Since 2023 the European Commission has been working on a roadmap for phasing out animal testing in chemical safety assessment, and in 2025 the US FDA published a roadmap to phase out animal testing for the pharmaceutical industry. We describe our Merck KGaA strategy across Life Science, Healthcare, and Electronics, introduced in 2021, to reduce animal testing by 50% by 2032 and by 75% by 2040 through our 4R program (Replace, Reduce, Refine, Responsibility). The approach focuses on what we can achieve today and avoids obstructive discussions about unresolved issues. We have categorized all animal tests currently required for our products into three "baskets" (3B). Basket 1 (Adoption) includes animal tests for which alternatives are available and accepted, including those still required in certain regions. Basket 2 (Adaptation) contains tests for which alternatives are proposed or being developed but that cannot yet be replaced. Basket 3 (Assessment) contains tests for which innovative replacement strategies still need to be developed. This strategy guides effective short-term replacement and directs investments into areas where innovation can replace animal testing in the future. It coordinates the way forward for all stakeholders and creates actionable milestones toward a genuine replacement of animal testing. The collaborative agreement in 2024 among the members of the European Federation of Pharmaceutical Industries and Associations to adopt the 3B approach, recognized by the European Commission and the European Medicines Agency, highlights its significance as a valuable tool for fostering a more ethical and sustainable science environment across Europe.
The whole blood pyrogen test was first described in this journal exactly thirty years ago. Its variant based on cryopreserved blood followed one year later. Together with other monocyte activation tests (MATs), it has fundamentally changed the landscape of pyrogen testing. In the five years since the 25th anniversary article in this series, progress has been remarkable: The European Pharmaco-poeia deleted the rabbit pyrogen test (RPT) effective January 2026, ending a 55-year era; ISO 10993-1:2025 removed material-mediated pyrogenicity from mandatory evaluation endpoints for medical devices; the U.S. FDA updated its guidance on pyrogen testing; and the MAT market grew to over $600 million. New validation studies have demonstrated MAT equivalence to the RPT for both endotoxin and non-endotoxin pyrogens, and the first product-specific MAT validations have been accepted in regulatory filings in Europe and the United States. Reporter cell lines and tran-scriptomic approaches are opening next-generation detection capabilities. Yet implementation gaps persist: The MAT is still underutilized for blood transfusions, cell therapies, and airborne pyrogens. Recombinant alternatives to the Limulus amebocyte lysate assay (LAL) have finally achieved phar-macopeial recognition, addressing horseshoe crab conservation concerns. This article reviews the developments of the last five years, updates the lessons learned, and reflects on three decades of bringing a human cell-based test from the laboratory bench to global regulatory acceptance.
OECD Guideline 497 includes defined approaches (DAs) that combine new approach methodologies (NAMs) to predict skin sensitization hazard and categorize potency according to the United Nations Globally Harmonized System of Classification and Labelling of Chemicals (UN GHS). To increase flexibility, the OECD allows the substitution in the DAs of NAMs addressing the same key events (KEs) in the adverse outcome pathway (AOP). This study evaluated the implementation of the U-SENS™ assay for dendritic cell activation (KE3) within two integrated testing strategy (ITS) DAs. First, EC150 threshold values were determined via computational modeling to convert CD86 stimulation index data into ITS DA scores. The performance of the modified ITS DAs was then evaluated against available Local Lymph Node Assay (LLNA) and human reference data. For hazard identification, the ITS DAs achieved 79% balanced accuracy compared with LLNA data, and 73-74% compared with human data. For potency categorization, correct classification rates were 67-68% against LLNA and 65-67% against human data. In comparison, the LLNA showed lower performance, with 58% balanced accuracy for hazard identification and 60% for potency categorization against human data. Selected case studies illustrate the practical application of these DAs following the Guideline 497 decision flowchart. Overall, findings underscore that incorporating U-SENS™ into ITS DAs enhances flexibility without compromising predictive capacity and outperforms traditional animal testing in providing reliable skin sensitization classification outcomes.
The use of terminology related to animal-free science has grown rapidly over the past two decades; however, definitions and interpretations of key terms remain inconsistent across global regulatory, scientific, and policy contexts. The term “new approach methodology(ies)” (NAM(s)) exemplifies this issue: though rarely used prior to 2019, its uptake in scientific literature has increased substantially in recent years. Despite this growth, the term, and its acronym, “NAM”, is used with varied meanings, leading to misunderstanding regarding the types of methods described. Similarly, the term “animal” is defined differently across common language, scientific discourse, and legal frameworks, resulting in further ambiguity in what constitutes “animal-free science”. These variations can hinder productive dialogue and collaboration, particularly in international settings. This manuscript maps the current landscape of definitions of key terms such as “NAM”, “animal”, and “animal-free”, drawing on regulatory, academic, and institutional sources to enhance understanding of the range of existing interpretations. By improving transparency and clarity in terminology, this effort seeks to support more coherent and effective global communication in the field of animal-free science. In addition, the definition of NAM, as agreed by the members of the International Collaboration on Cosmetics Safety, is described. Tabulated definitions and sources are provided as a reference tool.
Reporting standards have proliferated across biomedicine, yet incomplete methods reporting remains routine - less because the community doubts the value of transparency, but rather because compliance checking is tedious, inconsistently enforced, and poorly integrated into everyday writing and review. As a sequel to the Good In Vitro Reporting Standards (GIVReSt) argument that better reporting is essential infrastructure, this article explores a pragmatic next step: translating standards from static checklists into interactive, always-on guidance. We describe the development of three specialized "compliance copilots" built as custom GPT-based assistants - one aligned with the emerging GIVReSt, one reflecting the established ToxRTool reliability framework, and one mapped to ARRIVE for animal studies. The tools are designed to point to specific text evidence, flag missing essential information, and provide actionable suggestions while the manuscript is being written. Early benchmarking against expert assessments suggests that this approach can approx-imate human judgments for many checklist items in a fraction of the time and with high consistency. We also highlight why "strict" versus "lenient" interpretations matter, and why these systems should be framed as decision-support, not decision-makers. The central claim is cultural, not technical: arti-ficial intelligence (AI) will matter most when it makes rigorous reporting the path of least resistance, turning standards into routine practice rather than aspirational add-ons.
In vitro toxicology assays are widely used in the context of toxicology to study in vivo outcomes and to improve mechanistic understanding of toxic responses. For this purpose, a better characterization of actual exposure and the fate of chemicals within the cellular test system is required. The present study aimed to develop a novel model, INSIGHT (In Silico Guide for Harmonized in vitro Testing), that integrates physiological and physicochemical parameters to better describe chemical fate in vitro and to guide assay design. The newly developed model, integrating the dynamic features of the Virtual Cell Based Assay model with the partitioning framework of the Virtual In Vitro Distribution model, was calibrated using both a large literature dataset and original experimental data, comprising a total of 42 chemicals and 7 commonly used cell lines: HepaRG, HepG2, 3T3 Balb/c, PC12, MCF-7, RTgill-W1, and HEK293. These cell lines were selected for their diverse tissue and species origins, metabolic capacities, and potential for functional transport mechanisms, which are known to influence chemical kinetics. The INSIGHT model demonstrated flexibility and robustness across a range of cell lines when the parameters driving their metabolic activity or functional transport were informed. The present study underscores the pivotal function of logPow determination and the necessity for accurate calibration of partition coefficients, permeability, and metabolic processes to account for variability across cell lines and tested chemicals. This approach supports and facilitates enhanced experimental design and advancing quantitative in-vitro-to-in-vivo extrapolation (qIVIVE) in toxicology and strengthens next generation risk assessment (NGRA) workflows.
Read-across is frequently used in chemical risk assessment to predict the toxicological properties of data-poor compounds instead of performing new animal tests. The selection of relevant source compounds (SCs) for read-across to data-poor target compounds (TCs) is one of the main challenges, particularly for endpoints such as chronic and developmental toxicity, where the mechanisms leading to the observed apical toxicological findings are often unknown. In this study, the predictivities of using chemical, biological and/or metabolite similarity to inform read-across strategies were compared. Existing data from two reference compound groups, i.e., the pesticide classes of 27 triazoles and 8 triazinyl-sulfonylureas, were used to assess the performance of the three approaches and their modular combinations to identify relevant SCs from a large database of 468 pesticide active compounds. Metabolite similarity yielded high positive predictive values (PPV) of 85-100% while sensitivity was relatively low (12-65%). Basing metabolite similarity assessment on observed or predicted metabolites yielded comparable results. Chemical and biological similarity alone were less effective in separating SCs from irrelevant compounds. Modular approaches combining, e.g., chemical and metabolic similarity, enhanced SC selection (PPV up to 100%). This research underscores the potential of metabolite data to strengthen read-across justifications, thereby contributing to regulatory compliance and safety evaluations. We propose the integration of metabolite similarity assessment into an existing EU-ToxRisk workflow to enhance the reliability of read-across for chemical risk assessments.
New approach methodologies (NAMs), such as in vitro and in silico methods, provide opportunities to replace, reduce, and refine animal studies for chemical risk assessment, decrease experimental lead time, and increase predictivity for human health. However, starting from or including NAMs introduces uncertainties that need to be addressed and quantified to work towards regulatory implementation of NAMs within risk assessment. A theoretical risk assessment based on the developmental neurotoxicity (DNT) of chlorpyrifos and its active metabolite chlorpyrifos-oxon was performed to highlight important extrapolations and identify and quantify uncertainties when extrapolating NAM-based data to estimate human health risk. Cell viability of differentiating human neuronal progenitor cells was used as a proxy for DNT during early pregnancy. A physiologically based kinetic (PBK) model describing early pregnancy was developed to facilitate quantitative in-vitro-to-in-vivo extrapolation (qIVIVE) to estimate the external dose leading to DNT onset. An in vitro-based health-based threshold value (HBTV) was derived for the onset of DNT caused by chlorpyrifos and chlorpyrifos-oxon following chlorpyrifos exposure. This value was higher than the maternal exposure associated with the onset of DNT based on epidemiological data, showing the need for improvements in the NAM-based risk assessment of chemicals suspected of DNT. Improvements relate to the selection and refinement of the in vitro endpoint and assay and defining the quantitative relationship between the selected key event and the adverse outcome and provide a starting point for future research to aid regulatory implementation.
Next generation risk assessment (NGRA) strategies use animal-free new approach methodologies (NAMs) to generate information concerning chemical hazard, toxicokinetics (ADME), and exposure. The information from these major pillars of data gathering is used to inform risk assessment and classification decisions. While the required types of data are widely agreed upon, the processes for data collection, integration and reporting, as well as several decisions on the depth and granularity of required data, are poorly standardized. Here, we present the Alternative Safety Profiling Algorithm (ASPA), a broad-purpose, transparent, and reproducible risk assessment workflow that allows documentation and integration of all types of information required for NGRA. ASPA aims to make safety assessments fully traceable for the recipient (e.g., a regulator), delineating which steps and decisions have led to the final outcome and why certain decisions were made. An overarching objective of ASPA is to ensure that identical data input yields identical outcomes in the hands of independent assessors. Therefore, ASPA is not just a data gathering workflow; it also considers data interdependencies and requires precise justification of intermediate decisions. This includes the monitoring and assessment of uncertainties. To assist users, the ASPA-assist software was developed. It formalizes the reporting process in a reproducible and standardized fashion. By guiding an operator step-by-step through the ASPA workflow, a complete and comprehensive report is assembled, whereby all data, methods, operator activities, and intermediate decisions are recorded. Practical examples illustrating the broader applicability of ASPA across various regulations and problem formulations are provided through case studies.
The European Union’s main chemicals regulation, Registration, Evaluation, Authorization and Restriction of Chemicals (REACH), requires chemicals to be evaluated for health and environmental impacts, with animal tests the basis for many evaluations. Most discussions of REACH animal use focus on mammals, yet fish tests are also a significant component. Here we report the animal count for fish tests completed, ongoing, and pending under REACH, based directly on test reports. The estimated total to date is 382,000 fish used for short-term fish toxicity, long-term fish toxicity, endocrine disruption, and bioaccumulation tests. This count does not include new tests that will result from the 2022 REACH amendment, which extends requirements for long-term fish toxicity tests and removes the most common basis for waivers previously accepted for this test. An estimated 940−1,240 new long-term fish toxicity tests may result from these changes, requiring 520,000−680,000 fish. The count also does not include the potential expansion of endocrine disruption testing in the upcoming REACH revision. New non-animal alternatives to long-term fish toxicity and endocrine disruption tests are needed to reduce these impacts. For other fish tests, recently defined non-animal methods for short-term toxicity and the newly approved Hyalella azteca bioconcentration test for bioaccumulation should be evaluated for inclusion in REACH guidance, both to incentivize their use and to better comply with the REACH mandate to use animal testing only as a last resort.
A broad range of computational models is available for animal-free chemical safety assessment. The models are used to predict a variety of endpoints, including adverse effects or apical endpoints, toxicokinetic properties, and exposure, often from chemical structure or in vitro inputs alone. To support their wider use, such models need to be findable, accessible, interoperable, and reusable (FAIR). This study has reevaluated the existing FAIR principles applied to quantitative structure-activity relationships (QSARs) in order to adapt these principles to a wider range of computational models. Despite the breadth and variety of approaches, many computational models comprise common components including the training series, information about the modelling engine, and the model itself. As a result, a refined set of four FAIR Lite principles is proposed based on the methodological foundations of computational toxicology which are unambiguously understood by practitioners such as developers and end-users. To this end, it is proposed that to comply with the original FAIR principles, a computational toxicology model should be associated with (i) a globally unique identifier for model citation; (ii) the capture and curation of the model; (iii) the metadata for the dependent and independent variables and, where possible, data; and (iv) storage in a searchable and interoperable platform. The FAIR Lite principles are mapped onto the original FAIR principles applied to QSARs, thereby demonstrating that a simpler checklist approach covers all aspects.
Under the European Union’s REACH (Registration, Evaluation, Authorisation and Restriction of Chemicals) regulation, the European Chemicals Agency (ECHA) is required to assess the compliance of safety data submitted by chemical registrants. ECHA must check a proportion of registration dossiers for compliance. From 2010 to 2023, 4,854 compliance checks (CCHs) were conducted. When dossiers lack required studies or use inappropriate adaptations, ECHA issues decisions that are publicly available. As of April 2, 2024, 2,311 such decisions had been published. This study systematically analyzed these published CCH decisions, focusing on ECHA’s findings of non-compliant REACH registrations and the use of adaptations to standard information requirements. We found that over 70% of published CCH decisions included at least one adaptation, with “read-across” being the most common (48%). Among these, 83 documents contained read-across adaptations with justifications that ECHA deemed plausible. To understand what made these read-across hypotheses acceptable, we evaluated them using 17 assessment elements that capture specific arguments registrants proposed to justify the adaptation. Elements of “acceptable” read-across hypotheses included strong evidence of (i) toxicokinetic similarity between the registered substance and its analogues, and (ii) toxicodynamic similarity, supported by bridging studies. Additional support from in vitro studies and QSAR predictions further strengthened the accepted read-across hypotheses. Overall, this analysis provides insights into what constitutes a successful read-across under REACH data requirements. By identifying and evaluating accepted cases or read-across adaptations, we highlight best practices for establishing scientifically robust justifications for chemical similarity that can meet ECHA’s high regulatory standards.
This study aimed to develop a physiologically based kinetic (PBK) model for benzophenone-4 (BP-4) in humans based on in vitro and in silico input data and to achieve scientific confidence in predicted internal exposures of BP-4 in the absence of human kinetic data. The key steps included are: 1) establishing a core PBK model containing minimal required input for dermal absorption, liver metabolism, plasma protein binding, blood:plasma ratio, and tissue:plasma partition coefficients, 2) using chemical-specific characteristics to define additional key kinetic processes, which led to inclusion of transporter kinetics, and 3) conducting sensitivity analyses and assessing population variability. The in vitro kinetic results revealed limited skin penetration of BP-4 (< 0.4%), no metabolic conversion by the liver, and involvement of active transporters, including OAT1, OAT2, OAT3, BCRP, and MRP4. Inclusion of the transporter activity in the PBK model (scaled to kidney and liver) resulted in BP-4 active excretion and lowering of the plasma concentrations from 4 μM to 0.7 μM. Due to faster influx rates, by OAT1, OAT2, and OAT3, compared to efflux rates by BCRP and MRP4, relatively higher organ concentrations were predicted for the liver (0.31 μM) and kidney (0.18 μM) compared with other organs. While the PBK model results could not be evaluated against human data, we could evaluate the evolution of predicted concentrations in the process of developing the model. Increasing the physiological relevance of the model through inclusion of transporters increased confidence in the plausible ranges in plasma and organ concentrations.
N-nitrosamines (NAs) are potentially carcinogenic organic compounds, and nitrosamine drug substance-related impurities (NDSRIs) are currently regulated with class-specific thresholds in the low nanogram range according to the carcinogenic potency categorization approach (CPCA) classification schema. Beyond direct exposure, NDSRIs can form endogenously in the human organism after ingestion of secondary amines. As recently shown, enalapril, propranolol, and fluoxetine form NDSRIs under conditions mimicking the acidic environment in the stomach. The MUTAMIND project investigated whether such endogenously formed NA levels lead to plasma or liver concentrations which align with or exceed the acceptable intake limits based on the current CPCA. A generic physiologically based kinetic (PBK) model was built using compound-specific in vitro ADME parameters such as intestinal permeability and hepatic clearance. The predictions correlated well with measured in vivo ADME data for the data-rich APIs, so the same PBK approach was applied to the corresponding NDSRIs. While the modelling of propranolol was unremarkable, the highest NA conversion rate observed for N-nitrosoenalapril under gastric conditions resulted in plasma and liver levels exceeding those derived from the CPCA threshold by a factor of about 800 and 400, respectively. The long half-life of fluoxetine suggests a risk of bioaccumulation of its nitrosamine with chronic exposure. These findings indicate that PBK modelling could be a valuable tool as part of a weight of evidence approach in contributing to the risk assessment of nitrosamine impurities in pharmaceuticals.
Since its discovery as an innate bacterial immune system, the clustered regularly interspaced short palindromic repeats (CRISPR) associated nuclease 9 (CRISPR-Cas9) system has quickly landed on mammalian genomes to become the first-in-class editing technique. CRISPR-Cas9 offered an invaluable approach to correct pathogenic mutations, thus becoming a promising cure for diseases with highly unmet medical needs. To date, several attempts have been made to understand, categorize and predict the outcome of genetic manipulation with different degrees of success. The lack of an appropriate and translatable model to test CRISPR/Cas9 effects, both wanted and unwanted, has limited its applications to advance gene therapies. Herein we describe the potential of microphysiological systems (MPS) as an alternative to the classical models used in CRISPR safety studies, such as immortalized cell lines or small mammals (e.g., rodents), to facilitate the progress of new CRISPR medicines to the clinics.
While the genome codes for all proteins an organism can express, only certain sets of proteins are expressed in defined cell types. A cell's phenotype is influenced by the life-stage, exposure to signal molecules, as well as the exposome, i.e., external influences including physical stressors or chemicals from food, the environment or microorganisms. The interplay between genetics and these exposures is termed gene x environment interaction (GxE). Epigenetics contributes to GxE by modifying the accessibility of genes and thus their ability to be translated to proteins. Epigenetic mechanisms include DNA methylation, histone modifications, non-coding RNAs, and changes in local DNA packaging. As genetics and the exposome often jointly contribute to disease, understanding epigenetics may enable a better understanding of many human pathologies. Often, epigenetics will retain the memory of exposure, which changes an organism's susceptibility to subsequent, other exposures. This concept may allow new insights into mixture toxicity, especially when the exposures do not take place at the same time. Knowledge on epigenetic processes provides a basis for novel drugs that modify cell phenotypes (e.g., in cancer or neurodegenerative disease). Here, we provide an overview of the role of epigenetics in toxicology, and we call for a systematic assessment of epigenetic changes as part of investigative, and possibly regulatory, toxicity assessments. We propose tools and strategies for using human-relevant models, biomarkers, and AI to better predict who may be at risk. Ultimately, adding epigenetics to toxicology will help us create safer products and protect vulnerable individuals and future generations. Plain language summary Our genes are not the whole story when it comes to how our bodies respond to chemicals, medicines or environmental stress. A level of control-called epigenetics-determines which genes are turned on or off without changing the DNA itself. Detrimental epigenetic changes can be caused by external exposures like pollution, diet or drugs, and they may affect our health in long-lasting ways. This article explores how such "epigenetic toxicity" works, how it can lead to diseases like cancer or developmental disorders, and why current safety testing may miss it. We argue that epigenetic effects should be considered alongside traditional genetic damage in safety assessments of new drugs. We propose tools and strategies for using human-relevant models, biomarkers, and AI to better predict who may be at risk. Adding epigenetics to toxicology will help us create safer products and protectvulnerable individuals and future generations.
Microphysiological systems (MPS) are envisioned to improve drug approval success rates. Yet, integration of MPS into drug development processes has been hampered in part by uncertainties in data translation. To speed adoption, development of animal cell-based MPS is advocated by the pharmaceutical industry. In our view, animal MPS availability would fill a key gap in the ability to examine in vitro to in vivo translatability. Since in vivo animal data will be available and guide decision making in regulatory activities for the foreseeable future, there is significant opportunity for translational assessments. In vivo animal study findings that are recapitulated using in vitro models generated from the corresponding animal species provide validation that those models possess the relevant and necessary attributes, e.g., species-specific pharmacodynamics, metabolism, transport, susceptibility to toxicity, etc. Results from the corresponding human models can then be interpreted with greater confidence for the relevant context of use (COU). Some drugs do not get to the clinic due to adverse findings in animals, so there is considerably more data to directly compare to animal in vitro models than human systems. Another benefit of animal MPS is that drug candidates exhibiting animal safety findings might be easier to derisk, for example if the finding was observed in animal but not the corresponding human in vitro model. This paper reviews considerations and recommendations for adopting animal MPS models in drug discovery and development and describes how their deployment is consistent with 3Rs principles.
Reproducibility of cell culture experiments between laboratories needs to be improved by ensuring more complete reporting of methodology in scientific papers. The minimum reporting standards sug-gested here cover various cell culture methods including monolayers, stem cells, organoids, and microphysiological systems (MPS). The standards build on existing guidance like Good Cell Culture Practice (GCCP 2.0) and OECD test reporting guidelines on how to quality-assure in vitro work, focusing on transparency and completeness of reporting. Key elements to be reported include full details of cell source and identity, cell quality control and characterization, materials and reagents used, culture conditions and protocols, experimental design, data analysis, data availability, and legal and ethical aspects. For complex models, additional details need to be provided such as cell ratios, microenvironment conditions, functional characterization, etc. The guidance for Good In Vitro Reporting Standards (GIVReSt) is part of a broader initiative of evidence-based toxicology encompassing the improvement of the quality of in vitro studies for safety assessments and regulatory decisions. In summary, GIVReSt addresses incomplete reporting as a major factor affecting repro-ducibility of cell culture experiments by providing clear standards around transparency and rigor in reporting. The integration of agentic artificial intelligence (AI) is envisioned to streamline compliance checking, providing real-time feedback and accelerating scientific discovery by making high-quality evidence more accessible. This should lead to more reliable cell culture research overall.
New approach methodologies (NAMs) refer to any technology, methodology, or combination thereof used to inform on chemical hazard and risk, and support replacement, reduction, or refinement of animal use. While the development and application of NAMs has recently increased, their adoption in regulatory decisions is slow and awareness outside the community is low. The Consideration of Alternative Methods Working Group (CAMWG) within the Interagency Coordinating Committee on the Validation of Alternative Methods focused on understanding how NAMs are considered by stakeholders and identifying ways to encourage adoption. A set of questions was developed to focus stakeholder discussions; CAMWG members met with stakeholder representatives to collect perspectives on how alternatives to traditional animal tests are considered when developing toxicology testing and research programs. Participants represented agrochemical, industrial chemical, consumer products, and pharmaceutical companies; academic researchers in toxicology; and Institutional Animal Care and Use Committees. All stakeholders currently use or consider NAMs—some more than others. Challenges to broader NAM adoption were identified and five common themes emerged as potential barriers: perception, regulatory acceptance, scientific and technical limitations, education, and financial considerations. Solutions to overcoming barriers were identified, such as tailored education, proactive collaboration and improved communication. Additional recommendations were ensuring fit-for-purpose use of NAMs, developing harmonized national and global acceptance criteria, identifying funding sources, increasing awareness about NAMs strengths and limitations, and a need for a more central repository for NAMs information. Here, we detail these discussions about NAMs use, barriers, and proposed solutions, to successfully expand awareness, consideration, and adoption.
The use of non-human primates (NHPs) in biomedical research entails significant ethical considerations, demanding careful evaluation of both scientific necessity and research outcomes. This study presents a retrospective literature review comparing non-technical summaries (NTS) of research projects authorized in France between 2016 and mid-2019 with corresponding peer-reviewed scientific publications. The primary objective was to assess the publication rate of NHP-based projects, with secondary outcomes including time to publication, discrepancy in animal use reporting, and the scientific impact of published results. Literature searches were conducted primarily via PubMed, supplemented with additional methods such as author-based searches. Out of 191 projects analyzed, 56% led to at least one publication, the publication rate varying markedly, ranging from 83% in ophthalmology to 30% in immunology. In most cases, publications reported fewer animals than originally authorized: 1,751 actually used out of the 3,649 planned. 2,421 animals had been authorized for the projects for which no publication could be identified. The overall median Relative Citation Ratio (RCR), representing the field- and time-normalized citation rate for published studies, was 1.1, indicating a moderate scientific impact. These findings highlight the need for greater transparency in reporting, including the publication of negative or inconclusive results. The study underscores the importance of systematic retrospective assessments, improved harm/benefit evaluations under the EU Directive, and stronger upstream review mechanisms. Key recommendations include pre-registration of studies, mandated publication of all research outcomes, and the development of open-access platforms to facilitate data sharing, reduce unnecessary duplication, and enhance both ethical and scientific value.
