition, market withdrawal, and
way, the ingenious combination of biology and engineering has patient fatalities. By more closely modeling the in vivo microen­
created a second generation of in vitro models that are gradually vironment using human cells, CIVMs hold great promise to posi­
shifting the study of human biology [22]. By creating systems that tively disrupt the drug discovery and development process
strive to recreate in vivo microenvironments, incorporating crucial [20,26–28] and, therefore, it is not surprising that significant
elements such as accessibility to high quality human cells, cellular investment has been directed toward building liver CIVMs
organization, extracellular matrix scaffolds, intercellular communi­ [29,30] (Table 2).
cation, and mechanical stimuli, cells behave in a more in vivo-like To assist with the goal of i) building human-relevant
fashion and thus the resultant data should possess higher transla­ models of the liver that could be used to predict DILI,
tional relevance. The rapid expansion of this field has already been and ii) enhance understanding of the biological mechan­
extensively reviewed [23–25]; as such, this article will focus speci­ isms that drive the pathophysiology, the pharmaceutical
fically on liver CIVMs as well as the opportunities and challenges industry worked within the Innovation and Quality
for their application to DILI. Consortia’s IQ MPS affiliate to produce a paper describing
The liver is the largest solid organ in the body, which receives development of liver CIVMs [53]. The authors noted the
approximately 75% of its blood flow from the portal vein. It explosion of liver CIVMs but wanted to harmonize the
performs nearly 500 separate functions, including metabolism, characterization of baseline performance to accelerate
digestion, and detoxification. These functions are chiefly their use as favorable tools for drug discovery and

Table 1. A table highlighting commercial liver CIVMs; created with BioRender.com and adapted from dash and Proctor et al. [28].
Traditional liver 2D Models Liver complex in vitro models (CIVMs)
Standard Micropatterned co-culture 3D Microtissues – 3D Bio printed tissues Microfluidic systems
Hepatocyte systems spheroids, organoids
culture in 2D

Static/Flow Static Static Static/Flow Static/Flow Flow

Cell Types Used Primary Primary Hepatocytes, iPSC Primary or pooled Primary Hepatocytes, Primary Hepatocytes, HepaRG,
Hepatocytes, derived hepatocytes, Hepatocytes and Non- HepG2 and Non- HepG2, iHeps and Non-
HepaRG, stromal cells Parenchymal Cells Parenchymal Cells Parenchymal Cells
HepG2
Examples of Lonza, LifeNet, BioIVT micro-patterned In Sphero 3D Liver Molecular Devices 3D Emulate Human Liver-Chip,
Commercially Gibco hepatic island plates microtissue plate bioprinted tissues, Mimetas Liver MPS
available models (HEPATOPAC) ™
(InSight hLiMTs) Organovo 3D bioprinted
liver tissue
®
(OrganoPlate ),
CN BIO Organ-Chip platform

Cost $ $$ $$
(exVive3D )
$$$
™
$$$
®
(PhysioMimix )
 610

Table 2. A detailed summary of commercial liver CIVMs including the cell types used, through-put, treatment duration, selected readouts and drugs tested for DILI.
Treatment
Platform Perfusion Throughput Cells duration Select readouts Hepatotoxicity
Micropatterned Co- BioIVT micro-patterned None 24 or 96 well Primary hepatocytes (human or 28 days Albumin, Urea, CYP activity, Three-point concentration curve for 93 drugs
Culture Models hepatocyte cultures plates rat) or pooled hepatocytes, Bile canaliculi
(HEPATOPAC) [31] stromal/
fibroblasts
Microtissues without Primary human hepatocyte None 96 well plates Primary Human Hepatocytes 35 days Albumin, CYP activity, ATP Eight-point concentration-response curve to five
Flow (Spheroids, spheroids [32] content compounds
Organoids) 3D human liver None 96 well plates Primary Human Hepatocytes and 14 days ATP content Eight-point concentration response curve for
microtissues (hLiMT) [14] Non-parenchymal cells 110 drugs
3D Spheroids [15] None 96 well plates Primary Human Hepatocytes 14 days ATP content Three-point concentration response for 123
drugs
In Sphero Liver Micro No fluidic flow 96 well plate Primary Human Hepatocytes, 28 days Albumin, viability, IF and BF Five-to-seven-point concentration response
Tissues [33] Kupffer cells and liver sinusoidal imaging, curve for 28 compounds [33]
S. JADALANNAGARI AND L. EWART

endothelial cells LC-MS Five-point concentration response curve for

three compounds [34]
HUREL Micro Livers [35–37] None 6–384 well Primary Human Hepatocytes and 14 days LCMS, LDH, ATP, CTG One concentration for four compounds [37],
format Non-parenchymal cells Two concentrations for three compounds [35],
Six-point concentration response curve for 29
compounds [36]
3D Bio printed Bio printed hepatic Syringe pump with 100 X 100 HepG2/C3A 30 days Albumin, Ceruloplasmin, Single concentration of Acetaminophen
Tissues spheroids [38] recirculation Microwells alpha-1 antitrypsin,
per mold Transferrin
Organovo Bio printed None 3D bio printed Primary Human Hepatocytes, 28 days Albumin, ALT, gene Three-point concentration response curve for
Human Liver Tissue liver in 24- stellates, endothelial cells expression, histology eight drugs
(ExViveTM) [39,40] well
transwell
Liver Micro Mimetas Liver MPS Gravity flow with 96 Chips iHeps, HMEC-1 Endothelial cells, 15 days LDH, Albumin, Urea, Alpha Seven-point concentration-response curve to 21
physiological (OrganoPlate 2-lane)
® reciprocation THP-1 cells Fetoprotein, CYP activity compounds
Systems with Flow [41]
TissUse Primary human Micropump with 96 well liver Primary Human Hepatocytes and 7 days LDH, Albumin, Urea, CYP Two concentrations of two structurally related
hepatocyte spheroids recirculation plates/chips Non-parenchymal cells activity compounds
[42]
Emulate Human Liver Pressure driven, 12 Chips Primary Human Hepatocytes and 13 days Albumin, Urea, LDH, IF Two concentrations of eight compounds
Emulation System single pass flow liver sinusoidal endothelial cells imaging
[43,44]
Emulate Human Liver- Pressure driven, 12 Chips Primary Human Hepatocytes and 13 days Albumin, Urea, ALT Eight-point concentration-response curve to 27
Emulation System [45] single pass flow Non-parenchymal cells compounds
Emulate Human Liver- Pressure driven, 12 Chips Human liver organoids, 13 days Albumin, ALT, AST, One to two concentrations for six compounds
Emulation System [46] single pass flow Transcriptomics
CN Bio Hepatocyte-Kupffer Micropump with 1 Liver MPS Primary Human Hepatocytes and 12 days Albumin, LDH, CYP activity Six-point concentration-response curve to three
cell Organ Chip platform recirculation plate has 12 Kupffer cells compounds
[47] wells
Zonated, vascularized liver Syringe pump with Unknown PHH, LSEC, Stellate and Kupffer cell 14 days Albumin, LDH, protein Two concentrations for one compound
acinus vLAMPS [48,49] recirculation, lines analysis, metabolite
analysis, qPCR,
Javelin Bio Liver Tissue High flow Milli-fluidic PHH only or a co-culture available 12–15 Albumin, gene expression, One concentration for six compounds
Chip [50] (2 mL/h) chip days metabolism,
LDH, urea
Lena Biosciences Perfusion Pal – 12-well or 48- HepG2 and PHH 26 days Viability, Alamar Blue, LDH, Nine-point concentration response curve for 16
SeedEZ 3D culture scaffold, Syringe pump- well format CYP activity compounds
Perfused organ panel [51] based flow
Draper Labs - Microfluidic pump- 96 well plate Primary Human Hepatocytes 10 days Albumin, CYP enzymes, VEGF, No published data on DILI to date
PREDICT 96 [52] based flow Cytokines (IL-8, IL-6, TNF-α,
and TGF-β)
 EXPERT OPINION ON DRUG METABOLISM & TOXICOLOGY 611

development. Consequently, the authors formulated organ homeostasis, non-parenchymal cells (NPCs) such as
a three-stage framework to guide such model develop­ endothelial cells, Kupffer cells, and stellate cells should be
ment. Stage one focused on baseline characterization included. To add to the complexity, circulating immune cells
including measurement of albumin (>37 µg/day/106 hepa­ are likely also needed to fully emulate the human in vivo
tocytes) and urea production (>56 µg/day/106 hepatocytes) microenvironment. Thankfully, CIVMs are commonly regarded
as well as gene expression measurements for drug trans­ to be agonistic to cell type or source.
porters and enzymes involved in phase I and II drug meta­
bolism. Stage two described the assessment of parameters
3. Driving adoption: “Don’t let perfection be the
to measure model stability over time or said differently,
enemy of good”
a functional characterization. For stage three, the authors
proposed a 20-compound test set to enable a sensitivity While there is broad agreement that CIVMs are poised to
and specificity determination. The 20 compounds were tackle the challenge of translating preclinical data into clinical
mainly associated with clinical data and were proposed outcomes, like all tools, they have shortcomings that require
on the basis that the major known mechanisms of DILI continued investment. Therefore, it will be important to iden­
were represented while covering a range of DILI presenta­ tify when a model is ‘fit-for-purpose’ to address the question
tions from ‘severe’ to ‘no DILI concern.’ Interestingly, this a researcher is seeking to answer. The following section
compound test set included seven matched pairs. These describes published commercial liver CIVM model develop­
pairs were structural analogues, where clinical data indi­ ment in the context of DILI (Table 2) and highlights areas
cated that one member of the pair resulted in a more toxic where further work is required to address the expanding, yet
outcome compared to the other. By including this subset discrete, contexts of use.
of compounds, the industry affiliate was testing how well With a wide variety of researchers from academic institutions
a model could rank order compounds with the same as well as start-up, biotechnology, and biopharmaceutical com­
chemistry, a popular practice in the lead optimization panies working with an assortment of materials and cell types,
phase of drug discovery and development. the CIVM landscape has become challenging for end-users to
Researchers would agree that, in addition to this framework, identify the best tool for their needs, with some reporting a gap
the choice of cell within a model is important, as not all cells are between development, reproducibility, standardization, and
equal. Because hepatocytes are widely studied, it is known that qualification of these in vitro models [20,21,58,59]. To facilitate
primary human hepatocytes (PHH) have a short culture life gap-closing efforts, the IQ MPS Affiliate was formed in 2018 to
span, possess high phenotypic variability, are typically unstable support the implementation of CIVMs within drug discovery
in culture and lose functionality quickly [54]. On the other hand, [60,61]. There have also been workshops with the regulatory
a human immortalized cell line such as the HepaRG cell is community to drive education and awareness of regulatory
widely considered a close surrogate to PHHs with stable cyto­ qualification and regulatory community – led publications
chrome P450 expression; however, they lose their function [43,44,47,62]. These are important first steps that bring key sta­
under conventional static 2D conditions and have different keholders together, but continued, focused efforts are needed
characteristics from the human liver [55]. iPSCs have the poten­ for physiological and functional standardization of CIVMs to drive
tial to expand indefinitely and, with the appropriate growth industrialization and enable regulatory acceptance for DILI pre­
factors and signaling molecules, they can differentiate into dictions within defined contexts of use. An insight into progress
derivatives of all three primary germ layers (ectoderm, meso­ of this journey was recently reported by the IQ MPS Affiliate who
derm, and endoderm). However, while these cells are amenable surveyed their members in 2019 and again in 2021 on their
to genetic modulation and can differentiate into multiple hepa­ experience using preclinical data from CIVMs to support clinical
tic cell types from the same genetic background, iPSC hepato­ development. Their report also provided key longitudinal infor­
cytes are often considered immature due to the expression of mation on compound modalities tested, cell types used, non-
alpha fetoprotein (ALF) [56]. Thus, while PHHs are gold standard clinical species needed, organs of interest, as well as human
for predictive toxicology studies, they are limited in availability resources deployed. The survey findings ultimately showed
and have high inter-donor variability and unit costs, while iPSC CIVMs are in a strong position to improve drug development
derived hepatocytes can be expanded and produced in large decisions but emphasized the lack of confidence due to the
quantities. Thus, NIH (National Institute of Health) strongly limited data sets that characterize and qualify CIVMs [63].
encouraged using iPSCs and provided significant funding in Table 2 provides valuable insights into the current literature
this area, however, the differentiation protocols are new and available on CIVMs in DILI prediction and highlights the
warrant additional optimization to make them fully functional numerous readouts available from these platforms. While
[57]. In addition, in 2011–12 DARPA (Defense Advanced albumin is a relevant readout for baseline hepatocyte function
Research Projects Agency), FDA, and NIH also funded up to from these models, other endpoints such as ATP content and
$70 million dollars for five years to develop CIVMs to help LDH, which are generic to cell health, may also be informative.
predict drug safety, which in turn boosted the cell sourcing Most models also used urea and CYP activity as functional
industry by increasing the need for quality and quantity of readouts and very few measured clinical liver biomarkers such
human cells needed for these models. as ALT and AST. The duration for compound exposure and
Despite hepatocytes largely being responsible for perform­ testing also varied from 7 to 35 days, thus making the choices
ing the multiple functions of the liver, for greater mimicry of for end-users very difficult. Hence this table highlights the
612 S. JADALANNAGARI AND L. EWART

need for standardization of cell types, duration of treatment, troglitazone, compared to sandwich culture. Furthermore,
effluent readouts, and terminal endpoints across these models hepatocytes in the microfluidic model were functionally more
to drive a consensus for their use in defined contexts of use. stable with increased CYP3A4 activity and albumin secretion for
Amongst the data captured on Table 2, it is also worth high­ greater than 18 days while sandwich cultures and spheroids
lighting CIVM performance against clinically known hepatotoxins. lasted only 7- and 12 -days respectively. However, this work
Firstly, Bell et al., showed that 3D PHH spheroids grown in only used one microfluidic platform with PHH in combination
a serum free media retained their morphology and hepatocyte with Kupffer cells thus lacking cellular interactions with liver
specific functions for up to 5 weeks [32]. In addition, these specific endothelial cells and stellate cells which are essential to
spheroids were able to detect toxicity of 5 hepatotoxins at fully mimic a functional liver sinusoid. While evaluating one
clinically relevant concentrations. Similarly, Proctor et al., com­ microfluidic platform with other 2D/3D platforms is the first
pared 3D liver microtissues of PHH and NPCs with 2D PHH for step in comparing performance and sensitivities of different
110 drugs and demonstrated liver microtissues had increased platforms, there are several other comprehensive liver CIVMs
sensitivity to detect DILI while both platforms had similar speci­ such as quad-culture liver chips or liver chips with immune cells
ficity [14]. More recently, Ewart et al. tested 27 small molecule that also need to be contrasted to obtain a better understand­
drugs in a human Liver-Chip containing primary hepatocytes, ing of the reproducibility between the CIVMs.
Kupffer cells, stellate cells, and liver sinusoidal endothelial cells As a further effort to address the need for ‘validation’ of
[45]. This study highlighted that Liver-Chip had a sensitivity of CIVMs, in 2016, the National Centre for Translational Sciences
87% and a specificity of 100%, when using two different hepa­ (NCATS) provided funding to initiate two Tissue-Chip Testing
tocyte donors, indicating value to pharmaceutical scientists inter­ Centers (TCTC)-one at Texas A&M University and the other at
ested in predicting DILI liability ahead of progressing to a clinical Massachusetts Institute of Technology – to conduct indepen­
trial. The 100% specificity drives confidence that the model does dent testing of platforms through collaboration with develo­
not identify false positives. Finally, this work also demonstrated pers, regulators, and pharmaceutical researchers. Thus, to
that the model could discriminate toxic liability between struc­ show robustness, demonstrate reproducibility, and increase
tural analogues by correctly identifying the most toxic drug end-user adoption, Lim et al., compared PHH- and iPSC-
across seven structurally analogous pairs. Such a capability can derived hepatocytes in combination with and without NPCs
offer value in the lead optimization phase of drug discovery on a microfluidic platform (PhysioMimix™ LC12), 96- and 384-
where pharmaceutical researchers are required to rank order well plates, LAMPS (Liver Acinus MPS), and OrganoPlate plat­
chemically similar analogues to progress to the pivotal good forms [65]. Comparisons showed that PHHs with or without
laboratory practice (GLP) in vivo studies ahead of regulatory THP-1 or Kupffer cells demonstrated the highest metabolic
submission. Nevertheless, this Liver-Chip is made from PDMS function, while iPSC-derived hepatocytes or PHHs co-cultured
which binds the hydrophobic molecules to their exposed sur­ with NPCs on the microfluidic platform (PhysioMimix™ LC12),
faces. Thus, special protocols such as cell-free PDMS organ-chips showed similar urea production, but albumin and CYP3A4
or compound distribution kits are proposed to help design levels were lower, indicating a suboptimal performance on
studies that account for potential compound loss due to absorp­ the microfluidic platform While this study compared different
tion or adsorption [64]. MPS platforms and 2D plates, high inter-experimental varia­
Within the last three years, there has been a small rise in bility was reported. This was attributed to technical difficulties
publications designed to build confidence in CIVM data, includ­ in creating a uniform cell seeding across the different plat­
ing collaborations between model developers and pharma forms. The authors thus concluded that there was a need for
researchers or model developers and regulatory agencies higher replicates per condition but recognized that this would
[43,45,47,62]. These studies have been designed to evaluate increase the experimental costs and limit the throughput if
the functionality, robustness, reliability, and reproducibility of used in a drug screening mode. In another study, Kato et al.
liver CIVMs because ultimately end-users need to trust that the compared iHeps and PHH in combination with or without
platform and model consistently performs and that it can per­ NPCs (endothelial cells and THP-1 cells differentiated into
form without failure while producing the same result in the macrophages) on the OrganoPlate® 2-lane platform to char­
same experiment, ideally across different laboratories. To eval­ acterize its utility and reproducibility [66]. On this platform,
uate the reproducibility of predicting trovafloxacin and levo­ when albumin, urea secretion, CYP3A4 activity and drug meta­
floxacin hepatotoxicity, Rubiano et al., compared hepatocytes in bolism were assessed, iHeps and NPCs demonstrated stable
the microfluidic platform (PhysioMimix™ LC12) with two differ­ cell viability and function for 17 days, along with efficient drug
ent Kupffer cell lots and observed no significant changes in metabolism. However, to deliver robust results with low varia­
Lactate dehydrogenase (LDH) and CYP 3A4 activity [47]. They bility, they needed a high number of replicates, thus limiting
also evaluated the same toxicity experimental setup with one throughput.
hepatocyte and Kupffer lot at two independent sites and Developers also need to consider the depth to which mod­
observed similar toxicity with trovafloxacin at both the sites. els recreate the diversity of different tissues and cell types that
In another experiment evaluating platform performances, make up an organ such as the Liver. For example, within any
Rubiano et al., compared PHH on fluidic devices, spheroids, one tissue, there are hundreds of different cell types at various
and sandwich cultures for their toxicity to troglitazone, tamox­ stages of development, and these likely contribute to physio­
ifen and digoxin, and observed liver MPS and spheroids demon­ logical and pathophysiological responses. Tissue organization
strated increased sensitivity to tamoxifen and digoxin, but not can also influence cellular environment, oxygen levels, rate of
 EXPERT OPINION ON DRUG METABOLISM & TOXICOLOGY 613

nutrient supply, and intercellular communication; thus, they better at predicting DILI outcomes; however, the behavior and
must also be considered in design. Furthermore, to achieve manipulation of small fluid volumes have their limitations, such
clinical mimicry, the ways in which cells and tissues interact as the presence of surface effects dominating the volume effects
locally, and with the vasculature, nervous system, immune [76], the formation of bubbles, and flow obstructions. These
system, and germ cell development be reproduced; hence, could, in turn, affect the data quality, complicating study inter­
creating a CIVM of a single organ may be insufficient. pretation. The utmost care and attention to detail are needed
Consequently, there has also been an expansion of models when working with CIVMs to maximize the benefits and avoid
that couple a liver CIVM with other organs of interest, includ­ flow-related issues.
ing the kidney and/or intestine to study absorption, distribu­
tion, metabolism, and excretion [67,68]; the heart to detect
metabolite-induced toxicity [69]; and with heart, bone, and
4. The next frontier: what does the future hold?
skin to recapitulate the pharmacodynamics and pharmacoki­ With the development and application of CIVMs now in
netics of the anti-cancer drug doxorubicin [70]. their second decade, it is fitting to review what the future
Encouragingly, in a recent survey of pharmaceutical com­ holds for these models in DILI.
panies who are utilizing CIVMs, most respondents reported Considerable effort has been invested to build in silico
throughput to be satisfactory with 1–12 replicates or chips models from in vitro hepatotoxicity tests, but their predictive
while some favored 12–24 or 24–48 replicates or chips as the value is often limited [77]. Physiochemical properties, includ­
experimental size [63]. But the need for the number of repli­ ing cLogP of drugs, dose, disposition, and patient character­
cates varies per user, thus providing some flexibility in choos­ istics such as genetics or disease status affect the likelihood of
ing the sizes of experimental groups. Nonetheless, seeding DILI occurrence and its severity [9,78]. Risk factors include
more Chips or scaffolds are complex as it requires significant reactive metabolite formation [79], mitochondrial impairment
experimental planning and expertise to get uniform seeding [79] and inhibition of canalicular drug transporters such as the
density across all the replicates. Also, the high demand for the bile salt export pump (BSEP) [80]. Such knowledge has been
volume of cells to be used in larger study sizes and the translated into large-scale in vitro screening activities within
limitations of air bubbles restrict the number of replicates lead identification of drug discovery workflows, typically those
that can be realistically handled by one user which can hope­ involving cell lines or membrane vesicle assays. However, this
fully be balanced with automation and high throughput. approach does not factor in the downstream consequences of
Table 2 provides a detailed view of published studies on these pathways. For example, the hepatocyte drug transporter
commercial platforms and highlights that the throughput multidrug resistance protein 4 (MRP4) located on the basolat­
approach favored focuses on ‘chips’ in factors of 12. eral membrane, can be upregulated when BSEP is inhibited. In
Beyond biological considerations, device materials and a recent study, it was found that many drugs that inhibit BSEP
design are critical. Polydimethylsiloxane (PDMS) is a nontoxic, also inhibit MRP4, leading to a greater cholestatic risk for that
biocompatible, gas-permeable, and optically transparent mate­ drug [81]. Consequently, to build improved models for DILI
rial that is commonly used, while glass, elastomers, thermoplas­ prediction, it is important to ensure test systems can recreate
tic polymers, or hybrid combinations of PDMS-polycarbonate in vivo pathways, ideally capturing mechanism of action. This
are also used [52,71]. However, PDMS has strong absorption of in part is addressed by quantitative systems pharmacology
hydrophobic molecules, which could lead limit the compound (QSP), a discipline that uses computational methods to gen­
bioavailability and make it difficult to detecting toxicity [71,72]. erate an understanding of the dynamic interaction between
Auner et al., evaluated nineteen chemicals in two experimental drug and tissue at the cellular and biochemical level [82,83].
setups with different PDMS-surface-to-surface-volume ratios Separately, Williams et al. [84], combined Bayesian machine
and showed that PDMS binding is strongest in chemicals with learning with data on BSEP inhibition, mitochondrial toxicity,
a high octanol-water partition coefficient (log p > 1.85) and low cLogP and exposure (Cmax) to assess the predictive value of
H-bond donor number [73]. Thus, PDMS binding could change spheroids for DILI across a set of 96 drugs. They found that the
the compound exposed to the cells in the MPS systems and model delivered a sensitivity of 87%, a specificity of 85%,
developers need to ensure manufacturing quality to drive con­ a positive predictive value of 92%, and a negative predictive
fidence in product and, hence, utilization. However, as already value of 78% [84]. This approach indicated there is potential to
highlighted, there is a plethora of devices available, so calls for improve DILI risk prediction with human-relevant models and
standardization on devices are growing [74,75]. could likewise be applied to other characterized liver CIVM
Experimental results from traditional 2D and 3D liver cell models.
cultures have been used for decades in predicting toxicology. Deep learning models are increasingly widespread within
These conventional cultures are relatively easy to maintain, with drug discovery and development workflows primarily to con­
no fluidic flow. However, complex liver flow-based models (also firm drug candidate safety or efficacy at an earlier juncture of
highlighted in Table 2) have dynamic fluid flowing over the cells the drug discovery process. But, as already highlighted, DILI is
and tissues, thus, potentially exposing cells to nutrients while a multifactorial disorder where contributing mechanisms or
removing waste, establishing gradient flow, and enabling meta­ pathways are still not fully understood. Wang and coauthors
bolic zonation. This mimics the physiological microenvironment, have just published a model that claims enhanced predictive
as cells in vivo are exposed to a gradient of oxygen and hor­ capability by going beyond compound physicochemical and
mones. Therefore, research shows that liver CIVMs are generally structural attributes to discover mechanism-dependent and
614 S. JADALANNAGARI AND L. EWART

independent factors that provide deeper insights into the therapeutics, it is anticipated that further examples of immune-
complex dynamics of hepatotoxicity [67]. They used publicly mediated DILI will be seen, if not during clinical trial, then in the
available data to create detailed molecular fingerprints of post-marketing surveillance period. For example, it is widely
drugs labeled as DILI-positive or DILI-negative. Like all models, known that anti-TNF agents carry a higher risk of DILI with
the authors acknowledged that their model would benefit mechanisms that involve CD8+ T cells [90]. Thus, building
from further optimization, and as the computational powers human-relevant models that are immunocompetent would
continue to increase in parallel with CIVM capabilities, earlier facilitate the discovery and development of safer biological
and more accurate DILI prediction is likely to become a real medicines, especially as more of these are approved by
possibility, leading to reduced R&D costs and greater produc­ regulators.
tivity. Indeed, it is now widely recognized that multidisciplin­ Adaptive DILI can be described as a transient elevation of
ary approaches are required to tackle unresolved challenges. the transaminases ALT and AST which can spontaneously
Hence, the term ‘bioconvergence’ that emerged in 2005 but resolve (even without drug withdrawal) and is often consid­
gained significant recognition in 2020 for potential application ered as having no clinical significance. One hypothesis to
within life sciences. Bioconvergence speaks to the combina­ explain such observations is that liver regeneration is trig­
tion of biotechnology, engineering, and computerized systems gered by a mild injury resulting in newly grown cells which
that in this context can advance drug discovery and develop­ are more resistant to drug treatment. This hypothesis could be
ment. The amalgamation of highly differentiated, iPSC-derived explored further with microfluidic CIVMs that can control fluid
tissues creating multi-cellular human relevant CIVMs with clin­ flow, circulating cytokines, and/or paracrine interactions
ical mimicry combined with noninvasive imaging, multi-omics, between hepatocytes and the vasculature cells [91].
conceptual frameworks such as adverse outcome pathways Although current CIVMs may not be able to distinguish
(AOP), and computerized systems will undoubtedly be increas­ between adaptive and adverse responses, going forward
ingly commonplace before the end of this decade and may they could be to determine if regeneration can be triggered
finally lead to better prediction, diagnosis, and treatment of by a drug. When coupled with iPSC hepatocytes, drug devel­
conditions such as DILI. opers may be able to evaluate patient susceptibility to drugs
Cholestasis describes a condition where secretion of bile causing adverse or adaptive DILI.
from the liver is impaired. Consequently, bile acids accumu­ Metabolic dysfunction-associated steatotic liver disease
late within the liver and chronic conditions present clinically (MASLD, formerly known as Non-Alcoholic Fatty Liver
with jaundice. The mechanisms leading to cholestasis go Disease or NAFLD) has an estimated global incidence of 47
beyond inhibition of the BSEP transporter and are believed out of every 1000 people. The global prevalence has
to involve the enzymes controlling conjugation or sulfation increased from 26% to 38% within the last decade [92].
[68] of bile acids, nuclear receptors [85], and other drug Clinical reports suggest that obesity or MASLD may make
transporters [80]. Moreover, symptoms often occur after sev­ individuals more susceptible to DILI are rising [93] and linked
eral weeks or months; thus, longevity of cellular function to more frequent or severe reactions to drugs such as acet­
within a model will be key to helping predict drug-induced aminophen and tamoxifen. Mechanisms for increased hepa­
cholestasis. With more than 1000 drugs known to cause totoxicity in patients with liver disease are not well
cholestasis [86], drug hunters are eager to find improved understood, although it has been postulated that increased
ways of detecting this risk early in discovery. Hendriks et al. CYP2E1 coupled with decreased CYP3A4 activity play
(2016) have shown that 3D hepatic spheroids cultured with a critical role [94]. Additionally, mitochondrial dysfunction,
a nontoxic concentrated bile acid mixture for up to eight ER stress, and overproduction of reactive oxygen species
days demonstrated synergistic toxicity with troglitazone or may aggravate preexisting liver conditions [95]. Animal mod­
bosentan [87]. Thus, CIVMs can provide an additional tool for els of MASLD are insufficient, as they often show different
detection of drugs posing a cholestatic risk. Further enhance­ patterns of CYP regulation compared to humans [96]; there­
ments to CIVMs could also include separate channels for fore, liver CIVMs could be used to test for a change in toxicity
collecting bile and potentially the co-culture of hepatocytes frequency and/or severity as well as play a role in decipher­
with cholangiocytes. ing mechanisms of toxicity in diseased tissues, assuming they
Idiosyncratic DILI (iDILI) can be immune-mediated or non­ can correctly mirror the CYP patterns. Separate diagnosis of
immune mediated. By its very nature, iDILI is unpredictable, but DILI in patients with MASLD or metabolic associated steato­
with the emergence of CIVMs and the ability to recreate more hepatitis (MASH) is particularly difficult because diagnosing
complex cellular structures, biological mechanisms may be DILI relies on the same rudimentary biomarkers (e.g. transa­
unraveled, ultimately contributing to better drug candidates minases) as those which are often elevated in patients with
and/or patient outcomes. Specific human leukocyte antigen liver disease. Additionally, DILI can mimic other liver diseases
(HLA) haplotypes are known to be a risk factor for iDILI [88]; such as hepatitis. Therefore, liver CIVMs could improve this
therefore, when constructing a liver CIVM, it would be prudent situation by facilitating a greater understanding of cellular
to HLA type the donor cells, especially the hepatocytes. The disease targets and biology and, when coupled with proteo­
most prevalent form of iDILI is associated with an infiltration of mic or metabolomic approaches, may lead to the identifica­
CD8+ T cells. Microfluidic CIVMs have been shown to enable tion of new biomarkers for DILI. In the short term, researchers
immune cell recruitment [89] and thus offer a clear advantage working with liver CIVMs focus measurement of albumin or
for modeling iDILI. With the ascendance of biological ALT and AST, but urea is a good marker of mitochondrial
 EXPERT OPINION ON DRUG METABOLISM & TOXICOLOGY 615

function and has been measured in liver CIVMs [45,62,97] some conjugated to bile acids, yet no CIVM design has a bile
while miR122, which signals hepatocyte damage, has also canalicular structure which allows hepatobiliary drug excretion
been measured in liver CIVMs [98,99]. Other potential mar­ or assists in long-term liver maintenance. Furthermore, there is
kers that could be developed include detection of autoanti­ a lack of uniformity in cell types, device materials, validated
bodies for autoimmunity, identification of damage-associated endpoints, biomarkers for detection, and model set up, mak­
molecular patterns (DAMPs) and macrophage phenotypes as ing the comparability of the data complicated, potentially
biomarkers of immune stimulation, and glycochenodeoxy­ limiting end-user confidence. It is therefore imperative to
cholic acid (GCDCA) as a marker of cholestasis. simplify the platform and model landscape to drive utilization
The regulatory community is keenly watching the pro­ and adoption.
mise of CIVMs, as these models could improve the ability to Industry Consortia such as the IQ MPS affiliate, FDA partner­
predict DILI risk of a new drug, thus assisting with bringing ship with the Critical Path Institute, and regulatory pilot pro­
safer, more efficacious products to market faster. But multi­ grams like ISTAND have now been set up to improve and
ple steps are required to translate these new technologies qualify CIVMs for drug development as well as identify gaps.
into regulatory use, a process known as qualification. This has led to a community approach where pharmaceutical
Simply put, qualification is the conclusion of a process companies, regulatory agencies, academicians, and model
whereby a tool, within a stated context of use, adds value developers are working closely together and communicating
to data interpretation and thus benefits the drug develop­ regularly to characterize and qualify the models and reach
ment process. Once qualified, it can be used within regula­ a common agreement, so all the data is harmonized. Several
tory submissions without the regulator needing to assess workshops, surveys, and meetings have already been con­
the tool’s suitability. To enable CIVMs to qualify as ducted to encourage the use of CIVMs and discuss the chal­
a regulatory tool, the US FDA initiated the Innovative lenges in limiting their adoption. There are considerable
Science and Technology Approaches for New Drugs knowledge gaps and differences in terminology, so standardi­
(ISTAND) program [100]. Qualification of CIVMs is widely zation of nomenclature and process for characterization, qua­
expected to drive greater confidence by the end-user as lification, reliability, robustness, and reproducibility of CIVMs is
the process includes an assessment on reproducibility, needed. More published work sharing outcomes of collabora­
robustness, and reliability. tions or strategic partnerships between the regulatory agen­
In conclusion, over the last 30 years, there have been sig­ cies, model developers, drug developers, and academicians
nificant scientific advances that have enabled the develop­ are urgently sought to drive model characterization and qua­
ment of CIVMs (Figure 2). Specifically for liver, there are lification, improving data robustness and reliability while
many publications that highlight promising platforms and, building confidence. To reduce inter-laboratory and intra-
more recently, publications showing application toward pre­ laboratory variability and improve reproducibility between
dicting DILI. For the true value of CIVMS to be recognized, the studies, there could also be futuristic strategic partnerships
community of end-users, developers, regulators, and aca­ where the same studies are conducted at multiple sites and
demics need to continue collaborating to build models that data is shared without any barriers. The role of large-contract
are fit-for-purpose, have an appropriate cost and throughput, research organizations is strongly encouraged here. It is also
and – importantly – carry a high predictive value. Together possible for the model developers to offer certified learning
with AI technologies, the game-changing effect on the drug programs to harmonize training, leading to more unified and
discovery and development process is approaching reality, comparable experimental outcomes.
and the measure will be an increase in safer novel therapeu­ Until we build confidence in the CIVMs to predict candidate
tics that reach patients who desperately need them. drug safety with 100% specificity and sensitivity, a weight of
evidence-based approach using all the available in vitro and
in vivo data, including data generated using in silico
approaches such as Artificial Intelligence/Machine Learning
5. Expert opinion
(AI/ML), should be used to determine the overall risks asso­
There is no doubt that translation of data from preclinical ciated with any safety signals. The foundation for the accep­
studies into clinical practice remains a considerable burden tance of CIVMs by drug developers has been positioned with
to the pharmaceutical industry. This burden is measured by the FDA Modernization Act in 2022 and as 2023 closed, it was
poor success rates at each phase of the clinical trial process, announced that IniPharm used the PhysioMimix platform to
increased research budgets, and number of new drug generate efficacy data that was included within the regulatory
approvals by the US FDA. Drug attrition due to safety is submission for the candidate drug INI-822. This could be
a huge liability for drug developers, which in part is due to a significant turning point that recognizes the technology’s
the lack of clinically translatable in vitro and in vivo models. maturation and paves the way for CIVMs to be used routinely
CIVMs are well positioned to support the 3 R’s, improve toxi­ in drug discovery.
city and dose predictions, and aid with mechanistic investiga­
tions, thus transforming drug safety assessment while being
safer, faster, human-relevant, and translatable. Although Abbreviations
a variety of liver CIVMs exist, they only simulate partial organ
structure and function compared to a real liver. For example, DILI Drug Induced Liver-Injury
many drugs and drug metabolites are excreted in the bile, CIVMs Complex in vitro Models
616 S. JADALANNAGARI AND L. EWART

2D Two Dimensional ORCID

3D Three Dimensional
FDA Food and Drug Administration Sushma Jadalannagari http://orcid.org/0000-0002-6136-1059
Lorna Ewart http://orcid.org/0000-0002-8258-7460
R&D Research and Development
ALT Alanine Transaminase
AST Aspartate Transaminase
References
ALK P Alkaline Phosphatase
iPSc induce Pluripotent Stem Cells Papers of special note have been highlighted as either of interest (•)
ENT Equilabrative Nucleoside Transporter or of considerable interest (••) to the readers.
3Rs replacement, reduction, refinement 1. Senior M. Fresh from the biotech pipeline: fewer approvals, but
MPS MicroPhysiological Systems biologics gain share. Nat Biotechnol. 2023 Feb;41(2):174–82. doi:
IQ Innovation and Quality Consortia 10.1038/s41587-022-01630-6
PHH Primary Human Hepatocytes 2. Mullard A. 2023 FDA approvals. Nat Rev Drug Discov. 2024 [2024
Feb];23(2):88–95. doi: 10.1038/d41573-024-00001-x
ALF Alpha Fetal Protein
3. Buntz B. The big spenders: the top 25 pharma companies in R&D
NIH National Institute of Health investments: drug discovery & development. 2023. Available from:
NPC Non parenchymal cells https://www.drugdiscoverytrends.com/top-pharma-rd-spenders-2022/
LDH Lactate dehydrogenase 4. Scannell JW, Blanckley A, Boldon H, et al. Diagnosing the decline in
GLP Good lab practice pharmaceutical R&D efficiency. Nat Rev Drug Discov.2012 Mar 1;11
NCATS National center for Translational Sciences (3):191–200. doi: 10.1038/nrd3681
TCTC Tissue-chip Testing Centers 5. Ringel MS, Scannell JW, Baedeker M, et al. Breaking Eroom’s Law.
PDMS Polydimethylasiloxane Nat Rev Drug Discov. 2020 Dec;19(12):833–4. doi: 10.1038/d41573-
BSEP Bile Salt Export Pump 020-00059-3
MRP4 Multi Drug Resistance Protein 4 6. Scannell JW, Bosley J, Hickman JA, et al. Predictive validity in drug
discovery: what it is, why it matters and how to improve it. Nat Rev
QSP Quantitative System Pharmacology
Drug Discov. 2022 Dec;21(12):915–31. doi: 10.1038/s41573-022-
AOP Adverse Outcome Pathway 00552-x
iDILI Idiosyncratic DILI 7. Chen M, Suzuki A, Thakkar S, et al. Dilirank: the largest reference
HLA Human Leukocyte Antigen drug list ranked by the risk for developing drug-induced liver injury
MASLD Metabolic Dysfunction-Associated Steatotic Liver in humans. Drug Discov Today. 2016 Apr;21(4):648–53. doi: 10.
Disease 1016/j.drudis.2016.02.015
NAFLD Non-alcoholic Fatty Liver Disease 8. Thakkar S, Li T, Liu Z, et al. Drug-induced liver injury severity and
MASH Metabolic Associated Steatohepatitis toxicity (DILIst): binary classification of 1279 drugs by human
DAMP Damage associated molecular patterns hepatotoxicity. Drug Discov Today. 2020 Jan;25(1):201–8. doi: 10.
GCDCA Glycochenodeoxycholic acid 1016/j.drudis.2019.09.022
9. Weaver RJ, Valentin JP. Today’s challenges to de-risk and predict
ISTAND Innovative Science and Technology Approaches for
drug safety in human “mind-the-gap”. Toxicol Sci. 2019 Feb 1;167
New Drugs (2):307–21. doi: 10.1093/toxsci/kfy270
AI/ML Artificial Intelligence/Machine Learning •• Provides the reader with gaps in drug safety assessment and
DARPA Defense Advanced Research Projects Agency highlights the need for better translational models.
QSP Quantitative Systems Pharmacology 10. Gerussi A, Natalini A, Antonangeli F, et al. Immune-mediated
drug-induced liver injury: immunogenetics and experimental
models. Int J Mol Sci. 2021 Apr 27;22(9):4557. doi: 10.3390/
Funding ijms22094557
11. Hoofnagle JH, Bjornsson ES, Longo DL. Drug-induced liver
This paper was not funded. injury – types and phenotypes. N Engl J Med. 2019 Jul 18;381
(3):264–273. doi: 10.1056/NEJMra1816149
12. Tutty MA, Movia D, Prina-Mello A. Three-dimensional (3D) liver cell
Declaration of interests models – a tool for bridging the gap between animal studies and
clinical trials when screening liver accumulation and toxicity of
S Jadalannagari and L Ewart are employees of Emulate Inc and are eligible nanobiomaterials. Drug Deliv Transl Res. 2022 Sep;12(9):2048–74.
for stock options with the company. doi: 10.1007/s13346-022-01147-0
13. Bauer S, Wennberg Huldt C, Kanebratt KP, et al. Functional cou­
pling of human pancreatic islets and liver spheroids on-a-chip:
Reviewer disclosures towards a novel human ex vivo type 2 diabetes model. Sci Rep.
2017 Nov 6;7(1):14620. doi: 10.1038/s41598-017-14815-w
Peer reviewers on this manuscript have no relevant financial or other 14. Proctor WR, Foster AJ, Vogt J, et al. Utility of spherical human liver
relationships to disclose. microtissues for prediction of clinical drug-induced liver injury. Arch
Toxicol. 2017 Aug;91(8):2849–63. doi: 10.1007/s00204-017-2002-1
15. Vorrink SU, Zhou Y, Ingelman-Sundberg M, et al. Prediction of
Acknowledgments drug-induced hepatotoxicity using long-term stable primary hepa­
tic 3D spheroid cultures in chemically defined conditions. Toxicol
The authors would like to thank J Chaff, J Wells and BG Zani for their Sci. 2018 Jun 1;163(2):655–65. doi: 10.1093/toxsci/kfy058
support in reviewing this paper. 16. Horner S, Robinson S, Lees D, et al. Target organ profiles in toxicity
studies supporting human dosing: an assessment of recovery and
chronic dosing. Regul Toxicol Pharmacol. 2014 Oct;70(1):270–85.
Author contributions 17. McKenzie R, Fried MW, Sallie R, et al. Hepatic failure and lactic
acidosis due to fialuridine (FIAU), an investigational nucleoside
S Jadalannagari and L Ewart contributed equally to the conception of this analogue for chronic hepatitis B. N Engl J Med. 1995 Oct 26;333
paper, the manuscript preparation, writing, and review. (17):1099–105. doi: 10.1056/NEJM199510263331702
 EXPERT OPINION ON DRUG METABOLISM & TOXICOLOGY 617

18. Lee EW, Lai Y, Zhang H, et al. Identification of the mitochondrial 35. Chang JH, Sangaraju D, Liu N, et al. Comprehensive evaluation of
targeting signal of the human equilibrative nucleoside transporter bile acid homeostasis in human hepatocyte co-culture in the pre­
1 (hENT1): implications for interspecies differences in mitochondrial sence of troglitazone, pioglitazone, and acetylsalicylic acid. Mol
toxicity of fialuridine. J Biol Chem. 2006 Jun 16;281(24):16700–6. Pharm. 2019 Oct 7;16(10):4230–40. doi: 10.1021/acs.molpharma
doi: 10.1074/jbc.M513825200 ceut.9b00562
19. Administration USFD. About Alternative Methods 2023. Available 36. Novik EI, Dwyer J, Morelli JK, et al. Long-enduring primary
from: https://www.fda.gov/science-research/about-science- hepatocyte-based co-cultures improve prediction of
research-fda/advancing-alternative-methods-fda hepatotoxicity. Toxicol Appl Pharmacol. 2017 Dec 1;336:20–30.
20. Baran SW, Brown PC, Baudy AR, et al. Perspectives on the evaluation doi: 10.1016/j.taap.2017.09.013
and adoption of complex in vitro models in drug development: work­ 37. Burton RD, Hieronymus T, Chamem T, et al. Assessment of the
shop with the FDA and the pharmaceutical industry (IQ MPS affiliate). biotransformation of low-turnover drugs in the hmicrorel human
ALTEX. 2022;39(2):297–314. doi:10.14573/altex.2112203 hepatocyte coculture model. Drug Metab Dispos. 2018 Nov;46
21. Ekert JE, Deakyne J, Pribul-Allen P, et al. Recommended guidelines (11):1617–1625.
for developing, qualifying, and implementing complex in vitro 38. Bhise NS, Manoharan V, Massa S, et al. A liver-on-a-chip platform
models (CIVMs) for drug discovery. SLAS Discov. 2020 Dec;25 with bioprinted hepatic spheroids. Biofabrication. 2016 Jan 12;8
(10):1174–90. doi: 10.1177/2472555220923332 (1):014101. doi: 10.1088/1758-5090/8/1/014101
•• This article highlights the need to qualify CIVMs for specific 39. Crogan-Grunnndy CSR, Smith CR, Harwick NR, Nguyen GD, editor.
context of use such as drug discovery. Utilization of the ex vive human liver tissue model to assess
22. Ingber DE. Human organs-on-chips for disease modelling, drug drug-induced liver injury across a diverse set off chemical classes.
development and personalized medicine. Nat Rev Genet. 2022 Society of Toxicology Annual Meeting; 2017 March 12–16;
Aug;23(8):467–91. doi: 10.1038/s41576-022-00466-9 Baltimore, Maryland; 2017.
23. Low LA, Mummery C, Berridge BR, et al. Organs-on-chips: into the 40. Nguyen DG, Funk J, Robbins JB, et al. Bioprinted 3D primary liver
next decade. Nat Rev Drug Discov. 2021 May;20(5):345–61. doi: 10. tissues allow assessment of organ-level response to clinical drug
1038/s41573-020-0079-3 induced toxicity in vitro. PloS One. 2016;11(7):e0158674. doi: 10.
• Provides information on organ chips in general along with its 1371/journal.pone.0158674
advantages and disadvantages. 41. Bircsak KM, DeBiasio R, Miedel M, et al. A 3D microfluidic liver
24. Leung CM, de Haan P, Ronaldson-Bouchard K, et al. A guide to the model for high throughput compound toxicity screening in the
organ-on-a-chip. Nat Rev Meth Primers. 2022 May 12;2(1). doi: 10. OrganoPlate(R). Toxicology. 2021 Feb 28;450:152667. doi: 10.1016/j.
1038/s43586-022-00118-6 tox.2020.152667
25. Bhatia SN, Ingber DE. Microfluidic organs-on-chips. Nat Biotechnol. 42. Cox B, Barton P, Class R, et al. Setup of human liver-chips integrat­
2014 Aug;32(8):760–772. doi: 10.1038/nbt.2989 ing 3D models, microwells and a standardized microfluidic plat­
•• This article provides the basic understanding of Organ Chips form as proof-of-concept study to support drug evaluation.
including - fabrication, building biology, advantages and Biomater Biosyst. 2022 Aug;7:100054. doi: 10.1016/j.bbiosy.2022.
disadvantages. 100054
26. Ewart L, Fabre K, Chakilam A, et al. Navigating tissue chips from 43. Eckstrum K, Striz A, Ferguson M, et al. Evaluation of the utility of
development to dissemination: a pharmaceutical industry the beta human liver emulation system (BHLES) for CFSAN’s reg­
perspective. Exp Biol Med (Maywood). 2017 Oct;242(16):1579–85. ulatory toxicology program. Food Chem Toxicol. 2022
doi: 10.1177/1535370217715441 Mar;161:112828. doi: 10.1016/j.fct.2022.112828
27. Marx U, Akabane T, Andersson TB, et al. Biology-inspired microphysio­ • In this article US FDA evaluates the performance of an MPS
logical systems to advance patient benefit and animal welfare in drug platform for the toxicology assessment by demonstrating high
development. ALTEX. 2020;37(3):365–394. doi:10.14573/altex.2001241 sensitivity and specificity.
28. Ajist Dash WRP Hepatic microphysiological systems: current and 44. Eckstrum K, Striz A, Ferguson M, et al. Utilization of a model
future applications in drug discovery and development. In: hepatotoxic compound, diglycolic acid, to evaluate liver
Jeffrey T, Borenstein V, Sarah LT Joseph LC, editors. Microfluidic Organ-Chip performance and in vitro to in vivo concordance.
cell culture systems (Second Edition). Vol. 2. Micro and Nano Food Chem Toxicol. 2020 Dec;146:111850. doi: 10.1016/j.fct.2020.
Technologies: Elsevier; 2019. p. 159–186. doi:10.1016/B978-0-12- 111850
813671-3.00006-2 45. Ewart L, Apostolou A, Briggs SA, et al. Performance assessment and
29. Moradi E, Jalili-Firoozinezhad S, Solati-Hashjin M. Microfluidic economic analysis of a human liver-chip for predictive toxicology.
organ-on-a-chip models of human liver tissue. Acta Biomater. Commun Med (Lond).2022 Dec 6;2(1):154. doi: 10.1038/s43856-
2020 Oct 15;116:67–83. doi: 10.1016/j.actbio.2020.08.041 022-00209-1
30. Vernetti LA, Senutovitch N, Boltz R, et al. A human liver microphy­ 46. Zhang CJ, Meyer SR, O’Meara MJ, et al. A human liver organoid
siology platform for investigating physiology, drug safety, and screening platform for DILI risk prediction. J Hepatol. 2023 May;78
disease models. Exp Biol Med (Maywood). 2016 Jan;241(1):101–14. (5):998–1006.
doi: 10.1177/1535370215592121 47. Rubiano A, Indapurkar A, Yokosawa R, et al. Characterizing the
31. Kang W, Podtelezhnikov AA, Tanis KQ, et al. Development and reproducibility in using a liver microphysiological system for assay­
application of a transcriptomic signature of bioactivation in an ing drug toxicity, metabolism, and accumulation. Clin Transl Sci.
advanced in vitro liver model to reduce drug-induced liver injury 2021 May;14(3):1049–61. doi: 10.1111/cts.12969
risk early in the pharmaceutical pipeline. Toxicol Sci. 2020 Sep 48. Beaudoin JJ, Clemens L, Miedel MT, et al. The combination of
1;177(1):121–39. doi: 10.1093/toxsci/kfaa094 a human biomimetic liver microphysiology system with
32. Bell CC, Hendriks DF, Moro SM, et al. Characterization of primary BIOLOGXsym, a quantitative systems toxicology (QST) modeling
human hepatocyte spheroids as a model system for drug-induced platform for macromolecules, provides mechanistic understanding
liver injury, liver function and disease. Sci Rep. 2016 May 4;6 of tocilizumab- and GGF2-induced liver injury. Int J Mol Sci. 2023
(1):25187. doi: 10.1038/srep25187 Jun 2;24(11):9692. doi: 10.3390/ijms24119692
33. Albrecht W, Kappenberg F, Brecklinghaus T, et al. Prediction of 49. Li X, George SM, Vernetti L, et al. A glass-based, continuously
human drug-induced liver injury (DILI) in relation to oral doses zonated and vascularized human liver acinus microphysiological
and blood concentrations. Arch Toxicol. 2019 Jun;93(6):1609–37. system (vLAMPS) designed for experimental modeling of diseases
doi: 10.1007/s00204-019-02492-9 and ADME/TOX. Lab Chip.2018 Aug 21;18(17):2614–31. doi: 10.
34. Messner S, Agarkova I, Moritz W, et al. Multi-cell type human liver 1039/c8lc00418h
microtissues for hepatotoxicity testing. Arch Toxicol. 2013 Jan;87 50. Rajan SAP, Sherfey J, Ohri S, et al. A novel milli-fluidic liver tissue
(1):209–13. doi: 10.1007/s00204-012-0968-2 chip with continuous recirculation for predictive pharmacokinetics
618 S. JADALANNAGARI AND L. EWART

applications. AAPS J.2023 Oct 27;25(6):102. doi: 10.1208/s12248- • This article characterizes the utility and reproducability of an
023-00870-x MPS platform using different cell types.
51. Shoemaker JT, Zhang W, Atlas SI, et al. A 3D cell culture organ-on- 67. Wang J, Zhang L, Sun J, et al. Predicting drug-induced liver
a-chip platform with a breathable hemoglobin analogue augments injury using graph attention mechanism and molecular
and extends primary human hepatocyte functions in vitro. Front fingerprints. Methods. 2024 Jan;221:18–26. doi: 10.1016/j.ymeth.
Mol Biosci. 2020;7:568777. doi: 10.3389/fmolb.2020.568777 2023.11.014
52. Tan K, Keegan P, Rogers M, et al. A high-throughput microfluidic 68. Alnouti Y. Bile acid sulfation: a pathway of bile acid elimination and
microphysiological system (PREDICT-96) to recapitulate hepatocyte detoxification. Toxicol Sci. 2009 Apr;108(2):225–46. doi: 10.1093/toxsci/
function in dynamic, re-circulating flow conditions. Lab Chip. 2019 kfn268
Apr 23;19(9):1556–66. doi: 10.1039/c8lc01262h 69. McAleer CW, Pointon A, Long CJ, et al. On the potential of in vitro
53. Baudy AR, Otieno MA, Hewitt P, et al. Liver microphysiological organ-chip models to define temporal pharmacokinetic-pharmacody­
systems development guidelines for safety risk assessment in the namic relationships. Sci Rep.2019 Jul 3;9(1):9619. doi: 10.1038/s41598-
pharmaceutical industry. Lab Chip. 2020 Jan 21;20(2):215–25. doi: 019-45656-4
10.1039/c9lc00768g 70. Ronaldson-Bouchard K, Teles D, Yeager K, et al. A multi-organ chip
•• This manuscripts introduces guidance provided by the IQ MPS with matured tissue niches linked by vascular flow. Nat Biomed
affiliate for building CIVMs./ Eng. 2022 Apr;6(4):351–71. doi: 10.1038/s41551-022-00882-6
54. Lauschke VM, Hendriks DF, Bell CC, et al. Novel 3D culture systems 71. Campbell SB, Wu Q, Yazbeck J, et al. Beyond polydimethylsiloxane:
for studies of human liver function and assessments of the hepa­ alternative materials for fabrication of organ-on-a-chip devices and
totoxicity of drugs and drug candidates. Chem Res Toxicol. 2016 microphysiological systems. ACS Biomater Sci Eng. 2021 Jul 12;7
Dec 19;29(12):1936–1955. doi: 10.1021/acs.chemrestox.6b00150 (7):2880–99. doi: 10.1021/acsbiomaterials.0c00640
55. Josse R, Aninat C, Glaise D, et al. Long-term functional stability of 72. Toepke MW, Beebe DJ. PDMS absorption of small molecules and
human HepaRG hepatocytes and use for chronic toxicity and gen­ consequences in microfluidic applications. Lab Chip. 2006 Dec;6
otoxicity studies. Drug Metab Dispos. 2008 Jun;36(6):1111–8. doi: (12):1484–6. doi: 10.1039/b612140c
10.1124/dmd.107.019901 73. Auner AW, Tasneem KM, Markov DA, et al. Chemical-PDMS binding
56. Corbett JL, Duncan SA. iPSC-derived hepatocytes as a platform for kinetics and implications for bioavailability in microfluidic devices.
disease modeling and drug discovery. Front Med. 2019;6:265. doi: Lab Chip.2019 Feb 26;19(5):864–74. doi: 10.1039/c8lc00796a
10.3389/fmed.2019.00265 74. Piergiovanni M, Leite SB, Corvi R, et al. Standardisation needs for
57. Low LA, Tagle DA. Organs-on-chips: progress, challenges, and organ on chip devices. Lab Chip. 2021 Aug 7;21(15):2857–68. doi:
future directions. Exp Biol Med (Maywood). 2017 Oct;242 10.1039/d1lc00241d
(16):1573–1578. doi: 10.1177/1535370217700523 75. Reyes DR, Esch MB, Ewart L, et al. From animal testing to in vitro
58. Ewart L, Roth A. Opportunities and challenges with microphysiolo­ systems: advancing standardization in microphysiological systems.
gical systems: a pharma end-user perspective. Nat Rev Drug Discov. Lab Chip. 2024. doi:10.1039/D3LC00994G 24 5 1076–1087
2021 May;20(5):327–8. doi: 10.1038/d41573-020-00030-2 76. Danku AE, Dulf EH, Braicu C, et al. Organ-On-A-Chip: a survey of
59. Stresser DM, Kopec AK, Hewitt P, et al. Towards in vitro models for technical results and problems. Front Bioeng Biotechnol.
reducing or replacing the use of animals in drug testing. Nat 2022;10:840674. doi: 10.3389/fbioe.2022.840674
Biomed Eng. 2023 Dec 27. doi: 10.1038/s41551-023-01154-7 77. Przybylak KR, Cronin MT. In silico models for drug-induced liver
60. Fabre K, Berridge B, Proctor WR, et al. Introduction to a manuscript injury–current status. Expert Opin Drug Metab Toxicol. 2012 Feb;8
series on the characterization and use of microphysiological sys­ (2):201–17. doi: 10.1517/17425255.2012.648613
tems (MPS) in pharmaceutical safety and ADME applications. Lab 78. Chalasani N, Bjornsson E. Risk factors for idiosyncratic drug-induced
Chip. 2020 Mar 17;20(6):1049–57. doi: 10.1039/c9lc01168d liver injury. Gastroenterology. 2010 Jun;138(7):2246–59. doi: 10.
61. Shi Q, Arefin A, Ren L, et al. Co-culture of human primary hepato­ 1053/j.gastro.2010.04.001
cytes and nonparenchymal liver cells in the Emulate(R) liver-chip 79. Mihajlovic M, Vinken M. Mitochondria as the target of hepatotoxicity
for the study of drug-induced liver injury. Curr Protoc. 2022 Jul;2(7): and drug-induced liver injury: molecular mechanisms and detection
e478. methods. Int J Mol Sci. 2022 Mar 18;23(6):3315. doi: 10.3390/
62. Shi Q, Ren L, Sauld J, et al. Lessons learnt in establishing a reliable ijms23063315
and low-cost assay for urea production in human primary hepato­ 80. Pauli-Magnus C, Meier PJ. Hepatobiliary transporters and
cytes cultured in a Liver-Chip for the study of drug hepatotoxicity. drug-induced cholestasis. Hepatology. 2006 Oct;44(4):778–87. doi:
Society of Toxicology; 2023 March 20; Nashville, Tennessee; 2023. 10.1002/hep.21359
p. P106. 81. Welch MA, Kock K, Urban TJ, et al. Toward predicting drug-induced
63. Baker TK, Van Vleet TR, Mahalingaiah PK, et al. The Current status liver injury: parallel computational approaches to identify multidrug
and use of microphysiological systems by the pharmaceutical resistance protein 4 and bile salt export pump inhibitors. Drug Metab
industry: the international consortium for innovation and quality Dispos. 2015 May;43(5):725–34. doi: 10.1124/dmd.114.062539
microphysiological systems affiliate survey and commentary. Drug 82. Howell BA, Siler SQ, Barton HA, et al. Development of quantitative
Metab Dispos.2024 Feb 14;52(3):198–209. doi: 10.1124/dmd.123. systems pharmacology and toxicology models within consortia:
001510 experiences and lessons learned through DILIsym development.
64. Carius P, Weinelt FA, Cantow C, et al. Addressing the ADME chal­ Drug Discov Today Dis Models. 2016 Dec 01;22:5–13.
lenges of compound loss in a PDMS-Based gut-on-chip microphy­ 83. Woodhead JL, Watkins PB, Howell BA, et al. The role of quantitative
siological system. Pharmaceutics. 2024;16(3):296. doi: 10.3390/ systems pharmacology modeling in the prediction and explanation of
pharmaceutics16030296 idiosyncratic drug-induced liver injury. Drug Metab Pharmacokinet.
65. Lim AY, Kato Y, Sakolish C, et al. Reproducibility and robustness of 2017 Feb;32(1):40–5.
a liver microphysiological system PhysioMimix LC12 under varying 84. Williams DP, Lazic SE, Foster AJ, et al. Predicting drug-induced liver
culture conditions and cell type combinations. Bioeng (Basel). 2023 injury with bayesian machine learning. Chem Res Toxicol. 2020 Jan
Oct 14;10(10):1195. doi: 10.3390/bioengineering10101195 21;33(1):239–48. doi: 10.1021/acs.chemrestox.9b00264
•• This article compares different cell types on multiple MPS 85. Zollner G, Wagner M, Trauner M. Nuclear receptors as drug targets
platforms to boost end-user adoption. in cholestasis and drug-induced hepatotoxicity. Pharmacol Ther.
66. Kato Y, Lim AY, Sakolish C, et al. Analysis of reproducibility and 2010 Jun;126(3):228–43. doi: 10.1016/j.pharmthera.2010.03.005
robustness of OrganoPlate(R) 2-lane 96, a liver microphysiological 86. Bhamidimarri KR, Schiff E. Drug-induced cholestasis. Clin Liver Dis.
system for studies of pharmacokinetics and toxicological assess­ 2013 Nov;17(4):519–31. doi: 10.1016/j.cld.2013.07.015 vii
ment of drugs. Toxicol Vitro. 2022 Dec;85:105464. doi: 10.1016/j.tiv. 87. Hendriks DF, Fredriksson Puigvert L, Messner S, et al. Hepatic 3D
2022.105464 spheroid models for the detection and study of compounds with
 EXPERT OPINION ON DRUG METABOLISM & TOXICOLOGY 619

cholestatic liability. Sci Rep. 2016 Oct 19;6(1):35434. doi: 10.1038/ 95. Mantena SK, King AL, Andringa KK, et al. Mitochondrial dysfunction
srep35434 and oxidative stress in the pathogenesis of alcohol- and
88. Daly AK, Day CP. Genetic association studies in drug-induced liver obesity-induced fatty liver diseases. Free Radic Biol Med. 2008
injury. Drug Metab Rev. 2012 Feb;44(1):116–126. doi: 10.3109/ Apr 1;44(7):1259–72. doi: 10.1016/j.freeradbiomed.2007.12.029
03602532.2011.605790 96. Zhu Y, Chen L, He Y, et al. The alteration of drug metabolism
89. de Haan L, Suijker J, van Roey R, et al. A microfluidic 3D enzymes and pharmacokinetic parameters in nonalcoholic fatty
endothelium-on-a-chip model to study transendothelial migration liver disease: current animal models and clinical practice. Drug
of T cells in health and disease. IJMS. 2021 Jul 30;22(15):8234. doi: Metab Rev. 2023 Aug;55(3):163–180.
10.3390/ijms22158234 97. Jang KJ, Otieno MA, Ronxhi J. et al. Reproducing human and
90. Hernandez N, Bessone F. Hepatotoxicity induced by biological cross-species drug toxicities using a liver-chip. Sci Transl Med.
agents: clinical features and Current controversies. J Clin Transl 2019 Nov 6;11(517). doi: 10.1126/scitranslmed.aax5516
Hepatol. 2022 Jun 28;10(3):486–95. doi: 10.14218/JCTH.2021.00243 98. Hu J, Xu Y, Hao J, et al. MiR-122 in hepatic function and liver
91. Chhabra A, Song HG, Grzelak KA, et al. A vascularized model of the diseases. Protein Cell. 2012 May;3(5):364–71. doi: 10.1007/s13238-
human liver mimics regenerative responses. Proc Natl Acad Sci USA. 012-2036-3
2022 Jul 12;119(28):e2115867119. doi: 10.1073/pnas.2115867119 99. Liu Y, Li P, Liu L, et al. The diagnostic role of miR-122 in
92. Teng ML, Ng CH, Huang DQ, et al. Global incidence and prevalence drug-induced liver injury: a systematic review and meta-analysis.
of nonalcoholic fatty liver disease. Clin Mol Hepatol. 2023 Feb;29 Medicine (Baltimore). 2018 Dec;97(49):e13478. doi: 10.1097/MD.
(Suppl):S32–S42. doi: 10.3350/cmh.2022.0365 0000000000013478
93. Massart J, Begriche K, Moreau C, et al. Role of nonalcoholic fatty liver 100. Administration USFD. Innovative science and technology
disease as risk factor for drug-induced hepatotoxicity. J Clin Trans Res. approaches for new drugs (ISTAND) pilot program 2023 [cited
2017 Feb;3(Suppl 1):212–32. doi: 10.18053/jctres.03.2017S1.006 2024 Feb 21]. Available from: https://www.fda.gov/drugs/drug-
94. Jamwal R, de la Monte SM, Ogasawara K, et al. Nonalcoholic fatty development-tool-ddt-qualification-programs/innovative-science-
liver disease and diabetes are associated with decreased CYP3A4 and-technology-approaches-new-drugs-istand-pilot-program
protein expression and activity in human liver. Mol Pharm. 2018 Jul •• This article provides guidance on the use of CIVMs for drug
2;15(7):2621–32. doi: 10.1021/acs.molpharmaceut.8b00159 development from a regulatory perspective.