“Animal testing remains necessary to ensure drug safety.”
For decades, regulators required drugs to be tested in animals before human trials. In 2022 the US dropped that legal requirement, and in 2025 the FDA and NIH announced plans to reduce reliance on animal studies, starting with monoclonal antibodies. Organ-on-a-chip systems, human cell models and computational toxicology are maturing quickly.
Proponents of animal testing argue that a whole living organism remains the only way to see effects across interacting organ systems before risking humans. Critics point out that animal results often fail to translate, and that around 90% of drugs that pass animal testing fail in human trials. The empirical question is whether non-animal methods can replace animals for safety, not just supplement them.
Safety cannot currently be assured to an acceptable standard without animal studies for at least some drug types. "Necessary for all drugs" is a stronger claim.
Detecting toxicity before first-in-human trials and before approval: organ toxicity, reproductive and developmental effects, cancer risk.
Animal studies catch many toxicities before humans are exposed. A multi-company analysis found animal tests predicted most human toxicities observed in clinical trials.
Olson and colleagues examined 150 compounds with human toxicities from 12 companies: rodent and non-rodent tests together detected 71% of the human toxicities, with non-rodents (dogs, primates) more predictive.
Olson et al. (2000). Concordance of the toxicity of pharmaceuticals in humans and in animals. Regulatory Toxicology and Pharmacology, 32(1), 56–67.
Limitation
Concordance studies typically only look at drugs that reached humans, missing drugs wrongly abandoned because of animal toxicity that would not occur in people.
Many dangers emerge from interactions between organs, metabolism and development over time: a drug's metabolite made in the liver damaging the heart, or effects on a developing foetus. Isolated cell and chip models cannot yet reproduce a whole organism.
Thalidomide's birth defects led directly to mandatory reproductive toxicity testing. Developmental toxicity still relies heavily on animal studies, since no validated non-animal battery covers all stages of embryo-foetal development.
ICH S5(R3) guideline on detection of reproductive and developmental toxicity (2020).
Limitation
Thalidomide's effects did not appear in rats, the main species tested at the time, so it also illustrates the limits of animal models.
Regulators moving away from animals are proceeding cautiously, product class by product class. Even the FDA's 2025 roadmap targets monoclonal antibodies first because their effects are especially poorly predicted by animals, not because alternatives are ready for everything.
The FDA's April 2025 roadmap planned a stepwise approach, beginning with monoclonal antibodies and later extending to other biologics and eventually new chemical entities, with animal data still accepted where needed.
FDA (April 2025). Roadmap to Reducing Animal Testing in Preclinical Safety Studies.
Limitation
The roadmap's direction of travel is toward replacement; caution about speed is not a claim of permanent necessity.
About 90% of drugs that pass preclinical testing, mostly in animals, fail in human trials, many for safety reasons. Animals are a poor model for human biology.
Analyses of clinical development find around 90% of drugs entering phase 1 fail to reach approval; roughly 30% of failures are due to toxicity not predicted earlier.
Sun, Gao, Hu & Zhou (2022). Why 90% of clinical drug development fails and how to improve it? Acta Pharmaceutica Sinica B, 12(7), 3049–3062; Van Norman (2019). JACC: Basic to Translational Science, 4(7), 845–854.
Limitation
Failure rates also reflect lack of efficacy, commercial decisions and trial design. Without animal screening, more toxic drugs would reach humans.
Human-based models can already outperform animals for some toxicities. A liver-chip correctly flagged most drugs known to harm human livers that had passed animal testing.
Ewart and colleagues tested 870 liver-chips on 27 drugs: the chip identified 87% of hepatotoxic drugs (which had passed animal studies) with 100% specificity.
Ewart et al. (2022). Performance assessment and economic analysis of a human Liver-Chip for predictive toxicology. Communications Medicine, 2, 154.
Limitation
Study run by the chip's manufacturer, on a small set of drugs, for one organ. Liver toxicity is one of many safety endpoints.
Animal testing has already failed catastrophically in some cases. A first-in-human trial nearly killed six volunteers despite doses based on primate studies showing no harm.
In the 2006 TGN1412 trial in London, six volunteers suffered a cytokine storm and multiple organ failure within hours; the dose had been 500 times lower than one found safe in monkeys.
Suntharalingam et al. (2006). Cytokine Storm in a Phase 1 Trial of the Anti-CD28 Monoclonal Antibody TGN1412. NEJM, 355, 1018–1028.
Limitation
Monoclonal antibodies targeting human immune receptors are a special case; dosing rules were changed in response.
Animal models fail to predict some human toxicities, and some drugs that harm animals may be safe in people.
US law no longer requires animal testing before human trials (FDA Modernization Act 2.0, 2022).
Non-animal methods are advancing fastest for single-organ toxicities and for biologics.