ICCVAM Biennial Report 2024-2025

Current regulatory decision-making for potential thyroid-disrupting chemicals is based on in vivo apical endpoints, such as alterations in thyroid hormone levels. New approach methodologies (NAMs) to complement or replace traditional in vivo tests are being developed to enable higher-throughput mechanistic understanding of potential endocrine-disrupting chemicals.

A key element in the acceptance of NAMs in regulatory contexts is to establish confidence through validation studies that evaluate a NAM’s reliability and relevance for a specific application. NICEATM coordinated an interlaboratory validation study of the utility of a thyroid 3D microtissue model for evaluating chemical effects on thyroid hormone synthesis and tissue viability. The lead laboratory was within the EPA Office of Research and Development, with three agrochemical industry and commercial laboratories participating in the transferability phase (Deisenroth et al. 2026). Goals of this effort included (1) development of the study design, (2) test method harmonization and standardization and demonstration of intralaboratory reproducibility, and (3) method transferability, reference chemical testing, and demonstration of interlaboratory reproducibility. Reference chemical selection was based on previous work demonstrating disruption of thyroid hormone synthesis. These included one “true negative” and three “true positive” compounds. The study established standard operating procedures that were successfully transferred to multiple laboratories with no experience conducting the assay. Consistent results were generated among the testing laboratories using the same four blinded reference chemicals and among the three donor pairs. Independent human thyrocyte donors were matched between a testing laboratory and the lead laboratory to account for biological variability. In addition to the publication describing the study, a detailed report of the study is being prepared. A follow-up study to test the model with a larger chemical set and potentially improve replicability is currently being designed.

Skin sensitization, or an allergic skin reaction, is an important health concern in chemical testing. Assessment of skin sensitization hazard traditionally relies on animal models, such as the local lymph node assay (LLNA). Efforts to reduce animal use have led to the development of alternative methods based on the skin sensitization adverse outcome pathway (AOP).

The electrophilic allergen screening assay (EASA) is a promising in chemico method to identify substances that covalently bind to skin proteins, Key Event 1 in the AOP for skin sensitization. The assay was converted from a cuvette-based platform into a 96-well platform and characterized previously (Petersen et al. 2022). A multilaboratory validation study assessed intra- and interlaboratory reproducibility of the 96-well EASA. In addition to the CPSC/NIST lead laboratory, laboratories sponsored by FDA, DoW, and NIEHS participated in the validation study. Each participating laboratory confirmed transferability and assessed 20 reference chemicals. Results showed that the method is reproducible and reliable, supporting its potential inclusion in regulatory frameworks. A report of the validation study was published in August 2025, and a paper summarizing the study has been submitted for publication (Reinke et al. 2026).

Key Event 1 assays are incorporated in defined approaches (DAs) for skin sensitization hazard and potency assessments described in Test Guideline (TG) 497 issued by the Organisation for Economic Co-operation and Development (OECD; OECD 2025a). These DAs offer structured, fixed procedures for data interpretation, reducing reliance on expert judgment. By meeting performance standards for in chemico Key Event 1 assays, described by OECD in TG 442C (OECD 2025b), the EASA is a promising alternative to existing Key Event 1 assays. The EASA is currently under consideration by OECD for inclusion in TG 442C. Compared to the current TG 442C assays, the EASA offers higher throughput and requires less specialized equipment, making it more accessible and efficient. To support its inclusion in TG 442C, the OECD Skin Sensitization Expert Group has requested additional testing for predictive capacity and confirmation of molecular interaction between the probe and reference chemicals; this testing is in progress by NIST and the NIEHS-sponsored laboratory.

Scientists in the USAF Force Health Protection program’s Predictive Risk Team (PRT) performed a study using NAMs to characterize the risk of allergic contact dermatitis from occupational chemical exposures (Keebaugh et al. 2025). The goals of the study were to evaluate the sensitization potential of chemicals used by fabricators and maintainers and to develop preliminary candidate surface guidelines, i.e., guidelines for maximum surface concentrations to prevent the induction of sensitization. The kinetic direct peptide reactivity, KeratinoSensTM, and human cell line activation test assays were used to predict LLNA effective concentration values using the regression models published in Natsch et al. (2022a) and Natsch et al. (2022b). In vitro assay results, in silico model predictions, and available human and animal data were used to predict sensitizer status using a weight-of-evidence approach. Preliminary candidate surface guidelines, derived from predicted effective concentration values for predicted sensitizers, can be used to evaluate the presence of these chemicals on work surfaces to better understand the risk of USAF personnel developing allergic contact dermatitis from occupational exposures.

Skin sensitization is an allergic reaction triggered by chemical exposure, which can cause significant health issues. To ensure safety, regulatory frameworks rely on methods to assess both hazard, the likelihood of a chemical to be a sensitizer, and potency, a prediction of the severity of sensitization that could be caused by the chemical. Traditionally, sensitization hazard and potency have been estimated using animal models, such as the LLNA. However, advances in nonanimal methods, such as in chemico, in vitro, and in silico tests, have enabled more ethical and human-relevant approaches through integrated strategies such as DAs, which combine data from multiple nonanimal tests to predict sensitization potential.

A May 2025 reportsummarized a NICEATM evaluation of test substances nominated by ICCVAM member agencies. In the study, 181 “difficult” test substances of regulatory interest were tested in individual in chemico and in vitro methods, and the results applied to three DAs for predicting skin sensitization hazard or potency. The in vitro methods evaluated were the direct peptide reactivity assay, the KeratinoSens™ assay, and the human cell line activation test. The three DAs used were the Key Event 3/1 Sequential Testing Strategy, the Two-out-of-Three strategy, and the Integrated Testing Strategy. Predictions were compared against reference data from LLNAs and human predictive patch tests (HPPTs) to assess concordance.

Results showed varying concordance with human and LLNA reference data. Individual methods had concordance rates ranging from 30%–93% and 52%–78% with LLNA data and human data, respectively. The DAs demonstrated better agreement among themselves (70%–92%) and showed higher concordance with human reference data (42%–47%) compared to the LLNA (24%–51%). Importantly, the DAs proved to be more accurate predictors of human skin sensitization hazard and potency than LLNA methods.

These findings confirm that DAs provide a more reliable, efficient, and ethical alternative to animal models for assessing skin sensitization. By integrating multiple nonanimal methods, DAs improve human health predictions, supporting regulatory decision-making and reducing reliance on animal-based testing.

NICEATM scientists have been working to replace the use of animal models in skin sensitization studies with reliable and ethical alternatives. A June 2025 NICEATM report summarized an evaluation of the performance of a nonanimal test method called the Genomic Allergen Rapid Detection (GARD™skin) assay, which tests whether chemicals can cause allergic skin reactions. The study specifically assessed the GARDskin assay’s ability to predict both hazard, the likelihood of a chemical to be a sensitizer, and potency, a prediction of the severity of sensitization that could be caused by the chemical, utilizing defined approaches (DAs) for skin sensitization. Results were compared against traditional animal test results, including LLNA and HPPT data.

The study examined 31 substances using the GARDskin assay, in conjunction with results from other methods, such as the direct peptide reactivity assay, the human cell line activation test, and the KeratinoSens assay. Predictions from these tests were compared against human and LLNA data to evaluate accuracy (an overall assessment of performance), sensitivity (ability to identify true positives), and specificity (ability to identify true negatives). The GARDskin assay performed well, ranking highest among the methods in terms of accuracy (63%), balanced accuracy (62%), and sensitivity (81%) when compared to LLNA data. It also had the lowest rate of false negatives (19%). Inclusion of GARDskin data improved the performance of DAs, which combine data from multiple nonanimal tests to predict sensitization potential. For potency, the DA containing GARDskin had higher accuracy than other DAs (53%) and was less likely to overpredict or erroneously assign a more restrictive hazard classification compared to other approaches.

While only 31 substances were tested, limiting broader conclusions, this evaluation shows promising results for the GARDskin assay in providing human-relevant, reliable predictions. Combining multiple tests into integrated approaches adds confidence in safety assessments, helping advance nonanimal methods for regulatory use. This study supports the shift toward modern, ethical testing alternatives for chemical safety evaluations.

Skin irritation and corrosion testing for pesticides is traditionally conducted using the Draize rabbit test, which is required by many regulatory agencies. While the Draize test generally provides classifications designed to protect human health, it has limitations due to species differences between rabbits and humans, particularly for mild to moderate irritants. NAMs offer human-relevant alternatives, using reconstructed human tissues, biochemical skin models, and in vitro systems to better mimic human skin responses. However, these methods face challenges in regulatory adoption, mainly due to limited concordance with Draize test results and requirements to distinguish irritation potential across all hazard categories.

EPA researchers and collaborators evaluated how existing NAMs align with human skin anatomy, physiology, and the biological mechanisms underlying chemically induced irritation and corrosion (Raabe et al. 2024). The findings show that NAMs such as reconstructed human skin tissue models and synthetic barrier systems are biologically and mechanistically relevant to these endpoints and reflect human skin responses better than animal models. Notably, retrospective analyses revealed that the Draize test has low reproducibility (less than 50%) for mild and moderate irritants, with many replicate predictions varying across irritation categories. Based on these findings, the study recommends improved testing strategies for pesticide classification by using a combination of OECD-approved in chemico and in vitro methods that closely replicate human skin structure and function. Specifically, “top-down” approaches prioritize NAMs that identify corrosive substances, while “bottom-up” strategies improve sensitivity to accurately differentiate mild irritants or nonirritating substances from irritants and corrosives. Additionally, optimizing cytokine release protocols and analyzing reference test results could enhance discrimination within the mild irritation spectrum. This work underscores the potential of modern NAMs to replace animal models for skin irritation, advancing both regulatory science and human health protections by prioritizing more relevant and reproducible testing methods.

Skin permeation is a primary consideration in the safety assessment of cosmetic ingredients, topical drugs, and human users handling veterinary medicinal products. While excised human skin remains the standard model for in vitro permeation testing studies, unreliable supply and high cost motivate the search for alternative skin barrier models. FDA, NCATS, and NICEATM developed a standardized dermal absorption testing protocol to evaluate the suitability of alternative skin barrier models to predict skin absorption in humans (Salminen et al. 2023). Using this protocol, a reconstructed human epidermis model (EpiDerm-200, MatTek) and a synthetic barrier membrane model (Strat-M, Sigma–Aldrich) were assessed concurrently with standard excised human skin models. Ongoing follow-up studies to this project began in 2024. The number of reconstructed human skin models evaluated was expanded to seven (four reconstructed human epidermis and three full-thickness skin models; three batches per model), and the dermal chemical exposure time was increased from 6 to 24 hours. Caffeine, salicylic acid, mannitol, and testosterone were chosen as test chemicals for both models because of the abundance of historical in vitro and in vivo data and because they represent a wide range of physicochemical properties known to influence absorption. The endpoints assessed included permeability, transepidermal water loss, and histology. Tests were also run to characterize the lipid composition of the stratum corneum across skin models. All the in vitro models exhibited native human epidermis-like morphology, including a characteristic stratum corneum. The excised human skin models showed the lowest transepidermal water loss, with the Strat-M and reconstructed human epidermis models exhibiting a range of water loss equivalent to or higher than that of excised human skin. The full-thickness skin models exhibited the highest water loss and chemical permeation, making them the weaker skin barrier models. Overall, the results suggested that the reconstructed skin models have a weaker barrier than excised human skin, although the differences vary by model and test substance. Studies such as these to evaluate novel alternative skin barrier models may advance regulatory applicability of these models, potentially increasing the efficiency of regulatory review of new products. This project will be presented at the 2026 Annual Meeting of the Society of Toxicology (Modaresi et al.; Abstract 5193) and the American Society for Mass Spectrometry 2026 Conference on Mass Spectrometry and Allied Topics (Sun et al.; Abstract WP499). A peer-reviewed publication is in preparation for submission in 2026.

In 2025, NICEATM, PETA Science Consortium International e.V., EPA, and CropLife America member companies completed a collaborative effort to develop DAs for predicting United Nations Globally Harmonized System of Classification and Labelling (GHS) and EPA eye irritancy hazard classifications.

Using data from 29 agrochemical formulations tested in a common set of in vitro or ex vivo methods, four DAs were developed:

• The bovine corneal opacity and permeability assay (BCOP; OECD TG 437) with or without histopathology.
• The EpiOcular eye irritation test (OECD TG 492), followed by the BCOP with or without histopathology.
• The SkinEthic time-to-toxicity test for liquids (OECD TG 492B), followed by the BCOP with or without histopathology.
• The EyeIRR-IS assay (Cottrez et al. 2021), followed by the BCOP with or without histopathology.

Rather than evaluating direct concordance with historical in vivo rabbit eye test data, GHS and EPA predictions derived using the DAs were assessed using orthogonal validation and weight-of-evidence. Results showed that, for both classification systems, the predictions from each of the four DAs were as protective or more protective of human health than the in vivo test (Daniel et al. 2025). This study refines and builds upon two DAs described in a previous publication (van der Zalm et al. 2024). This work underscores the limitations of current animal-based reference data and highlights the need for a more human-relevant approach to eye irritation hazard classification. By improving integrated testing strategies, this research lays the groundwork for safer, more ethical, and scientifically meaningful alternatives to animal testing in regulatory toxicology.

The application of pesticides to control invasive species remains a key component of integrated pest management. Historically, developing new chemical controls required extensive testing in live animals. In vitro cytotoxicity assays using mammalian or fish cell lines are increasingly utilized because they reduce the number of animals required and enable high-throughput screening (HTS) of candidate chemicals. However, it is still unclear how well in vitro cytotoxicity values represent in vivo toxicity values.

Scientists in the DOI U.S. Geological Survey have developed and evaluated fish cell line cytotoxicity assays for several indigenous and nonindigenous species to demonstrate proof-of-concept for pesticide screening targeting nuisance fish. Recent work evaluated antimycin A using the RTgillW1 cell line for rainbow trout (Oncorhynchus mykiss), following OECD TG 249 and using the Promega CellTiterGlo 2.0 assay. Results showed that in vitro–in vivo toxicity ratios were approximately one for most endpoints. In addition, a grass carp (Ctenopharyngodon idella) gill cell line assay demonstrated repeatability using the OECD positive control chemical 3,4-dichloroaniline. Together, these proof-of-concept assays and the strong agreement between in vitro and in vivo responses indicate that fish gill cell lines can reliably predict whole-animal toxicity. Continued development of these systems shows strong promise for pesticide candidate screening and related scientific applications.

USAF Predictive Risk Team (PRT) is developing a new approach methodologies-based approach to support rapid risk assessment of novel and poorly characterized chemicals. A tiered approach is being evaluated, where lower tiers use in silico and high-throughput in vitro models to prioritize hazards, while higher tiers use organotypic in vitro models to perform quantitative risk assessments. To test the utility of this NAM-based approach on operationally relevant scenarios, researchers conducted a case study using several chemicals with known liver and central nervous system toxicity. The chemicals were assessed using the Edwards et al. 2022 in silico model (Tier 0) to evaluate mechanism of action and acute systemic toxicity. In vitro tests were then performed using assays appropriate for the mechanism of action identified by the in silico model (Tier 1A). Additional in vitro tests evaluated cell health (cell viability, cytokines, oxidative stress) in glial and liver monoculture models (Tier 1B). Finally, Tier 2 human liver microtissue (3DInsight) and central nervous system (CNS-3D, 28bio) organoid models were used to evaluate tissue functional effects. Points-of-departure (PODs) were derived using benchmark dose modeling, converted to human-equivalent doses using toxicokinetic modeling, and compared to traditional testing data. Human-equivalent doses were generally more conservative than in vivo acute oral toxicity values, indicating that use of this metric to set exposure limits would be at least as health-protective as using metrics derived from traditional approaches. Neurotoxicity human-equivalent doses were generally similar to, or more conservative than, literature-derived traditional PODs. Liver microtissue testing results were also qualitatively representative of in vivo effects, and quantitative analysis of human-equivalent doses is ongoing. Findings from the project were presented at the 2025 and 2026 Society of Toxicology meetings and the 2025 Military Health Systems Research Symposium, and two publications are planned for 2026.

Supported by funding from the Defense Threat Reduction Agency’s Joint Science and Technology Office, the U.S. Army Chemical Biological Center (CBC) is exploring the use of organ-chip technologies and other NAMs to advance medical countermeasures for chemical toxicity syndromes. The first project aims to create a detailed database of cellular markers from various organs exposed to chemical warfare agents or other toxic compounds. This will help identify new agents by analyzing overlapping mechanisms and symptoms, potentially leading to new drug targets for medical countermeasures development. The second study will identify the most predictive NAMs and develop a testing protocol or toolkit that will allow the rapid identification of potential neurotoxicants and promising medical countermeasures without extensive animal testing. NAMs that target key events in each of the endpoints of interest will be tested against known neurotoxicants, and in vitro qualitative and quantitative responses will be compared to concurrent in vivo studies to determine the most reliable and sensitive assays for inclusion in a testing protocol. Both efforts will result in the digital curation of endpoint-specific biomarkers that can be used by researchers across DoW to identify additional targets for medical countermeasure development.

Comparing in vitro results to in vivo data is necessary to establish suitability of the in vitro models as a replacement for in vivo acute lung toxicity for volatile chemical hazards. USAF PRT, in collaboration with the Air Force Research Laboratory’s MPS team and the Naval Medical Research Unit Dayton, are evaluating the mechanistic relevance and the quantitative predictivity of in vitro lung models by comparing dose responses for key events to in vivo data following acute exposure to sulfur dioxide. Key events in the acute lung injury AOP, including oxidative stress, cytotoxicity, inflammation, and barrier integrity, are being measured in monoculture cell models (A549, HuArlo) and human reconstructed tissue models (MatTek EpiAirway™ and EpiAlveolar™; Epithelix MucilAir™) using gas exposure at the air–liquid interface. In vitro data collection is nearly complete. Concurrent studies in a rat model with corresponding sulfur dioxide exposures and correlated endpoints were recently completed at the Naval Medical Research Unit Dayton. Dose–response modeling is being used to identify points-of-departure from each endpoint, and in vitro to in vivo extrapolation (IVIVE) is being used to describe gas disposition in the lung (Tsujino et al. 2005; Nowak et al. 2025). Assays and measured endpoints that are most predictive of in vivo response will be incorporated into a recommended rapid chemical testing protocol. These NAMs are expected to facilitate scientifically sound chemical risk assessments on a timescale that would realistically support operational and acquisition decisions.

Rapid advancements in engineering have driven the need to understand human exposure to ultrafine particles and volatiles. To this end, U.S. Air Force (USAF) scientists developed an in vitro inhalation model to quickly assess toxicity. An alveolus-on-a-chip platform was developed using primary alveolar epithelial and microvascular endothelial cells cultivated at the air–liquid interface. The platform incorporates cyclical stretch and breathing mechanics to better emulate human physiology in a laboratory model. This model has been utilized to identify unique biomarkers of soot-based exposures, an emerging respiratory global concern to human health and performance. Initial results after treating the alveolus-on-a-chip platform with diesel particulate matter resulted in increased cytotoxicity and a statistically significant decrease in lung–blood barrier integrity. Extracellular vesicles isolated from particulate matter-exposed alveolus-on-a-chip platforms exhibited increased RNA content compared to the controls, suggesting that exposure alters the signaling cargo released by the lung. These exosomes were successfully labeled and traced for uptake by recipient cells. Multiple downstream models, including liver, skin, and immune cell systems, showed significant exosome uptake, supporting a mechanism for systemic communication beyond the lung following alveolar exposure. RNA from the recipient models after exosome exposure was isolated for global transcriptomic profiling. Investigators now aim to characterize pathways by which soot-based exposures drive whole-body effects through exosome-mediated transfer of molecular information to aid in biomarker identification for countermeasure development.

This project aims to identify a suite of NAMs for predicting the qualitative and quantitative health risk from exposure to fine respirable particles (PM_2.5, or particulate matter less than 2.5 micrometers in diameter) using crystalline silica dioxide as a test material. Cytotoxicity, oxidative stress, tissue barrier disruption, and injury dose–response will be compared between concurrent in vitro and in vivo test models. Acute exposure and repeated exposure (daily for 1–2 weeks) experiments are currently nearing completion in HuArlo, BEAS-2B, Epithelix MucilAir, and MatTek EpiAlveolar models in studies performed by the Air Force Research Laboratory MPS Team. Points-of-departure (PODs) will be derived for each endpoint. In vitro PODs will be converted to in vivo equivalent doses using multiple path particle dosimetry modeling, which will then be compared to in vivo exposures performed at the Naval Medical Research Unit Dayton. USAF PRT is performing multiple path particle dosimetry modeling, benchmark dose modeling, and risk assessment strategy. The product is a predictive risk capability that will facilitate rapid, efficient, and scientifically grounded decisions for ensuring Force health protection in complex operational environments.

Researchers at Battelle Memorial Institute are testing dose responses for multiple key events in lung injury using the EpiAirway reconstructed human airway in vitro model with the prototype lung irritants propylene oxide and epichlorohydrin. The goal of this project is to identify a suite of NAMs for testing lung toxicity risk from operational exposures to inhaled hazards. EpiAirway cells were exposed at the air–liquid interface to propylene oxide for four hours at concentrations of 0, 350, 750, or 1,250 ppm in a VITROCELL 24/48 exposure system. Initial range-finding studies were used to identify the three propylene oxide concentrations used in the definitive study. EpiAirway tissues exposed to propylene oxide demonstrated a concentration-dependent increase in cytotoxicity, barrier stress, cytokine production and oxidative stress in human bronchial reconstructed airway models. All endpoints were altered at 1250 ppm propylene oxide, with no effects at 350 ppm and only the most sensitive endpoints (e.g., adenylate kinase-indicated cytotoxicity, cilia loss) being disrupted at 750 ppm. A similar experiment using epichlorohydrin is currently ongoing. USAF PRT is comparing the dose–response results to in vivo data to evaluate the mechanistic relevance, predictivity, and repeatability of the in vitro model for testing reactive gases in support of health risk assessments for novel chemicals.

To reduce animal use in the investigation of human-relevant inhalation toxicity, the NIEHS Division of Translational Toxicology is evaluating the use of in vitro lung models, including air–liquid interface airway tissues. Potential applications of these models include screening-level assessments to help predict the adverse airway effects of inhaled substances and prioritize them for further testing.

Initially, proof-of-concept studies were conducted to characterize and optimize the VITROCELL 48 2.0 Plus exposure system with human and rat tracheobronchial-derived air–liquid interface airway tissues. The chemical used was 2,3-pentanedione, a highly volatile component of artificial butter flavorings that has been associated with airway injury and fibrosis via occupational inhalation exposure. The studies were designed to explore the validity of extrapolations among rat in vivo, rat in vitro, and human in vitro airway effects in these models. They showed that exposure of the air–liquid interface tissues to 2,3-pentanedione vapor induces concentration-dependent changes in relevant airway toxicity endpoints. Changes in transepithelial electrical resistance, cytotoxicity, histopathology, and secreted biomarkers were comparable between species for most of the endpoints. Measurements of 2,3-pentanedione concentration were also performed on apical wash, basolateral culture media, and tissue lysate samples as indicators of dosimetry. One paper describing these studies has been submitted for publication and a second is in preparation. Further method development and optimization work are also in progress, including the use of alveolar-derived air–liquid interface models and incorporation of transcriptomics.

To help better assess the utility and application of in vitro lung (including air–liquid interface) models, NIEHS issued a Request for Information in April 2024 to gather information on the use of these models in inhalation toxicology research including knowledge gaps and technical challenges and limitations. Most responders indicated that they currently use in vitro lung models for inhalation/respiratory toxicity screening, although many also use them for chemical-specific mechanistic studies or IVIVE modeling. There is also interest in developing and validating the models for regulatory decision-making. The primary challenge or limitation identified in the use of these models was reproducibility (encompassing both data variability and method standardization), but other concerns included dosimetry (including test article generation and exposure), physiological/mechanistic relevance (model complexity, or lack thereof), model validation, and translational relevance/prediction (correlating in vitro model findings to data in vivo and in exposed humans).

To address the reproducibility concern, the NIEHS Division of Translational Toxicology has assembled a steering committee to oversee an interlaboratory reproducibility study that will involve multiple independent testing facilities. The steering committee is determining test article generation and exposure systems and the source of air–liquid interface airway tissues used, appropriate chemistry and toxicological endpoints, standardizing the protocol according to the approach used by the Division of Translational Toxicology testing facility, formulating a plan for statistical analyses, and selecting three to five partnering facilities (federal or nonfederal) to conduct testing using 2,3-pentanedione.

Mast cells rapidly release a variety of mediators such as histamine, heparin, proinflammatory cytokines, and chemokines in response to foreign substances or endogenous damage-associated molecular patterns. These factors facilitate immune cell recruitment and play a key role in pulmonary toxicity due to mast cell locations within airways and mucosal surfaces of the lungs. The pulmonary system is an ideal exposure route for chemical and biological agents, making mast cells a prime choice to assimilate into a lung-on-a-chip.

The U.S. Army CBC is incorporating immune components into commercially available MPS systems. CBC collaborated with the University of Colorado Anschutz Medical Campus to integrate a mast cell immune component into the Emulate system. CBC scientists observed significant increases in specific cytokine and chemokine production compared to chips without mast cells, which aligns with literature reporting. The findings from this study are captured in CBC Technical Report DEVCOM-CBC-TR-1941 “Incorporation of Immune Response into Organ-on-a-Chip Technology”.

This work has contributed to substantial efforts to incorporate other immune cells into the system to build a complete model of immune system infection to study prevention and pathogen defense. CBC recently received funding to establish a multiorgan chip system capable of recapitulating components of both the innate and adaptive immune system. Using the TissUse Chip3, they will combine liver, lung, and lymphoid follicle tissues along with peripheral blood mononuclear cells and other resident immune cell types. The long-term goal is to use this chip for viral infection challenge and vaccine candidate screening. Additionally, CBC has been working to bioprint human lung and skin tissues, complete with tissue-specific macrophages. This project aims to not only increase physiological accuracy for these tissue types but will also provide a more robust model of immune response to microbial infections, which is largely lacking in most commercial organ-chip systems.

The FDA Center for Drug Evaluation and Research has marked a major milestone by accepting the first organ-on-a-chip technology into its Innovative Science and Technology Approaches for New Drugs (ISTAND) Program. This program promotes the use of innovative tools to improve how new drugs are developed. The accepted technology, called the Liver-Chip, is an MPS designed to predict drug-induced liver injury (DILI), a common reason drugs fail during early testing.

The Liver-Chip recreates the human liver environment by growing four types of liver cells in a small, engineered device. These cells mimic the natural physiology and mechanical forces they experience in the human body. This setup allows scientists to monitor in real time how drugs may cause liver damage compared to safe conditions. Importantly, the Liver-Chip evaluates whether a drug poses less, similar, or more risk of liver injury than other drugs in the same class. By improving preclinical testing, it can provide better data for drug investigational submissions and can help identify safer treatments earlier in development.

Initial Liver-Chip assessment has revealed the following:

• Conventional toxicity assays do not perform well in liver MPS.
• ALT activity is not stable in the Liver-Chip.
• Lactate dehydrogenase activity is prone to interference by medium components.
• ELISA is recommended for toxicity measurement in liver MPS.
• The chemical method is recommended for the urea assay in the Liver-Chip, offering adequate and improved specificity and sensitivity.

These results highlight the importance of reliable and reproducible assays for translational DILI biomarkers in liver MPS. Work is ongoing, demonstrating FDA’s commitment to advancing new technologies, such as organ-on-a-chip systems, that have the potential to bring safer and more effective therapies to patients while reducing reliance on animal testing.

Compared to conventional cell culture approaches, liver-on-a-chip models have the potential to better maintain cultures of primary hepatic cells, the current gold-standard model for drug metabolism and hepatotoxicity, and improve the prediction of DILI. Albumin, urea, ALT, AST, and LDH are proposed translational biomarkers for the prediction of DILI using liver MPS such as the Emulate® Liver-Chip. However, the performance of commonly used assays for these biomarkers in liver MPS may differ from clinical samples, posing a significant challenge in using NAMs data to predict clinical responses. 

ALT and AST are both FDA-endorsed clinical biomarkers for DILI. While using the Liver-Chip, scientists in the FDA National Center for Toxicological Research (Shi et al. 2025) observed that when these biomarkers were measured using extensively validated reference procedures, the activity of ALT but not AST spontaneously decreased by approximately 50% every 24 h. Furthermore, ALT and AST activity remained below the detection limit (5 U/L) after acetaminophen treatment, when their protein concentrations were increased by approximately 20-fold as determined by ELISA. The protein concentrations of ALT and AST, but not their activity, showed good correlation with LDH activity, which was nearly eliminated after a freeze-thaw cycle. For viability of nonparenchymal hepatic cells, an unusually high background signal prevented the use of LDH activity assays; this was largely attributable to CultureBoost and fetal bovine serum in the perfusion medium, which are required to maintain key cell functions. However, LDH ELISA appeared to be a useful alternative. Of two widely used urea assays, the enzymatic approach was profoundly affected by medium supplements, wherein sample dilution increased the limit of detection, making it impossible to observe drug-induced urea inhibition. In contrast, the chemical assay offered adequate and improved specificity and sensitivity. For albumin, an ELISA had to be performed since routinely used dye-binding methods were not sensitive enough.

These findings provide a plausible explanation for many inconsistent reports in the literature and highlight the significance of establishing reliable and reproducible assays for translational DILI biomarkers in liver MPS. Efforts are ongoing and will continue into 2026 and beyond to examine the performance of newly FDA-qualified and promising emerging clinical DILI biomarkers in liver NAMs models. 

DILI is a critical safety concern in drug development. The FDA is committed to reducing reliance on animal models in preclinical safety studies by promoting the use of NAMs. A review by scientists in the FDA NCTR evaluated the performance of NAMs, animal studies, and advanced MPS in predicting DILI risks (Shrimali et al. 2025). The study found that while in vitro NAMs showed advantages over traditional animal models in some contexts, there was significant variability in their predictive performance. Furthermore, the limited availability of data for MPS, a cutting-edge technology simulating human tissue-level responses, restricted the ability to make meaningful comparisons between these and other approaches. Overall, these findings highlight both the promise and challenges of integrating NAMs into DILI testing.

To systematically improve and benchmark NAMs for DILI prediction, the researchers proposed "DILIference", a curated reference drug list compiled from scientific literature. DILIference provides a standardized set of compounds with established liver toxicity profiles, serving as a critical resource for evaluating and refining DILI-predictive NAMs.

This work emphasizes the need for consistent benchmarks and robust datasets to further advance NAMs for regulatory use. By leveraging NAMs and tools like DILIference, researchers and regulatory bodies can reduce reliance on animal models while enhancing the reliability and human relevance of liver toxicity predictions in preclinical drug development.

Researchers at FDA NCTR used complementary human liver cell systems to examine toxicity mechanisms of CBD and its analogs and metabolites. Together the studies illustrate the value of selecting the appropriate in vitro liver model for resolving metabolism-dependent versus metabolism-independent toxicity mechanisms.

• Chen et al. (2025) used mRNA sequencing in HepG2 cells to identify key cellular responses to CBD and its metabolites 7-hydroxy-CBD and 7-carboxy-CBD. Direct enzymatic assays confirmed strong inhibition of mitochondrial respiratory complexes IV and V by CBD and metabolite 7-hydroxy-CBD, with mitochondrial dysfunction further validated by glucose-galactose assay and loss of membrane potential by JC-1 staining. Autophagy, evidenced by autophagosome formation and enhanced autophagic flux, was the common cytotoxic mechanism shared across all three compounds, while mitochondrial complex inhibition was specific to CBD and 7-hydroxy-CBD.
• Campasino et al. (2024) exposed iPSC-derived hepatocytes to CBD and hemp extract across a concentration range of 0.1 to 30 μM to assess hepatic viability and function. Overt cytotoxicity was observed only at 30 μM, but subcytotoxic concentrations of 0.1 to 3 μM produced significant changes in CYP enzyme activity, mitochondrial membrane potential, and lipid accumulation. Concentrations of 3 μM and greater additionally affected albumin and urea secretion, caspase 3/7 activity, and glutathione levels. The iPSC-derived hepatocyte model was thus able to detect functionally relevant perturbations below the cytotoxicity threshold, supporting its utility for early-stage safety evaluation of botanical compounds.
• Primary human hepatocytes and HepG2 cells were used by Li et al. (2025) alongside a panel of HepG2 lines individually overexpressing 14 human CYP enzymes to characterize the toxicity and metabolism of 7-hydroxy-CBD and 7-carboxy-CBD. Both metabolites induced cell cycle disturbances, apoptosis, and ER stress comparable to the parent compound. CYP2D6 was identified as the enzyme responsible for metabolizing 7-hydroxy-CBD, and its activity attenuated the associated cytotoxicity, apoptosis, and endoplasmic reticulum stress; 7-carboxy-CBD was not a substrate for any of the 14 CYPs tested. The study highlights how CYP-overexpressing cell lines can resolve metabolism-dependent differences in toxicological outcomes between structurally related metabolites.

Many promising medications have failed to demonstrate safety and efficacy in human clinical trials despite encouraging results in preclinical studies. MPS, also known as tissue chips or organs-on-chips, offer a promising approach to generate more predictive preclinical data and improve the likelihood of clinical success. Built from human cells, MPS can replicate the structure and function of major organ systems, including the heart, kidneys, and lungs. Scientists are increasingly using tissue chips to evaluate drug effects on human tissues in a faster and more physiologically relevant manner than traditional preclinical methods, such as 2D cell cultures and animal models.

Since 2011, NCATS has supported the development and application of tissue chips for assessing drug safety and efficacy, modeling both rare and common diseases, informing the design of human clinical trials, and advancing precision medicine. In 2024, NCATS established four Translational Centers for Microphysiological Systems (TraCe MPS) to promote the broader adoption of tissue chip technologies, particularly in drug discovery and development (RFA-TR23-001). TraCe MPS is a partnership with the FDA ISTAND Program, with additional collaboration from the Critical Path Institute. The goal of the partnership is to qualify MPS, within defined contexts of use, as FDA-designated drug development tools. The overarching objective is to develop fit-for-purpose MPS platforms with clearly defined contexts of use that can support regulatory decision-making for new drugs and therapeutics.

The four Translational Centers are located at Texas A&M University in partnership with the University of Texas Medical Branch; the University of Pittsburgh; the University of Rochester in collaboration with Duke University; and the University of Washington in partnership with the Icahn School of Medicine at Mount Sinai. Current funded projects include models for kidney function (University of Washington and Mount Sinai), DILI and metabolic-associated steatotic liver disease (University of Pittsburgh), barrier tissues such as the blood–brain barrier (University of Rochester and Duke University), and placental–fetal development (Texas A&M University and the University of Texas Medical Branch). In addition to engagement with the FDA’s ISTAND Program, NIH-funded awardees collaborate with biotechnology and pharmaceutical partners to refine contexts of use, broadening the pathway for adoption of MPS technologies. 

Several translational centers have achieved notable progress within the ISTAND Program, with submissions publicly available on the FDA ISTAND website to support transparency and advancement in this rapidly evolving field. Accepted Letters of Intent for these models include:

• Two from the University of Pittsburgh: one focused on using their model to inform drug dosing in clinical trials for liver disease (DDT-IST-000034), and another on quantifying hepatotoxicity to support dosing decisions (DDT-IST-000044)
• One from the University of Washington and Mount Sinai team for assessing relative nephrotoxicity using a human kidney chip (DDT-IST-000045)
• One from the Texas A&M and University of Texas Medical Branch team for using a human chorio-decidual interface organ-on-chip to derisk positive rodent developmental and reproductive toxicity findings for novel investigational drugs (DDT-IST-000047)

Collectively, these efforts demonstrate the application of MPS technologies across multiple organ systems and regulatory contexts, including dose selection, toxicity assessment, and risk deconvolution. As of the end of 2025, the TraCe MPS is positioned to make a substantial contribution to advancing regulatory acceptance and adoption of MPS-based approaches.

The ability to predict placental transfer of drugs administered during pregnancy is critically important to help manage the health of pregnant women and to include women of child-bearing potential in clinical trials. However, placental transfer assessments remain dependent on in vivo reproductive and developmental toxicology studies because there are no well-established NAMs-based approaches. Given the limitations of individual NAM platforms, a combinatorial approach integrating in silico and in vitro NAMs has greater potential for accurate predictions of placental transfer of drugs. FDA NCTR is evaluating the potential for an integrated approach using in silico, in vitro and ex vivo NAMs to predict placental transfer for a set of drugs. The in silico approach will combine rigorous model-building procedures utilizing 3D spatial molecular descriptors and advanced self-consistent algorithms free of overparameterization to generate accurate categorical models of placental permeability. A combinatorial NAMs approach to predict placental transfer could inform weight-of-evidence-based decisions to enable earlier inclusion of women of child-bearing potential in clinical trials and expand access for pregnant women to experimental therapeutics, while reducing reliance on animal data.

An FDA NCTR study investigated the combined effects of verapamil (a drug used to treat high blood pressure, angina, and certain heart rhythm disorders) and sodium arsenite (a form of arsenic used in some pesticides, antiseptics, dye manufacturing, and chemical laboratories) on zebrafish (Danio rerio) embryos (Twaddle et al. 2025). Researchers aimed to understand how the interaction between these two chemicals affects development since verapamil can alter arsenic processing in the body. The results showed that zebrafish embryos exposed to sodium arsenite or verapamil alone survived without any major issues. However, when embryos were treated with both verapamil and moderate levels of sodium arsenite, survival rates decreased significantly to 23% and 17%, and the embryos showed developmental defects and heart damage. Interestingly, using a lower verapamil dose with arsenite resulted in no embryo deaths but caused developmental defects (cardiac and growth). Further analysis revealed that verapamil caused higher arsenic levels in the embryos by potentially blocking its removal. Verapamil also slightly reduced arsenic-induced cell death in the main body region (not yolk sac) but caused developmental and heart-related toxicity when paired with arsenic. However, the addition of acetyl-L-carnitine, a compound known to protect cells and mitochondria, completely prevented these toxic effects.

This study is the first to show that acetyl-L-carnitine can counteract the harmful effects of combined verapamil and arsenic exposure in zebrafish embryos, providing valuable insights into the mechanisms of these toxicities and potential therapeutic approaches. A second study showed N-acetylcysteine, another antioxidant and dietary supplement, also eliminated arsenic-induced cell death in zebrafish embryos (Gu et al. 2026). The zebrafish model has proven to be a useful tool for studying toxic elements, chemical interactions and their developmental impacts. Further studies are ongoing to determine the combined effects of arsenic and cadmium on development and the nervous system.

Prenatal developmental toxicity research explores how environmental substances, drugs, and chemicals can negatively affect the growth of embryos and fetuses. Traditionally, these studies have relied on animal models, but ethical concerns and the need for human-relevant models have led researchers to explore new methods. One promising tool is pluripotent stem cells, which can differentiate into any type of human cell. Pluripotent stem cell models are increasingly used to simulate human development in the laboratory. Two-dimensional models are easier to use and can screen many chemicals quickly, but they lack the complexity of real human tissues. On the other hand, 3D models, such as organoids, create miniature versions of organs and tissues that mimic the conditions of a living body. While more realistic than 2D models, 3D models are challenging to produce and analyze, often requiring expensive tools and techniques.

A review by NIEHS scientists (Wu et al. 2024) highlights progress in the use of pluripotent stem cells to create models of brain and heart development, which are critical areas for understanding toxicity during pregnancy. Researchers are also combining pluripotent stem cell models with innovative technologies to improve testing methods and produce data that is more relevant to humans. These advancements help scientists predict potential developmental risks posed by chemicals, offering tools that are more ethical and accurate than traditional animal-based methods. Ultimately, these breakthroughs are paving the way for better decision-making in prenatal toxicity testing, helping regulators protect human health while advancing the science of developmental biology and toxicity assessment.

Organophosphorus flame retardants (OPFRs) are widely used chemicals that persist in the environment, yet limited information regarding their potential toxicity exists. Their structural similarity to organophosphate pesticides raises concerns about developmental neurotoxicity (DNT), particularly for human health. However, traditional in vivo testing methods are not suitable for evaluating the wide range of OPFRs that lack safety data.

To address this gap, NIEHS researchers applied an in vitro testing battery designed to assess key processes in neurodevelopment and function (Kreutz et al. 2024). Behavioral data from small model organisms were also included. This study used an IATA combining data from experimental assays, the NIH Integrated Chemical Environment (NICE) database, and existing literature. Human exposure data and physiologically based toxicokinetic models were used to compare in vitro toxicity thresholds with predicted human plasma concentrations.

Eight OPFRs, including aromatic OPFRs and halogenated flame retardants, were evaluated alongside two brominated flame retardants known to cause DNT. Results showed that several aromatic OPFRs (e.g., triphenyl phosphate and isodecyl diphenyl phosphate) demonstrated activity at concentrations comparable to the toxic brominated flame retardants, suggesting a need for further investigation. While the in vitro assays highlighted potential hazards, they provided limited insights into mechanisms of toxicity. By integrating additional information, endocrine disruption was identified as a potential mechanism of action for some OPFRs.

This study underscores the utility of an IATA for prioritizing chemicals of concern. It also found that human exposure to some OPFRs may result in plasma concentrations similar to those that trigger activity in vitro, raising concerns about potential human health risks. These findings provide a foundation for future investigations into OPFR safety and the mechanisms driving their neurotoxicity.

Chemical exposure is believed to play a role in neurodevelopmental disorders, but most chemicals have not been tested for DNT potential. Traditional testing methods rely on animal models, but these tests are expensive, require many animals, and sometimes produce results that are not reliable for regulatory decisions. To overcome these limitations, the NIEHS Division of Translational Toxicology launched the DNT Health Effects Innovation program in 2019 to improve how chemical risks to the developing nervous system are evaluated.

The program applied a DNT screening battery to test chemicals quickly and efficiently using modern methods. This battery includes both 2D and 3D in vitro tests with human and rodent cells to assess key brain development processes, such as cell growth, migration, neurite development, and neural network function. It also uses zebrafish (Danio rerio) embryos to study neurobehavioral effects. Over 300 chemicals nominated by stakeholders have been selected and distributed for testing using this battery.

A key accomplishment of the program is the creation of a unified data analysis system that integrates information from different tests. This system, which includes tools such as the Toxicological Prioritization Index (ToxPi), helps evaluate DNT hazards and prioritize chemicals for further study. The strategy demonstrates how the DNT screening battery can effectively identify chemicals for further evaluation, providing a systematic way to interpret results and guide decision-making. This innovative approach simplifies and modernizes DNT testing, ensuring timely, ethical, and scientifically reliable evaluations of chemicals for their effects on brain development. A paper describing the first testing phase of 95 chemicals has been accepted for publication in Neurotoxicology (Hall et al. 2026).

Two FDA NCTR studies used commercially available human fetal neural stem cells as a platform to assess the DNT potential of drugs often subject to abuse. 

• Neural stem cells derived from fetal hippocampal tissue were exposed to CBD, its major metabolites 7-OH-CBD and 7-COOH-CBD, and delta-9 tetrahydrocannabinol across a range of clinically relevant concentrations for up to 7 days (Latham et al. 2025). All compounds caused dose- and duration-dependent cell death, and CBD and 7-OH-CBD additionally reduced the G1-phase cell population in cell cycle analysis. Combination exposures of cannabidiol with 7-OH-CBD produced additive cytotoxicity, providing direct evidence that the metabolite contributes to overall neurotoxicity in vivo. Gene expression studies also suggested that prenatal cannabinoid exposure may impair astrocyte maturation and endocannabinoid signaling in the developing brain.
• Human neural stem cells and cells differentiated from them were exposed to fentanyl at concentrations of 0.5 to 100 μM for 24 hours. Standard cytotoxicity assays detected no significant cytotoxicity or mitochondrial impairment at any concentration tested. However, chromatographic lipidomic analysis of differentiated neural cells revealed dose-dependent effects indicating disruption of the ceramide/sphingolipid pathway at the highest dose. These findings demonstrate that lipidomic profiling can uncover early-stage neurochemical effects of opioid exposure that are invisible to conventional cell viability endpoints, highlighting the value of multiendpoint approaches when evaluating DNT (Wang et al. 2025).

Exposure to opioids during early pregnancy has been associated with a risk of neural tube defects in nonclinical animal models and human epidemiological studies. However, challenges stemming from study design limitations and conflicting outcomes in human and animal research have complicated risk assessment of the drug class. While animal studies are still required for embryo–fetal development assessments for therapeutics, the evolving regulatory landscape in drug development has recognized the value of NAMs to supplement current animal studies. Specifically, human iPSC-based models are increasingly being used in assessments of DNT potential.

FDA NCTR scientists used these models to recapitulate aspects of neural tube formation in culture to (1) determine which opioids pose greatest risk to the developing fetus, (2) identify the most susceptible neural cell types, and (3) investigate toxicological mechanisms of action without the need for animal studies. In 2024 and 2025, three human iPSC lines derived from healthy female donors were differentiated to neural progenitor cells using an optimized protocol for establishment of neural rosettes of embryoid bodies, three-dimensional aggregates formed by pluripotent stem cells. Opioids, including morphine, methadone, codeine, and buprenorphine, were tested in this protocol at concentrations ranging from 3x to 1000x the clinical maximum plasma concentration (Cmax). Opioid treatment had no clear effect on iPSC proliferation, differentiation, embryoid body development, or neural rosette formation at concentrations less than 30x Cmax. Methadone, however, inhibited neural rosette formation at concentrations exceeding 300x Cmax. Subsequent studies also identified potential adverse effects on neural differentiation and activity following high-dose methadone treatment. Follow-up gene expression analysis identified a potentially important pathway in methadone-induced DNT. This work was presented at the 2024 Annual Meeting of the Society of Toxicology (Cai et al., Abstract 4461), with peer-reviewed manuscripts anticipated for publication by 2027.

Safety concerns about general anesthetics, such as desflurane, have arisen due to evidence of potential harmful effects on the developing brain after early-life exposure. A study by FDA NCTR and the FDA Center for Drug Evaluation and Research used neural stem cells from the hippocampus of monkey brains to investigate how desflurane affects brain development processes, including cell health, growth, and differentiation into specialized brain cells, namely neurons and glial cells (Wang et al. 2025). The goal was to establish a direct biological link between historical in vivo nonhuman primate data and cellular responses, providing evidence that could help reduce or replace the need for future nonhuman primate animal studies.

When exposed to desflurane at clinically relevant levels (5.7%), the neural stem cells showed no notable damage or changes in their ability to grow or survive after both short-term (3 hours) and longer-term (24 hours) exposures. However, once the neural stem cells were guided to differentiate into different types of neural cells, prolonged exposure to desflurane caused significant damage. Markers of cell injury, such as lactate dehydrogenase and increased inflammation-related proteins, were elevated after 24 hours. In addition, the number of neurons, as identified by a key marker called PSA-NCAM, was greatly reduced after prolonged exposure to desflurane. These results suggest that while desflurane does not harm neural stem cells directly, it can cause damage to newly formed neurons after long-term exposure. The study highlights PSA-NCAM as an important marker for understanding how anesthetics affect neuron health and provides insight into the mechanisms of anesthesia-induced neurotoxicity. These findings could help improve the safe use of desflurane and other anesthetics in children, reducing potential risks to brain development during early life.

Montelukast (Singulair®) is a G protein-coupled receptor antagonist targeting a specific receptor in the lung. Montelukast is prescribed to millions of patients every year, including very young children, to prevent and treat asthma, allergic rhinitis, and exercise-induced bronchoconstriction. However, montelukast has been associated with reports of serious neuropsychiatric adverse events. FDA NCTR scientists are using a central nervous system-3D organoid NAM to confirm the potential for off-target central nervous system receptor activation and downstream neuromodulation of human central nervous system cells. In 2024 and 2025, extensive secondary pharmacology studies were conducted; these studies identified several off-target candidate receptors expressed in the brain with significant binding and functional activity in non-neuronal cells overexpressing off-target candidate receptors. The human iPSC-derived brain organoids are being used to investigate the neuromodulatory effects of montelukast at endogenous receptors in the human brain and toxicological mechanisms of action. This study represents application of a human-based NAM to address a complex neuropsychiatric context of use for drug risk assessment. Planned future activities include further investigation of the neuromodulatory effects of montelukast at endogenous receptors in human brain 3D organoids and toxicological mechanisms of action.

Assessment of health risks posed by chemical exposures is a fundamental component of warfighter health protection. Neurotoxicity is of particular concern for the military due to the potential for diminished performance and loss of able forces and an elevated risk of mission failure. USAF PRT is leading a cross-service collaboration with the Air Force Research Laboratory MPS Team, Army CBC, Naval Medical Research Unit Dayton and Sanford Burnham Prebys Medical Discovery Institute to evaluate the predictive capacity and reproducibility of several in vitro (2D, 3D, brain organoid, organ-on-a-chip) and ex vivo (rodent-derived brain slice) neurotoxicity models. These are being applied to qualitative (hazard identification) and quantitative (derivation of a POD) risk assessments of chlorpyrifos, chlorpyrifos-oxon, diisopropyl fluorophosphate, and carbaryl. In vitro data will be compared to concurrent in vivo rodent study data. Dose ranges for testing were initially estimated using in silico toxicokinetic models and published literature. Data collection efforts are nearly complete and will be used to estimate PODs using benchmark dose modeling. PODs from in vitro and ex vivo models will then be translated to in vivo doses using IVIVE modeling, and derived external dose PODs will be compared to in vivo PODs. The end product will be a recommended nonanimal test strategy for acute neurotoxicity via acetylcholinesterase inhibition that can also serve as a foundation for other neurotoxicity pathways. Data and strategies developed in this project will improve our ability to protect the warfighter from neurotoxic injury due to operational exposures.

Animal studies have long been used to assess toxicity in drug development and environmental safety, particularly for central nervous system effects such as seizures. However, these animal studies often fail to accurately predict seizure risks in humans, leading to unexpected adverse events during clinical trials or leaving certain environmental compounds, like pesticides, untested due to limited resources.

To improve seizure risk assessment, NICEATM scientists and collaborators explored a novel approach integrating mechanism-based insights (Behl et al. 2025). The investigators identified 27 biological target families associated with seizure risk, combining drug discovery data with key events derived from AOPs. Over 100 in vitro assay endpoints were annotated to 26 of these target families, which included vital components such as neurotransmitter receptors, transporters, and calcium channels. Researchers reviewed a dataset of seizure-inducing and non-seizure-inducing chemicals, with 80% of these compounds tested in the in vitro assays. While the approach showed promise, certain target families lacked sufficient data, highlighting areas for further research.

The findings demonstrate that seizure risk has potential to be assessed using an AOP framework combined with in vitro testing. This approach improves early identification of seizure risk during drug development and environmental chemical evaluations. Expanding the assay panels to include additional targets will further enhance the predictivity of this methodology. By adopting this mechanistic and human-relevant strategy, researchers can better prioritize compounds for study, reduce reliance on less-predictive animal models, and optimize resource use. Ultimately, this approach strengthens central nervous system safety evaluations, supports public health protection from neurotoxicants, and bridges critical gaps in scientific knowledge. This work was highlighted together with other NAMs in a Drug Discovery Today review of this topic (Roberts et al. 2026).

Alzheimer’s disease is characterized by excess free and aggregated amyloid-beta peptides and by intracellular neurofibrillary tangles that contain hyperphosphorylated and misfolded tau protein. Moreover, research has reported neurovascular dysfunction in patients with mild cognitive impairment and Alzheimer’s disease. These data suggest early neurovascular impairment may contribute to or exacerbate the underlying pathophysiology of Alzheimer’s disease.

Researchers at FDA NCTR and collaborators at Emulate developed a brain-chip model of the neurovascular unit using patient-derived human iPSCs (Shen et al. 2026). Cells from an Alzheimer’s disease patient and a cognitively normal individual were used to differentiate cell types that constitute the neurovascular unit, including brain microvascular endothelial cells, neurons, astrocytes, microglia, and pericytes. The model allowed researchers to provide proof-of-concept evidence that a brain-chip model was sensitive to neurovascular unit endpoints that are affected in Alzheimer’s disease. The study found that Alzheimer’s disease brain-chips showed increased paracellular permeability, lower tight-junction protein expression, and reduced activity of P-glycoprotein, a key blood–brain barrier efflux transporter, compared to cognitively normal brain-chips. Additionally, tau protein buildup was observed in the vascular channel, along with increased markers of inflammation. Notably, Alzheimer’s disease brain-chips had lower levels of amyloid-beta 42 compared to the cognitively normal brain-chips, a finding that is not consistent with Alzheimer’s disease pathology and supports the need to include multiple patient-derived human iPSC cell lines when studying heterogenous brain diseases like Alzheimer’s disease. This brain-chip model provides a basis for researchers to investigate blood–brain barrier and neurovascular unit changes using NAMs. Importantly, the study emphasizes the use of a fully isogenic model and neurovascular unit-related endpoints, such as paracellular permeability and P-glycoprotein functioning. While promising, the findings from the study highlight the need to include multiple human iPSC donor lines, as significant genetic variability is part of the model’s limitation. Complete characterization and validation of these models will be necessary before they can be used with confidence in initial evaluations of safety and efficacy of novel therapies being developed for Alzheimer’s disease.

Many cancer drugs, such as doxorubicin, can cause cardiotoxicity in some patients, reducing drug effectiveness as lifesaving treatments. To better understand differences in responses, FDA NCTR researchers used human iPSC-cardiomyocytes, human heart cells created from donor-specific stem cells, to study variability in how individual patients respond to doxorubicin-induced cardiotoxicity (Pang et al. 2024). The study found that iPSC-cardiomyocytes from different donors had varying responses to doxorubicin, resembling patterns seen in real-world clinical practice. By categorizing iPSC-cardiomyocytes into “resistant” and “sensitive” groups based on their reactions to doxorubicin doses, FDA researchers demonstrated that laboratory results could predict how likely a patient is to experience heart damage from treatment. Importantly, they identified a key gene, DND1, which was expressed differently in doxorubicin-resistant versus doxorubicin-sensitive cells, potentially playing a role in cardiotoxicity.

This research highlights the usefulness of donor-specific iPSC-cardiomyocytes for studying differences in drug-induced heart risk among individuals. The project will be continuing into 2026 and beyond to evaluate larger panels of iPSC-cardiomyocytes from diverse donors, aiming to identify new genetic and molecular markers for predicting cardiotoxicity from doxorubicin and other cancer drugs. By providing a more personalized method to understand drug risks, this approach could help reduce heart-related side effects, improve precision medicine in cancer care, and ensure safer treatment strategies for vulnerable patients.

Advances in technology and a push to reduce reliance on animal models are driving the adoption of human-relevant models in drug development. At a May 2024 workshop held by the Health and Environmental Sciences Institute, FDA and other experts discussed how NAMs can improve the safety evaluation of drugs, especially in detecting potential heart-related risks (Pierson et al. 2025).

The workshop focused on innovative tools including human iPSC-cardiomyocytes, microfluidic systems such as BioFlux™, and engineered heart tissues. These systems recreate human heart functions more accurately than traditional animal models, allowing researchers to better predict how drugs might cause cardiac problems. Regulatory representatives emphasized the importance of creating uniform standards and validation protocols for integrating these methods into drug approval processes.

Case studies presented during the workshop highlighted how NAMs often outperform traditional methods in predicting drug safety and identifying issues early in development. These successes show the potential for NAMs to refine cardiovascular drug research, reduce failures during later clinical trials, and support the adoption of more ethical preclinical testing methods. However, the transition to the widespread use of NAMs will require collaboration between academic researchers, industry experts, and regulatory agencies. Together, these groups aim to validate the methods, improve drug testing efficiency, and help create a new era of drug research that is both scientifically advanced and less reliant on animal models.

Cardiovascular toxicity from chemotherapy drugs affects men and women differently, but the underlying mechanisms remain poorly understood. Traditional preclinical cardiac safety studies rely heavily on animal models, often including only male participants, which fails to capture sex-specific differences observed in human patients. An FDA NCTR project addressed this critical gap by developing a human-based laboratory model to study sex-specific cardiac responses to cancer drugs. This model utilizes engineered 3D cardiac tissue constructs created from human iPSCs that can contract and function similarly to actual heart muscle. The engineered heart tissues offer a human-relevant alternative that can maintain stable contractile function for over 100 days, enabling long-term toxicity assessment without animal use. This approach is anticipated to be faster and more cost-effective and provide human-specific data that may predict clinical outcomes better than animal models.

The ongoing project has three main objectives: (1) characterize baseline differences between engineered male and female heart tissues, (2) evaluate the role of sex hormones in cardiac function, and (3) test cardiotoxic chemotherapy drugs such as doxorubicin and paclitaxel. In 2024 and 2025, the research team successfully established protocols for generating stable engineered heart tissues in-house from human iPSCs and demonstrated their capability to detect drug-induced contractile dysfunction. In 2026 and beyond, the team will evaluate the effects of sex hormones and chemotherapy drugs on cardiac contractility using this platform. This innovative platform has significant potential for regulatory applications, enabling sex-specific risk assessment of pharmaceuticals while advancing the FDA's mission to expand nonanimal testing methodologies.

Human iPSC-cardiomyocytes are increasingly being used in advanced NAMs to assess heart-related drug safety in regulatory toxicology. These stem cell-derived heart cells provide a promising alternative to traditional animal-based methods, offering human-specific insights in nonclinical drug evaluations. Scientists in the FDA Center for Drug Evaluation and Research reviewed the landscape of human iPSC-cardiomyocyte submissions to the FDA, highlighting trends and areas for improvement in their regulatory use (Simpson et al. 2025). The search method used was more comprehensive than those used in previous studies and emphasized the importance of standardized keywords in study titles for easier identification of NAMs submitted to the FDA. The study identified an increasing number of human iPSC-cardiomyocyte studies being submitted to the FDA. Most studies relied on the multielectrode array platform, which measures electrical activity in cardiomyocytes. While the review confirmed increasing adoption of human iPSC-cardiomyocyte methodologies, the researchers, who are also involved in nonclinical review of drug applications, noted several challenges. Specifically, study methodologies, context of use, and results of archived submissions were often unclear, limiting their impact on FDA regulatory decisions. More detailed reporting of clinical relevance of human iPSC-cardiomyocyte findings in submitted studies and a better understanding of their predictive accuracy compared to traditional nonclinical methods could broaden the application of human iPSC-cardiomyocytes in regulatory contexts.

This review emphasizes the need for standardized reporting to build confidence in the reliability and relevance of human iPSC-cardiomyocyte data. Such improvements would help the FDA better integrate these innovative methods into decision-making processes. By systematically submitting robust and detailed human iPSC-cardiomyocyte data to the FDA and further demonstrating their predictivity, researchers and applicants can advance the regulatory acceptance of these human-based methodologies, paving the way for safer and more efficient drug evaluations.

Breast cancer is the most common cancer in women and second leading cause of cancer-related death. Breast cancer-specific mortality results from metastasis to other organs, which often develops many years after successful surgery and recovery from the initial primary diagnosis and treatment. NCI is investigating the molecular mechanisms that help cancer cells establish metastases, which are often unique when compared to the primary tumors. Through extensive multipronged characterizations, researchers have demonstrated that a specialized yet simple 3D liquid culture model more accurately replicates how tumor cells behave in the body than traditional 2D systems (Balamurugan et al. 2026; Sterneck et al. 2025). This model helped reveal how inflammatory signals from molecules, such as prostaglandins, stabilize adhesion proteins such as E-cadherin, enhancing cancer cell survival and metastatic potential (Balamurugan et al. 2023). Additionally, the 3D culture model allowed the researchers to discover a targetable molecular link between the E-cadherin cell–cell adhesion molecule and mitochondrial metabolism, which sustains tumor growth (Balamurugan et al. 2026). This innovative model is also being applied to better understand the biology of rare circulating tumor cell clusters, which are most efficient in seeding metastases, and the potential for liquid biopsy applications. This research provides new insights into metastasis, creating opportunities to improve cancer diagnosis, personalized treatments, and patient outcomes.

Common epithelial cancers, including colon, breast, pancreatic, and biliary cancers, have patient-specific mutations that lead to the development of antigens that drive the T cell response to these cancers, called neoantigens. Therapies using T cells often succeed in treating these cancers, but when a patient does not respond to this approach, the treatment failure may be due to reasons related to the patient’s specific tumor. NCI researchers are developing patient-derived tumor organoids to investigate the mechanisms of treatment failure and optimize subsequent treatment options (Gustafson et al. 2026). Because common epithelial cancers rarely grow in conventional 2D culture, 3D matrix-based organoids are generated from each treated patient to serve as autologous tumor surrogates. Studies using these models showed that organoid recognition by tumor-infiltrating lymphocytes correlates with higher clinical response rates. Ongoing work involves investigating discrepancies between organoid recognition and neoantigen presentation by normal antigen-presenting cells. In addition, whole-exome sequencing is being used to identify mutations, and the neoantigens related to these mutations are being cloned for further study.

With thousands of chemicals present in the environment, identifying which ones may pose risks to human health is a critical task. Scientists are using advanced laboratory methods like high-throughput transcriptomics to quickly determine how chemicals might harm biological systems. High-throughput transcriptomics looks at changes in gene activity (transcriptomics) when cells are exposed to chemicals, helping pinpoint which molecular targets or pathways might be affected.

EPA scientists developed an approach to link transcriptomic results to specific molecular targets, such as nuclear receptors, and combine those findings with more focused receptor-level tests to confirm their accuracy (Rogers et al. 2025). Using special gene-based “signatures” developed from known reference chemicals, researchers screened 1,126 chemicals, identifying 201 that showed specific bioactivity linked to those targets. Of these, 57 chemicals were confirmed as selective nuclear receptor activators. When the researchers lacked data to confirm predictions, they tested 37 chemicals using five receptor-level assays, ultimately confirming that 18 (49%) of these had bioactivity as predicted. For prioritized chemicals, transcriptomics provided better or equally sensitive thresholds for potential effects compared to traditional pathway-based measures in 95.4% of cases. This innovative approach demonstrates how integrating high-throughput transcriptomics with targeted testing is highly effective at identifying chemicals with specific biological activity. The approach can help prioritize chemicals for further study, supporting risk assessment strategies that are faster, more efficient, and less reliant on traditional animal models. It offers a promising step forward in identifying potential chemical hazards more effectively.

NAMs can be used to rapidly evaluate chemical safety without relying on traditional animal models. One key tool is the combination of in vitro high-throughput laboratory tests and in silico computer models to predict how chemicals might affect human health. These methods can help set a POD, or the level of chemical exposure that starts to cause harmful effects, and compare it to estimated exposure levels, resulting in a bioactivity-to-exposure ratio. This ratio helps prioritize chemicals for further testing based on potential risks. Additionally, these tools flag potential hazards, such as effects on hormones, brain development, or the immune system, for further investigation.

In a case study conducted by an international group of researchers, including researchers from the EPA, 200 chemicals with little traditional toxicity data were tested using NAMs (Paul Friedman et al. 2025). Scientists combined several advanced methods, including transcriptomics (the study of gene activity), phenotypic profiling (observing cellular traits), biochemical tests, and high-throughput toxicokinetic models, to estimate safe exposure levels for these chemicals. The results provided bioactivity-based PODs and helped identify potential hazards for further study.

This case study shows how NAMs can efficiently address gaps in chemical safety data, ensuring human health is protected while reducing the need for animal testing. By offering a reusable and flexible approach, this effort supports the use of cutting-edge science in managing chemical risks around the world. It demonstrates that combining NAM tools can provide reliable data for screening chemicals, improving our ability to assess risks quickly and ethically.

Cleanup of industrial sites contaminated with hazardous waste is administered through the EPA Superfund program. At these sites, soil, air, and water may contain hazardous materials, including metals. Individuals residing and working in these areas can be exposed to mixtures of these substances, and such exposures can lead to many adverse health outcomes, including cancer. However, the risk presented by mixtures of hazardous materials compared to that of individual components has not been studied thoroughly. NCATS and NIEHS researchers investigated the toxicity of metal mixtures in five categories of in vitro assays (Ooka et al. 2025) measuring cytotoxicity, oxidative stress, genotoxicity, cytokine release, and blood vessel development (angiogenesis). Testing of these mixtures in primary cells and cell lines demonstrated that nickel/arsenic/cadmium and beryllium/arsenic/cadmium combinations exhibited higher cytotoxicity than their individual components, suggesting that the mixtures amplified the cytotoxic effect. Because oxidative stress is a common factor in most metal-related toxicities, the mixtures were further tested in cell-based assays to measure oxidative stress. Results showed that cadmium-induced oxidative stress was increased in some mixtures. The mixtures that induced oxidative stress also exhibited increased DNA damage, inhibited DNA synthesis, and activated the p53 transcription factor pathway. In addition, compared to their component compounds, the mixtures significantly increased interleukin-8 secretion, a marker of inflammatory response, and angiogenesis. These results demonstrate the utility of in vitro NAMs to explore mechanisms of mixture-induced toxicity.

N-nitrosamines are chemical impurities increasingly detected in human drugs, and there is concern about their potential to cause mutations and cancer. To better understand their effects, FDA NCTR researchers (Seo et al. 2025) conducted a study comparing the DNA damage caused by eight nitrosamines in primary human hepatocytes and primary rhesus macaque (Macaca mulatta) hepatocytes. They measured DNA damage and analyzed cellular responses to validate potential genotoxic risks using tests including the comet assay. This is a sensitive technique used to detect DNA damage at the level of individual cells. It involves the electrophoresis of lysed cells in agarose gel, allowing damaged DNA to migrate and form a "comet" shape under fluorescence analysis.

The results revealed that all eight nitrosamines caused significant DNA damage in hepatocytes from both species. Interestingly, most nitrosamines resulted in higher DNA damage in macaque hepatocytes than in human hepatocytes, with variations linked to enzyme activity differences. Human hepatocytes showed greater variability in responses between donors, particularly in CYP enzymes, which influence how toxic chemicals are broken down in the body. Further tests, including western blot analysis, suggested that the presence of specific chemical groups, such as ethyl groups, may influence how nitrosamines promote DNA damage and cell death. Modeling results revealed more conservative threshold values (called benchmark concentrations) for DNA damage in human hepatocytes than in macaque hepatocytes, suggesting that human hepatocytes may have higher sensitivity to nitrosamines. The results align with findings from earlier studies using 3D HepaRG spheroid models, supporting the relevance of these results in human health risk assessments. Overall, this research highlights critical human-relevant findings of the genotoxic potential of nitrosamines and underscores the importance of investigating species differences to improve drug safety standards and toxicological evaluations.

Silver nanoparticles (AgNPs), which are widely used for their antimicrobial properties in food packaging and dietary supplements, may pose risks to gut health. Previous studies showed that AgNPs affect gut microbial composition and epithelial function, and research by FDA NCTR explored how AgNPs impact host-pathogen interactions at the intestinal barrier, particularly in the context of Salmonella serovar Heidelberg, a common cause of foodborne illness (Hanafy et al. 2026).

In laboratory tests, human intestinal epithelial cells were exposed to small (10 nm) AgNPs at safe, nontoxic levels. While AgNP exposure did not affect the ability of the bacteria to adhere to the cells initially, it significantly increased bacterial invasion and survival inside cells, particularly at the 10 µg/mL concentration. This indicates a weakened epithelial barrier, suggesting AgNP exposure can make cells more vulnerable to infection. FDA researchers observed changes in gene expression related to intestinal permeability, with affected genes controlling tight junctions, cell connections, and structural remodeling. Even though the cells activated repair mechanisms, these efforts were not strong enough to prevent Salmonella invasion and persistence. AgNP exposure also increased proinflammatory cytokines, worsening inflammation, while reducing key anti-inflammatory and repair factors.

These findings reveal that even low doses of AgNPs can disrupt the gut barrier, making it easier for pathogens including Salmonella to invade and persist. Given their increasing presence in consumable products, this study emphasizes the need to reassess the safety of AgNPs, especially regarding gastrointestinal health and infection risk. These findings also reveal that in vitro methods can successfully characterize toxicity of a nanomaterial. 

The U.S. Army CBC recently completed a project aiming to determine if multiorgan chips can be effective models for the establishment of bacterial infection and for characterization of antibiotic efficacy and safety. This effort enabled CBC to be the first end-user of the Chip4 platform, a multiorgan chip that encompasses four different organ models containing two independent media circuits, representing blood and excretion. To evaluate this application, CBC used methicillin-resistant Staphylococcus aureus (MRSA) and its primary antibiotic, vancomycin. MRSA was selected as the target bacteria because vancomycin has been characterized in clinical data. The efficacy of vancomycin within the multiorgan chip platform after MRSA exposure was analyzed while the platform and potential compatible downstream analysis techniques were established. CBC scientists were able to establish MRSA infections in the multiorgan chips and observe the efficacy of vancomycin. They were also able to characterize the safety of the antibiotic, specifically pertaining to cytotoxicity. The data generated from this project are included in the CBC technical report DEVCOM-CBC-TR-1996 “Incorporation of a Multi-Organ-on-a-Chip Platform to Assess Drug Candidate Characterization for Methicillin-Resistant Staphylococcus aureus (MRSA)”. Follow-up work for this study is ongoing. The current goal is to establish a multiorgan chip viral infection model capable of characterizing biosafety level 3 and 4 viral pathogens of concern for medical countermeasure development and testing.

The CBC has also been working toward creating a more physiologically relevant organ-chip infection model through the incorporation of commensal microbial communities to human tissues. Skin and buccal chip models will be cocultured with microbial species representative of their specific microbiomes, and the chips will be subsequently exposed to relevant bacterial pathogens. By combining cutting-edge organ-chip technology with microbiome integration and sophisticated downstream analysis, this program seeks to advance the understanding of biological agent pathogen toxicity and improve predictive models for diagnostics development.

While MPS technologies are gaining traction as a viable alternative model for toxicity studies, further characterization is necessary to explore their full translational potential as well as the utility of these platforms to serve as a diagnostic tool. For human-relevant omics data to be truly reliable, samples must be collected and analyzed from patients or individuals infected by or exposed to the drugs or agents in question. U.S. Army CBC scientists investigated how well bioinformatic data collected from organ chips match bioinformatic data collected from human samples, with the aim of using organ chips as an alternative approach for biomarker discovery in toxicity assessment and exposure identification. To do this, TissUse Chip3 multiorgan chips seeded with kidney organoids, liver organoids, and full-thickness lung tissue were exposed to acetaminophen and analyzed to characterize effects on protein content and metabolism. Data from the organ chips largely matched the data in the published literature, including the identification of several known acetaminophen metabolites and biotransformation products. These data suggest that organ chips may be a suitable surrogate for human biomarker identification and drug exposure diagnosis. A paper describing this work is currently under review with the Journal of Proteome Research. Follow-up work that aims to establish animal organ chips and examine their translational potential was recently funded. Currently, CBC is developing rat and nonhuman primate multiorgan chip systems and will compare the effects of acetaminophen on the incorporated tissues to results reported in historic and published literature.

Per- and polyfluoroalkyl substances (PFAS) are a large group of human-made chemicals widely used in industry and consumer products. They are highly persistent in the environment, can bioaccumulate, and pose potential toxicity concerns. While a handful of PFAS have been studied for their hazard profiles, most remain poorly understood due to the number of chemicals involved. Evaluating thousands of PFAS using traditional methods would require substantial time, cost, and animal resources, prompting researchers to seek more efficient alternatives.

EPA and the NIEHS Division of Translational Toxicology launched a research project to address this challenge (Patlewicz et al. 2026). The study screened approximately 150 PFAS using a range of NAMs, including small model organisms such as zebrafish (Danio rerio); in vitro assays that examine cellular and biochemical activities; and toxicokinetic assays that study absorption, distribution, metabolism, and excretion. The data were analyzed to identify chemical categories, prioritize PFAS for further study, and apply read-across approaches to predict toxicity for untested chemicals based on similarities within categories.

The findings from this effort helped categorize PFAS based on their structural and biological properties, facilitating streamlined prioritization for future research and regulatory actions. By leveraging alternative testing methods and computational tools, this work represents an efficient and ethical approach to tackling the extensive hazard evaluation needed for PFAS. It also provides a foundation for tiered-testing strategies, reducing reliance on traditional animal models while advancing understanding of PFAS toxicity and environmental safety.

PFAS are synthetic chemicals widely used in commercial products. These chemicals are a global concern due to their persistence in the environment and extensive associations with adverse health outcomes. While legacy PFAS have been extensively studied, many nonlegacy PFAS lack sufficient toxicity information. NCATS and NIEHS researchers analyzed the bioactivity of PFAS using Tox21 screening data (Ooka et al. 2024). Over 75 assay endpoints including nuclear receptors, stress response, and metabolism were examined to understand the toxicity of nonlegacy PFAS and investigate potential new targets of PFAS. This analysis confirmed several known PFAS targets and pathways and identified several potential novel targets and pathways. The novel PFAS-affected targets included the metabolic enzyme CYP2C9, as well as ROR and sonic hedgehog, targets both related to tissue development. To confirm the effect of PFAS on these novel targets and pathways, follow-up studies were conducted using several cell- and enzyme-based assays. PFAS were found to inhibit cytochrome P450 metabolism enzymes, especially CYP2C9, which is important for metabolism of both drugs and endogenous compounds, such as fatty acids. This PFAS–CYP2C9 interaction was further investigated using molecular docking analysis, which suggested that PFAS directly bind to the active sites of CYP2C9. These findings are important contributions to understanding the mechanism of PFAS action and toxicity.

Results from in vitro assays used in the Tox21 and EPA ToxCast HTS programs may be confounded by overt cell stress and cytotoxicity, such that a decrease in viable cells could erroneously be attributed to a chemical’s mechanistic effects. Integration of cytotoxicity assessment with assay endpoints can bolster confidence in the interpretation of assay outcomes. To address this, concurrent cytotoxicity readouts from Tox21/ToxCast assays were mapped for hundreds of assay endpoints, allowing direct comparison of bioactivity potency against cytotoxicity. For chemicals tested in HTS assays that lack directly relevant cytotoxicity data, NICEATM scientists are developing machine learning algorithms to predict chemical- and cell type-specific cytotoxicity concentrations across time trajectories. The models’ predictions will be integrated with bioactivity data to strengthen confidence in assay outcome interpretation by identifying mechanistic vs. nonspecific/cytotoxicity-confounded bioactivities. The comparisons and predictions will be integrated into future versions of concentration–response visualizations for the curated HTS data in the NICE Curve Surfer tool. The direct comparisons will also be incorporated into the NICEATM project to harmonize curated HTS data to the OECD Harmonised Template 201, providing additional context for bioactivity calls to help facilitate the international adoption of standardized NAMs.

Biomarkers are measurable biological indicators, such as specific proteins or genes, that can provide insights into how the body reacts to chemical or drug exposure. Concepts from translational medicine, particularly the use of biomarkers, can help advance alternative testing methods to replace animal studies in preclinical safety assessments. By studying these indicators, researchers can improve the accuracy and human relevance of NAMs used in safety testing.

In a review article (Hartung et al. 2024), NIEHS and NCATS researchers outline how principles from translational medicine, such as “fit-for-purpose” validation (matching tools to specific tasks), evidence-based methods, and integrated testing strategies, are already being used to refine and validate NAMs. Biomarkers can help researchers better understand the mechanisms behind toxicity and relate observations in alternative models to outcomes from animal studies or clinical trials. This improves confidence in NAMs as replacements for animal testing.

Challenges, such as standardizing methods, demonstrating relevance, and achieving regulatory acceptance, still exist, but the field is advancing thanks to collaboration, technological innovations, and evolving regulations. The review also considers future trends, including combining data from multiple biological systems (multiomics), using organ-like MPS, and applying super intelligence to analyze data. Successful case studies show how biomarkers are already being applied in preclinical safety testing. By integrating biomarkers into NAMs, researchers can develop more predictive, ethical, and human-relevant safety assessment tools, ultimately reducing reliance on animal models and improving public health outcomes.

Toxicological testing methods are used to assess the safety of chemicals, but the final results depend heavily on how the raw data is processed. This study focused on the importance of understanding and managing the steps involved in processing and interpreting data generated by NAMs. When toxicological tests are conducted, multiple layers of data are analyzed in sequence, starting with raw outputs from laboratory equipment and ending with a final classification of the test chemicals (e.g., whether they are toxic or not). Each step in the process involves specific data handling methods, and even small changes in these steps can alter the final outcome. For instance, the same raw data, when processed differently, could lead to different conclusions about a chemical’s safety. This inconsistency can affect how reliable, robust, and acceptable NAMs are for regulatory use.

To address this, a review by FDA and NIEHS researchers emphasizes that detailed descriptions of data processing steps should be included in NAM reporting templates such as ToxTemp (Blum et al. 2025). This transparency helps ensure that results are reproducible, building trust in the method and supporting its acceptance by regulatory agencies. The review provides an overview of the key data levels involved in NAMs, explains how data is processed between these levels, and highlights the impact of these steps on the final results. By raising awareness of the importance of careful data management, this work aims to make NAMs more reliable, trustworthy, and widely accepted in toxicology, ultimately advancing the use of modern, animal-free testing methods.

Scientific communities and industry have long sought alternatives to traditional cell cultures and animal studies to improve human-relevant models for applications such as drug development, toxicology assessments, and preclinical evaluations. One promising innovation is microfluidic-based organ-on-a-chip technology, which mimics the structure and function of organs and organ systems more accurately than conventional methods. Despite increasing interest from regulatory bodies and industry, implementation of this technology has been slowed by the absence of standardized guidelines.

To address this gap, a multidisciplinary team of stakeholders was formed, including representatives from FDA, NIST, the European Union, academia, and industry (Reyes et al. 2024). This collaboration aimed to establish a framework for developing guidelines and standards for the engineering and application of organ-on-a-chip models. By promoting consistency and reliability, these guidelines will help overcome technical and regulatory challenges, advancing the adoption of these systems.

The resulting framework encourages further conversations among stakeholders, including standards development organizations, to refine and finalize best practices for organ-on-a-chip use. This effort represents a critical step toward optimizing the potential of this innovative technology, ensuring it meets regulatory and industry needs while enhancing human-relevant testing.

Ultimately, this work lays the foundation for widely accepted standards that will improve the reliability, scalability, and impact of organ-on-a-chip models, supporting advancements in drug development, toxicology, and personalized medicine.