The chemical universe to which humans and ecosystems are exposed is expanding at a rate that traditional toxicology cannot match. Estimates of the commercial global chemical inventory have moved from tens of thousands in the 1980s to roughly 350,000 distinct substances registered for commercial use in 2019, and exposures now arrive as mixtures, transformation products, and engineered materials whose number grows faster than any list can capture. Against this curve, conventional regulatory testing has at best advanced linearly, and in some jurisdictions it has stagnated outright. The mismatch is not a regulatory failure of will; it is a structural impossibility, because animal-based assays cannot be made exponentially cheaper, faster, or more numerous. The escape from the bottleneck has come from a different family of methods entirely. New approach methodologies (NAMs), and in particular high-throughput in vitro screening, computational toxicology, microphysiological systems, and most recently artificial intelligence, are themselves on exponential growth curves. This article argues that the only coherent response to exponential exposure is exponential testing, and that the field must now think and plan in exponential rather than linear terms. I examine why this is hard, drawing on cognitive science and on Hartmut Rosa's diagnosis of social acceleration, and propose how an exponential toxicology can be built without sacrificing scientific rigor. Plain language summary The number of chemicals being made and sold is doubling and redoubling, and people now meet thousands of them every day in food, air, water, and consumer products. Traditional safety testing, by contrast, has barely sped up in fifty years, because it depends on slow and expensive animal experiments. The gap between what is being produced and what is being tested is widening, not closing. The way out is to use modern tools that can match this pace, including cell-based assays run on robots, computer models that learn from existing data, and miniature organs grown from human cells. This article explains why this matters, why people find exponential growth so hard to imagine, and what regulators, scientists, and citizens can do to keep safety testing in step with a world that is changing faster every year.
Recommendation 013/2026 of the German Federal Institute for Risk Assessment (BfR) considers a reduction of concurrent control animals possible when historical control data demonstrate stability. However, the recommendation provides no method for quantifying stability or determining the magnitude of reduction. We present a framework that addresses this gap for designs whose control group size follows from a power calculation on pre-specified continuous endpoints, each assessed separately. Designs whose group size is prescribed by a test guideline, as in standard repeated-dose toxicity studies, fall outside its scope. The framework offers two approaches scaled to the available expertise. Both begin with a shared planning step that estimates how much the control-group mean of an endpoint varies across studies of the same design and assigns that endpoint, not the laboratory, to one of three stability zones (stable, moderate, or unstable). In the stable zone, a frequentist approach maintains 80% power while accepting a modestly larger minimum detectable effect and requires no Bayesian expertise. A full Bayesian approach achieves a greater reduction at constant power and effect size, provided concurrent and historical controls agree, but requires statistical expertise. We demonstrate, using 89 artificial historical control animals across 8 studies, that the frequentist approach saves 12 of 37 control animals per study and the Bayesian approach up to 32, compared with zero under the status quo. Naive pooling, the intuitive but incorrect alternative, can inflate the false positive rate from 5% to over 30%. A freely available planning tool (C*Power) supports the planning step. Plain language summaryWhen researchers test new substances in animals or run intervention and behavioral studies, they need an untreated control group. Its size is often set by a statistical formula that ignores all previous experiments, although many laboratories have years of consistent control data. We developed a framework that uses this evidence to reduce control animal numbers for measured outcomes such as body weight, wherever such a formula sets the group size. It does not apply when a testing guideline prescribes how many animals each group must contain. The framework offers two options: a simpler approach that needs no specialist statistics, and a more powerful one for laboratories with statistical support that saves the most animals. A free online tool guides researchers through the planning and drafts the documentation for ethics applications. This supports the 3R principle of reduction, as encouraged by the German Federal Institute for Risk Assessment (Recommendation 013/2026).
Vaccines are administered to large populations of healthy individuals, including vulnerable groups, making an exceptionally rigorous safety profile a prerequisite for their approval and use. Nonclinical vaccine safety assessment has traditionally relied on standardized animal testing methodologies, which face increasing challenges: the ethical imperative to reduce animal use, the need for faster vaccine development, and the limitations of animal models in predicting human-specific immune responses. The advent of modern vaccine platforms such as mRNA technologies and viral-vector vaccines necessitates a re-evaluation of safety frameworks. This article explores emerging strategies to modernize vaccine safety assessment, integrating new approach methodologies (NAMs), including in vitro systems, artificial intelligence, and adverse outcome pathways (AOPs). Ethical refinements to animal husbandry practices, including group housing, humane endpoints, pain management and scaled blood sampling volumes, are also discussed to align with the principles of the 3Rs (replacement, reduction, refinement). The adoption of molecular and omics-based techniques, along with innovative validation strategies, offers opportunities to improve predictive accuracy and mechanistic insight. The article highlights a strategic framework for implementing NAMs across the three stages at which animals are still used – early screening, late nonclinical development, and batch release – while ensuring alignment with regulatory requirements. Implementation must be careful enough to ensure that the use of NAMs in the safety evaluation of vaccines maintains or exceeds current safety assessment standards. By addressing scientific, ethical, and regulatory challenges, NAMs have the potential to transform vaccine safety assessment, enhancing human relevance and accelerating innovation in vaccine development. Plain language summaryThis article examines the evolving landscape of vaccine safety assessment, focusing on opportunities to modernize testing approaches through new approach methodologies (NAMs). Drawing on recent discussions with stakeholders across industry, academia, and regulatory bodies, we outline key challenges and strategic pathways for implementing NAMs in vaccine development while maintaining robust safety standards. Special attention is given to emerging technologies, such as human-relevant test systems, artificial intelligence, and virtual control groups, to the replacement of animal-based potency, pyrogen and adventitious agent tests used for batch release, and to considerations for regulatory acceptance and implementation strategies.
Validation underpins confidence in test methods and integrated testing strategies, yet while the methods under evaluation have been transformed, the design of validation studies remains governed by convention and feasibility rather than prospective optimization. I propose model-informed validation design (MIVD), a quantitative framework that treats a validation study as a decision problem to be designed before resources are committed. This expands on our e-validation concept from 2024. Inspired by model-informed drug development, MIVD represents the pathway from target substance population and reference construct through biological and laboratory variability, analysis, and decision rules as an executable computational model, a digital validation twin. Candidate designs are then compared by simulation using assurance, information economics, adaptive rules, and multi-objective optimization, with the aim of maximizing the probability of the correct, appropriately scoped validation decision rather than minimizing effort or maximizing the chance that a method will pass. Because reference standards are imperfect and contexts of use evolve, a validation becomes cumulative and can be updated along a lifecycle. A mean-based proxy reconstruction of the 1996-1997 EPISKIN skin corrosion validation shows how the preferred validation design shifts with the acceptance criteria, reference uncertainty, and chemical selection, while demonstrating that additional testing does not necessarily produce proportionately greater decision value. MIVD reframes validation from the evaluation of methods to the prospective design of the evidence by which those methods are judged. Plain language summaryBefore a new safety test can be used, it must be checked, or validated, to show that it produces reliable and correct results. Such validation studies are expensive, because many chemicals are tested in several laboratories. Today validation studies are mostly designed based on past practice, rather than by working out in advance which experiments would settle the important questions. This article proposes a way to design validation studies deliberately. A digital twin imitates and compares different study designs before the laboratory work begins. Revisiting a skin corrosion validation study done in 1996-1997, the approach shows that the amount of testing needed depends strongly on how much confidence in the results is required, how reliable the reference classifications are, and which chemicals are selected. Designing validation studies, rather than assembling them from templates, is the next step for safety science.
