How many animals are used in experiments each year?

The objection

“It’s a handful of mice in a clean lab, surely. People exaggerate this stuff. How bad can it really be?”

The answer

Researchers at a campaign group estimated that 192 million animals were used for science worldwide in 2015, 79.9 million of them in scientific procedures, and nobody knows the true figure because most countries publish none. What is done to them ranges from a blood sample to procedures designed to end in death, and European law spells out which is which.

How many animals are used in experiments each year?

About 79.9 million animals were used in scientific procedures worldwide in 2015, rising to about 192 million with animals killed for their tissues, kept to maintain genetically modified strains, or bred for laboratories and never used, on an estimate published in 2019 by two researchers at the campaign group Cruelty Free International [1]. They standardised the published figures from 37 countries to the definitions in EU law and estimated the other 142 from how much animal research each publishes, and the 79.9 million was up 36.9% on their equivalent figure for 2005 [1]. Great Britain, one of the countries that does count, recorded 2.54 million procedures on living animals in 2025, the lowest under its Animals (Scientific Procedures) Act 1986, and 48% of them were to create and breed genetically altered animals, according to the UK government (2026) [2].

Why is the true number unknown?

Most countries publish no national figures: of the 179 in the 2019 estimate only 37 did, and the authors found no evident increase in that number since 2005 [1]. The United States publishes figures but leaves out most of its laboratory animals. Its Animal Welfare Act regulations cover only warm-blooded animals and exclude birds, rats and mice bred for research [3], and the US Department of Agriculture’s annual summary of research use leaves out the species the Act does not cover [4]. A 2021 study in Scientific Reports gathered figures from 16 large US research institutions and found that rats and mice made up about 99.3% of the mammals they used; scaled up from the 780,070 animals reported under the Act in 2017-18, that suggests about 111.5 million rats and mice a year, a figure its author says is higher than any other estimate and could be too high or too low [9]. In Britain, whose law covers all vertebrates, mice, fish, birds and rats accounted for 95% of procedures in 2025 [2].

How much do the animals suffer?

Most experimental procedures in Great Britain are assessed as causing mild suffering or less, but almost three in ten cause moderate suffering and some are severe: in 2025, on UK government figures, 49% were mild, 17% sub-threshold and 4% done entirely under a general anaesthetic the animal never woke from, while 29% were moderate and 1.3% severe, down from 3.4% the year before [2]. That 1.3% of 1.32 million experimental procedures still comes to about 17,000 [2]. EU law (Directive 2010/63/EU) classes a procedure as severe when it is likely to cause severe pain, suffering or distress, or long-lasting moderate pain, and its examples include toxicity tests where death is the end-point, inescapable electric shock to produce learned helplessness, forced swimming to exhaustion, immobilising rats to induce stomach ulcers or heart failure, prolonged isolation of dogs and monkeys, and tumours left to spread or ulcerate [5]. Opening the skull or chest under anaesthesia and pain relief, with pain afterwards, counts as moderate [5]. The same law requires anaesthesia unless it is inappropriate, judged partly on whether it would be more traumatic than the procedure or incompatible with its purpose, and requires pain relief once the anaesthetic wears off only where that is compatible with the purpose [6]. Procedures involving serious injuries that may cause severe pain must never be done without anaesthesia [6].

Are the Draize and LD50 tests still used?

Both survive in changed form: the rabbit eye test is still an OECD test guideline [11], revised in 2012 to make anaesthetic and pain relief routine [10], and the classical oral LD50 was replaced by tests that can still kill [7]. The EU’s version of the eye test, adopted in 2017 to match the OECD’s 2012 guideline, says animals should be used only when testing is judged necessary after alternatives have been considered, and that some regulatory frameworks still need or require it [10]. Under it a single dose goes into one eye of an albino rabbit, with the other eye as a control; topical anaesthetics, systemic painkillers and humane end-points “should be routinely used”, and any animal in severe pain, or with lesions such as a perforated cornea, should be humanely killed [10]. Great Britain recorded no procedures for UK regulatory eye irritation testing in 2025 [2]. The American Anti-Vivisection Society, a US campaign group, lists the test, devised at the US Food and Drug Administration in 1944, among those performed today, and describes restrained rabbits suffering redness, bleeding, ulcers and even blindness [8].

The classical oral LD50, OECD Test Guideline 401, gave at least three groups of five or more rodents different doses, spaced to produce a range of toxic effects and death rates, and used the deaths to calculate the dose expected to kill half; the OECD deleted it from its guidelines in 2002 [7]. Of the three oral replacements adopted between 1992 and 1998, one can stop at signs of evident toxicity instead of death [7], while in the other two, the American Anti-Vivisection Society says, animals often suffer convulsions and seizures [8]. The classical dermal LD50 (Test Guideline 402) was still the OECD’s only guideline for skin exposure in 2010 [7], and EU law still names single-dose acute toxicity testing among its examples of severe procedures [5]. A 2010 review by scientists from industry, the European Commission, the UK’s National Centre for the 3Rs and Humane Society International found acute lethality testing largely historic for medicines and food additives, but still driven by chemical classification and labelling rules, and saw “little or no scientific basis” for keeping death as the end-point [7].