Why do traditional animal models fail to predict human drug toxicity?

Traditional animal models fail to predict human drug toxicity due to fundamental species differences in drug metabolism enzymes (especially CYP450 isoforms), transporter protein expression, cellular sensitivity to toxic insults, and immune system responses. For example, humans express CYP3A4 as a major metabolic enzyme while rodents express different CYP3A isoforms with different substrate specificities. Additionally, the BSEP transporter responsible for many cases of DILI has different inhibition profiles across species. These differences result in only 50-60% concordance between animal toxicity findings and human outcomes.

How accurate are human liver chips at predicting drug-induced liver injury (DILI)?

Human liver chips demonstrate 87% accuracy in predicting DILI, with sensitivity rates of 80-90% and specificity above 90%. In head-to-head comparisons, liver chips correctly identified hepatotoxic drugs that passed animal testing but failed in humans. The improved accuracy comes from using primary human hepatocytes or iPSC-derived cells, maintaining proper cell polarity and bile canaliculi formation, including non-parenchymal cells like Kupffer cells, and enabling long-term culture for chronic toxicity assessment.

What is the CiPA initiative for cardiac safety testing?

The Comprehensive in vitro Proarrhythmia Assay (CiPA) is an FDA-supported initiative that modernizes cardiac safety testing by combining ion channel assays, in silico modeling, and human iPSC-derived cardiomyocyte (iPSC-CM) testing. CiPA moves beyond simple hERG channel testing to assess multiple cardiac ion channels and integrated cellular responses. The iPSC-CM component provides a human-relevant model that can detect drug effects on action potential duration, contractility, and arrhythmogenic risk that may be missed by traditional methods.

Can organ-on-chip systems detect metabolite-mediated toxicity?

Yes, multi-organ chip systems excel at detecting metabolite-mediated toxicity by connecting metabolically active liver compartments to target organ compartments. When a drug passes through the liver module, it generates human-relevant metabolites that then flow to heart, kidney, or other tissue modules. This captures toxicity from reactive metabolites that would be missed in single-organ assays. Body-on-chip systems have successfully identified liver-mediated cardiotoxicity and nephrotoxicity from prodrugs and compounds with toxic metabolites.

What regulatory pathways exist for qualifying organ-on-chip toxicity data?

The FDA ISTAND (Innovative Science and Technology Approaches for New Drugs) program provides a pathway for qualifying novel toxicology methods including organ-on-chip platforms. Companies can submit qualification packages demonstrating analytical validation, biological relevance, and predictive performance. Additionally, the FDA Modernization Act 2.0 explicitly allows non-animal testing methods for IND applications. The EPA NAMs (New Approach Methodologies) Workplan promotes organ-chip adoption for environmental toxicity, while ICH guidelines increasingly accept human cell-based assays for specific endpoints like cardiac safety (ICH S7B/CiPA).

APPLICATIONSSafety AssessmentPredictive ToxicologyFDA ISTAND Qualified

Application Domain

Toxicity Testing

Predictive Human Safety Assessment with Organ-on-Chip Technology

Written by J Radler | Patient Analog

Last updated: January 2025

Key Applications

Direct clinical applications improving patient outcomes
Cost-effective alternatives to traditional methods
Scalable solutions for pharmaceutical development
Regulatory-accepted methodologies for drug approval

← Back to Applications

⚠ WHY TOXICITY TESTING MATTERS

50%

Drug-induced liver injury (DILI) causes 50% of acute liver failures in the United States^[1]

17%

Cardiotoxicity responsible for 17% of drug withdrawals from market post-approval^[2]

20%

Nephrotoxicity accounts for 20% of clinical trial failures in Phase II/III^[3]

87%

Human liver chips predict DILI with 87% accuracy vs 50% for animal models^[4]

The Toxicity Prediction Problem

Drug toxicity represents one of the most significant challenges in pharmaceutical development, accounting for approximately 30% of drug failures^[5] during clinical trials and post-market withdrawals. The fundamental problem lies in the poor translatability of preclinical safety data to human outcomes.

Species Differences in Drug Metabolism

Human cytochrome P450 enzymes, particularly CYP3A4, CYP2D6, and CYP2C19, have different substrate specificities and expression patterns compared to rodent orthologs. A drug that is safely metabolized in mice may generate toxic metabolites in humans, or vice versa. This metabolic divergence underlies many cases of unexpected human toxicity.

Transporter Expression Differences

Drug transporters like BSEP (bile salt export pump), OATP, and OCT have species-specific expression patterns and inhibition profiles. BSEP inhibition is a major mechanism of DILI, but rodent BSEP has different sensitivity to many drugs compared to human BSEP. This explains why some hepatotoxic drugs pass animal testing.

The Human Cost of Prediction Failure

Traditional animal models achieve only 50-60% concordance with human toxicity outcomes. This means nearly half of all drug toxicity predictions are wrong - either missing toxic drugs that harm patients or falsely flagging safe drugs that never reach patients who need them. The pharmaceutical industry spends over $2.5 billion per approved drug partly due to these late-stage toxicity failures.

⚙ Hepatotoxicity Testing (DILI Prediction)

Drug-induced liver injury (DILI) is the leading cause of acute liver failure and the primary reason for post-market drug withdrawals. The liver's role in drug metabolism makes it uniquely vulnerable to toxic insults, while species differences in metabolism make animal prediction unreliable.

Liver-on-Chip Technology

Primary human hepatocytes maintain phenotype 28+ days
Proper bile canaliculi formation for BSEP studies
Flow-based media delivery mimics sinusoidal blood flow
Integration of non-parenchymal cells (Kupffer, stellate)
Real-time albumin and urea secretion monitoring

3D Liver Microtissues

InSphero 3D InSight liver models
Spheroid architecture improves cell-cell contacts
Extended viability for chronic toxicity studies
Enables repeat-dose toxicity assessment
Compatible with high-throughput screening

DILI Mechanism Detection

Mitochondrial toxicity (JC-1, MitoTracker)
Oxidative stress (ROS, GSH depletion)
BSEP inhibition and cholestasis
Reactive metabolite formation
Idiosyncratic DILI with immune co-culture

Performance Data: Liver Chip vs Animal Models

In validation studies using drugs with known human DILI outcomes, liver-on-chip platforms achieved 87% sensitivity (correctly identifying toxic drugs) and 100% specificity (not falsely flagging safe drugs). This compares to approximately 50% sensitivity for traditional rodent studies. Notably, liver chips correctly identified trovafloxacin, troglitazone, and other drugs that caused human DILI but passed animal testing.

♥ Cardiotoxicity Testing (iPSC-CMs & CiPA)

Cardiotoxicity remains a leading cause of drug withdrawals, with arrhythmogenic risk being particularly difficult to predict. The Comprehensive in vitro Proarrhythmia Assay (CiPA) initiative represents a paradigm shift in cardiac safety assessment, incorporating human iPSC-derived cardiomyocytes as a central component.

The CiPA Framework

Ion Channel Assays

hERG, Nav1.5, Cav1.2 channel screening

In Silico Modeling

O'Hara-Rudy cardiac AP simulation

iPSC-CM Testing

Human cardiomyocyte functional assays

Clinical Validation

ECG-based QT assessment confirmation

iPSC-Cardiomyocyte Endpoints

Action potential duration (APD90) measurement
Contractility and beat rate assessment
Calcium transient dynamics
Arrhythmia detection (EADs, DADs)
Structural cardiotoxicity (cell death, hypertrophy)

Heart-on-Chip Advantages

3D tissue architecture with aligned fibers
Mechanical loading mimics cardiac physiology
Multi-cell type composition (CMs, fibroblasts, ECs)
Long-term culture for chronic effects
Real-time contractility monitoring

Beyond hERG: Multi-Ion Channel Assessment

Traditional cardiac safety focused on hERG channel blockade (QT prolongation), but this approach has generated both false positives (blocking safe drugs) and false negatives (missing non-hERG cardiotoxicity). CiPA's integrated approach assesses Nav1.5, Cav1.2, and multiple potassium channels to provide a complete proarrhythmic risk profile. iPSC-CMs naturally express these channels and provide an integrated readout of drug effects.

◒ Nephrotoxicity Testing (Kidney Chips)

The kidneys are particularly susceptible to drug-induced injury due to their role in drug excretion and high blood flow. Nephrotoxicity accounts for 20% of clinical trial failures, with aminoglycosides, cisplatin, and NSAIDs among the most problematic drug classes. Kidney-on-chip platforms replicate the proximal tubule's unique transport and metabolic functions.

Proximal Tubule-on-Chip

Human primary RPTECs or iPSC-derived cells
Polarized epithelium with brush border
Functional transporters (OAT1, OAT3, OCT2)
Flow-induced shear stress
Tubular secretion assessment

Toxicity Biomarkers

KIM-1 (Kidney Injury Molecule-1)
NGAL (Neutrophil gelatinase-associated lipocalin)
Clusterin and osteopontin
NAG enzyme activity
GGT and LDH release

Glomerulus Models

Podocyte-endothelial co-culture
Glomerular filtration barrier
Albumin permeability assessment
Podocyte foot process modeling
Glomerulonephritis disease models

Case Study: Cisplatin Nephrotoxicity

Kidney-on-chip platforms have successfully modeled cisplatin-induced acute kidney injury, demonstrating dose-dependent proximal tubule cell death, elevated KIM-1 secretion, and disrupted transporter function. These models identified protective compounds that reduce cisplatin nephrotoxicity without compromising anti-cancer efficacy - a finding validated in subsequent clinical studies.

🧠 Neurotoxicity Testing (Brain Organoids & Nerve Chips)

Neurotoxicity assessment is particularly challenging due to the blood-brain barrier's protective role and the complexity of neural circuits. Drug-induced neurological adverse events include seizures, peripheral neuropathy, cognitive impairment, and movement disorders. Human neural models provide species-specific insights impossible to obtain from rodent brains with different neurotransmitter systems and receptor distributions.

Brain Organoid Applications

Cortical organoids for developmental neurotoxicity
Midbrain organoids for dopaminergic toxicity
Cerebral organoids with multiple brain regions
Electrophysiology via MEA recordings
Calcium imaging for network activity

Peripheral Nerve Models

AxoSim NerveSim for CIPN detection
Dorsal root ganglion (DRG) cultures
Sensory neuron function assessment
Nerve conduction velocity measurement
Axonal degeneration quantification

BBB-on-Chip

Brain endothelial-astrocyte-pericyte co-culture
Tight junction formation and TEER
P-gp efflux transporter function
CNS drug penetration prediction
Neuroinflammation modeling

Chemotherapy-Induced Peripheral Neuropathy (CIPN)

CIPN affects up to 70% of patients receiving taxanes, platinum compounds, or vinca alkaloids, yet animal models poorly predict human susceptibility. Human iPSC-derived sensory neurons and DRG organoids can detect CIPN-inducing compounds with high sensitivity, enabling identification of neuroprotective agents and dose optimization strategies that minimize neurotoxicity while maintaining anti-cancer efficacy.

★ Multi-Organ Toxicity (Body-on-Chip)

Many drug toxicities result from metabolite formation in one organ affecting another - a phenomenon impossible to capture in single-organ assays. Body-on-chip systems connect multiple organ modules through a common circulation, enabling systemic toxicity assessment and ADME-Tox integration.

Multi-Organ Connectivity Configurations

Liver-Heart

Detects cardiotoxicity from hepatic metabolites (e.g., terfenadine to fexofenadine pathway)

Liver-Kidney

Models hepatorenal syndrome and metabolite-mediated nephrotoxicity

Gut-Liver

First-pass metabolism and gut microbiome-drug interactions

10-Organ Systems

TissUse HUMIMIC platform integrates comprehensive organ coverage

Key Capabilities

Metabolite-mediated toxicity detection
PK/PD parameter estimation
Systemic immune response modeling
Organ-organ crosstalk via cytokines
28+ day repeat-dose studies

Leading Platforms

Hesperos Human-on-Chip (up to 14 organs)
TissUse HUMIMIC (10 organ capability)
CN-Bio PhysioMimix (liver-centric)
Emulate multi-organ configurations
Mimetas OrganoPlate (scalable)

✓ Regulatory Framework for Toxicity Testing

Understanding the regulatory landscape is essential for implementing organ-on-chip toxicity testing in drug development. Key guidelines from ICH, OECD, FDA, and EPA provide the framework for validating and applying these new approach methodologies (NAMs) in regulatory submissions.

ICH S7A: Safety Pharmacology Studies

ICH S7A establishes requirements for safety pharmacology core battery studies including cardiovascular, central nervous system, and respiratory function assessment. While traditionally animal-based, regulatory agencies increasingly accept human cell-based alternatives for specific endpoints when properly validated. The guideline emphasizes functional endpoints that iPSC-derived tissues can readily provide.

ICH S7B: Cardiac Safety (QT Prolongation)

ICH S7B specifically addresses delayed ventricular repolarization (QT interval prolongation) and was updated in 2022 to incorporate CiPA principles. The revised guideline explicitly recognizes human iPSC-cardiomyocyte assays as part of an integrated risk assessment strategy, reducing reliance on the traditional hERG assay and in vivo QT studies in animals.

OECD Test Guidelines for In Vitro Toxicity

OECD has adopted multiple test guidelines accepting in vitro methods: TG 439 (skin irritation using reconstructed human epidermis), TG 442C/D/E (skin sensitization defined approaches), TG 491 (eye irritation), and TG 455/456 (estrogen/androgen receptor activity). These guidelines provide internationally harmonized protocols for regulatory acceptance of human cell-based toxicity data.

FDA ISTAND Qualification Pathway

The Innovative Science and Technology Approaches for New Drugs (ISTAND) pilot program provides a formal pathway for qualifying novel drug development tools including organ-on-chip platforms. Successful qualification allows the tool to be used in regulatory submissions for a defined context of use without requiring additional case-by-case justification.

FDA Modernization Act 2.0 (2022)

This landmark legislation removes the mandate requiring animal testing before human clinical trials. The law explicitly lists "cell-based assays, organ chips, microphysiological systems, computer modeling, and other human biology-based test methods" as acceptable alternatives. This regulatory shift creates unprecedented opportunities for companies to develop drugs using human-relevant toxicity data from the outset of development.

EPA New Approach Methodologies (NAMs)

The Environmental Protection Agency's strategic plan aims to eliminate mammalian animal testing for chemical safety by 2035. The ToxCast and Tox21 programs have generated over 75 million data points using high-throughput human cell-based assays. Organ-on-chip systems are being evaluated for more complex endpoints requiring integrated tissue responses, including hepatotoxicity, developmental toxicity, and endocrine disruption.

📄 Case Studies: Drugs That Failed Due to Toxicity

These high-profile drug withdrawals illustrate the critical need for human-relevant toxicity testing. In each case, animal studies failed to predict the toxicity that ultimately harmed patients and led to market withdrawal. Retrospective analysis with organ-on-chip platforms has demonstrated that human models could have detected these toxicities earlier.

Vioxx (Rofecoxib) - Cardiotoxicity

Withdrawn 2004 | Estimated 88,000-140,000 excess cardiac events

This COX-2 inhibitor for arthritis pain passed extensive animal safety testing but caused heart attacks and strokes in humans. The mechanism involved COX-2 inhibition in vascular endothelium altering prostacyclin/thromboxane balance - a pathway with significant human-rodent differences. Human iPSC-derived endothelial-cardiomyocyte co-cultures can now model this vascular toxicity mechanism.

Rezulin (Troglitazone) - Hepatotoxicity

Withdrawn 2000 | 63 confirmed liver failure deaths

This diabetes drug caused severe idiosyncratic liver injury that was not predicted by animal studies. The toxicity involved mitochondrial dysfunction and BSEP inhibition - mechanisms now detectable with human liver chips. Retrospective testing showed liver-on-chip models correctly identified troglitazone as hepatotoxic at clinically relevant concentrations.

Propulsid (Cisapride) - Cardiotoxicity

Withdrawn 2000 | 80+ deaths from cardiac arrhythmias

This gastrointestinal motility drug caused fatal cardiac arrhythmias via hERG channel blockade. The drug had minimal effects in animal cardiac studies due to species differences in ion channel expression. Human iPSC-cardiomyocyte assays now routinely identify hERG-blocking compounds that animal studies miss.

Duract (Bromfenac) - Hepatotoxicity

Withdrawn 1998 | 4 deaths, 8 liver transplants

This NSAID for pain relief caused severe hepatotoxicity with prolonged use. Animal studies showed no liver concerns. The mechanism involved reactive metabolite formation by human CYP enzymes with different specificity than rodent orthologs. Liver-on-chip models with human hepatocytes can detect reactive metabolite-mediated toxicity.

Fialuridine (FIAU) - Multi-organ Toxicity

Clinical trial halted 1993 | 5 deaths, 2 liver transplants

This antiviral for hepatitis B caused fatal mitochondrial toxicity affecting liver, pancreas, and muscle. Extensive animal studies (including primates) showed no toxicity. The mechanism involved human mitochondrial polymerase sensitivity not present in animal enzymes. This tragedy catalyzed development of human-specific mitochondrial toxicity assays.

Baycol (Cerivastatin) - Myotoxicity

Withdrawn 2001 | 52 deaths from rhabdomyolysis

This cholesterol-lowering statin caused fatal muscle breakdown (rhabdomyolysis), particularly in combination with gemfibrozil. The drug-drug interaction was not adequately captured in animal studies. Human skeletal muscle organoids and liver-muscle chip systems can now model statin-induced myotoxicity and identify high-risk drug combinations.

The Common Thread: Species Differences

Each of these drug failures shares a common element: animal models failed to predict human toxicity due to species-specific differences in drug metabolism, transporter function, cellular sensitivity, or immune responses. Human organ-on-chip and organoid models directly address this translational gap by providing physiologically relevant human tissue responses. Retrospective validation studies have demonstrated that these platforms could have identified the toxicity signals that animal studies missed.

Toxicity Testing Platforms: Comprehensive Comparison

Feature	Animal Models	2D Cell Culture	3D Organoids	Organ-on-Chip
Human Relevance	Low (50-60%)	Moderate	High	Very High (87%+)
DILI Prediction Accuracy	~50%	60-70%	80-85%	87%+
Tissue Architecture	Native	None	Self-organized	Engineered
Mechanical Cues	Present	Absent	Limited	Controllable
Multi-Organ Integration	Native	Not possible	Challenging	Integrated
Throughput	Very Low	Very High	Moderate-High	Moderate
Cost per Compound	$$$$$	$	$$	$$$
Study Duration	Weeks-Months	Days	Days-Weeks	Days-Weeks
Ethical Concerns	Significant	Minimal	Minimal	Minimal
Regulatory Acceptance	Established	Established	Growing	Growing (FDA ISTAND)
OECD Guidelines	TG 407, 408, 451-453	TG 442C/D/E, 455/456	TG 439, 492	In development
ICH Guidelines	S1-S11 (all)	S7B (CiPA)	S7B (CiPA)	S7B (CiPA), S2R1

CRITICAL CHALLENGE

Toxicity accounts for ~30% of drug failures in clinical development, with drug-induced liver injury (DILI) being the leading cause of post-market withdrawals. Traditional animal models poorly predict human toxicity due to species differences in drug metabolism, transporter expression, and cellular sensitivity. Human simulation technologies address this gap with physiologically relevant models that achieve 87%+ accuracy in predicting human outcomes.

ORGAN-SPECIFIC APPLICATIONS

Hepatotoxicity (DILI): InSphero 3D liver microtissues predict idiosyncratic and intrinsic liver toxicity with 80%+ sensitivity
Cardiotoxicity: Emulate Heart-Chip and Hesperos cardiac modules detect QT prolongation and contractility effects
Nephrotoxicity: Mimetas kidney tubule models identify proximal tubule damage biomarkers
Neurotoxicity: AxoSim NerveSim detects chemotherapy-induced peripheral neuropathy (CIPN)
Pulmonary Toxicity: Lung-on-chip models assess inhaled drug and environmental toxicant effects

REGULATORY ACCEPTANCE

FDA ISTAND: Qualification pathway for novel toxicology methods
OECD Test Guidelines: TG 442C/D/E for skin sensitization, TG 439 for skin irritation
ICH S7B: CiPA initiative accepts human iPSC-cardiomyocyte assays
EPA NAMs: ToxCast/Tox21 high-throughput screening replacing animal tests

PERFORMANCE METRICS

Human liver organoids and organ-chips demonstrate 80-90% sensitivity and 90%+ specificity for DILI prediction, compared to 50-60% for animal models. Multi-organ systems capture systemic toxicity from metabolite formation, while long-term culture (28+ days) enables repeat-dose toxicity assessment matching regulatory requirements.

Application Comparison

Aspect	Traditional	Organ-on-Chip
Predictive Accuracy	50-60% for animal models	85-95% clinical correlation
Development Speed	10-15 years	5-7 years accelerated
Total Cost	$2.6 billion per drug	$800M-$1.2B with early detection

Explore More

🔬

Platform Technology

Learn about organ chip systems

📜

Regulatory Landscape

FDA approval pathways

🚀

Getting Started

Implement in your lab

Frequently Asked Questions

How do organ chips predict human toxicity?

Organ chips use human cells responding to drugs with human-specific metabolism, transporters, and toxicity pathways. Liver chips detect hepatotoxins that clear safely in rodents but cause liver failure in humans. Heart chips identify QT prolongation missed by animal tests.

What types of toxicity can chips assess?

Chips assess hepatotoxicity (liver), nephrotoxicity (kidney), cardiotoxicity (heart arrhythmias and damage), neurotoxicity (brain chips), developmental toxicity (embryonic tissues), immunotoxicity (immune cell chips), and multi-organ toxicity in linked chip systems.

How accurate are chip toxicity predictions?

Liver chips predict human hepatotoxicity with 85-95 percent accuracy versus 50-60 percent for rat studies. Cardiac chips achieve 85-90 percent accuracy for arrhythmia versus 75 percent for dog studies. Accuracy continues improving as platforms mature and validation data grows.

Can chips detect idiosyncratic drug reactions?

Partially. Idiosyncratic reactions involve genetic susceptibility, immune responses, and rare circumstances. Chips using panels of cells from different genetic backgrounds increase likelihood of detecting susceptible individuals. Immune chips model allergic mechanisms underlying some idiosyncratic reactions.

What is high-throughput toxicity screening on chips?

Platforms like Mimetas OrganoPlate and Emulate multi-chip arrays enable testing 96-384 compounds simultaneously. Automated liquid handling, integrated imaging, and data analysis provide toxicity profiles for entire compound libraries within weeks instead of months.

How do chips reduce animal testing for cosmetics?

Skin chips, eye irritation models, and sensitization assays have achieved regulatory validation replacing rabbit and guinea pig tests. OECD guidelines now accept reconstructed human tissue models enabling cruelty-free cosmetics complying with EU ban.

What is ADMET testing and how do chips help?

ADMET covers Absorption, Distribution, Metabolism, Excretion, Toxicity. Multi-organ chips link gut (absorption), liver (metabolism), kidney (excretion), and heart (toxicity) modeling complete drug journey through body. This predicts human pharmacokinetics without animal dosing studies.

Can chips test chronic toxicity?

Increasingly yes. Extended culture platforms maintain organ function for weeks to months. Repeated dosing protocols mimic chronic exposure. However, full lifetime carcinogenicity assessment still requires longer-term solutions or computational modeling integrated with chip data.

What regulatory acceptance exists for toxicity chips?

EPA accepts certain organ chip data for chemical safety under TSCA. FDA recognizes microphysiological systems for hepatotoxicity and metabolism studies. OECD is developing test guidelines that will create global acceptance. Acceptance continues expanding as validation data accumulates.

What is the future of toxicity testing?

Future is animal-free using integrated testing strategies combining organ chips, computational toxicology, high-throughput screening, and population genomics. Chips will be standard preclinical platform with regulatory requirements updated to accept human-relevant data over animal studies.

References

[1]

Lee WM. Drug-induced hepatotoxicity. N Engl J Med. 2003;349(5):474-485. DOI: 10.1056/NEJMra021844 | PubMed: 12890847

[2]

Onakpoya IJ, Heneghan CJ, Aronson JK. Post-marketing withdrawal of 462 medicinal products because of adverse drug reactions: a systematic review of the world literature. BMC Med. 2016;14:10. DOI: 10.1186/s12916-016-0553-2 | PubMed: 26843061

[3]

Perazella MA, Moeckel GW. Nephrotoxicity from chemotherapeutic agents: clinical manifestations, pathobiology, and prevention/therapy. Semin Nephrol. 2010;30(6):570-581. DOI: 10.1016/j.semnephrol.2010.09.005 | PubMed: 21146122

[4]

Sison-Young RL, Mitsa D, Jenkins RE, et al. Comparative Proteomic Characterization of 4 Human Liver-Derived Single Cell Culture Models Reveals Significant Variation in Drug Disposition Protein Content. Drug Metab Dispos. 2015;43(4):485-493. DOI: 10.1124/dmd.114.062315 | PubMed: 25568491

[5]

Arrowsmith J, Miller P. Trial watch: phase II and phase III attrition rates 2011-2012. Nat Rev Drug Discov. 2013;12(8):569. DOI: 10.1038/nrd4090 | PubMed: 23903212

[6]

Fermini B, Hancox JC, Abi-Gerges N, et al. A New Perspective in the Field of Cardiac Safety Testing through the Comprehensive In Vitro Proarrhythmia Assay Paradigm. J Biomol Screen. 2016;21(1):1-11. DOI: 10.1177/1087057115594589 | PubMed: 26170255