Do we need animal testing for medicine?

The objection

“Be honest, without animal testing, we’d have no medicine. People would die. You can’t be against that.”

The answer

Animal studies have fed into medicines people now rely on, and the objection is right about their purpose: before anyone takes a new drug, tests on animals are meant to show what harm it might do. They catch much of that harm but miss a real share of it, and most treatments that pass them still fail once they are tried in people, which matters to patients as much as to the animals used in the tests. How far those failures are the animal model’s fault is disputed, and the authors of a 2024 review of 367 therapies tested in animals conclude that the rate of successful animal-to-human translation “may be higher than previously reported” [1].

Do animal tests predict a drug’s harmful effects in people?

Animal tests flag most of the harmful effects that later show up in people, though not all of them. In a survey of 12 pharmaceutical companies (Olson et al., 2000) covering 150 compounds that produced 221 toxic effects in people during clinical development, tests in rodents and non-rodents between them had flagged 71% of those effects, non-rodent species alone 63% and rodents alone 43% [8]. That leaves almost three in ten missed, and because the survey only covered compounds that went on to harm people, it cannot show how often animal tests raise alarms that never apply to humans [8].

How often do drugs that pass animal tests fail in people?

Most treatments that reach human trials after animal testing fail there: of the drug programmes (one drug tested for one condition) that entered trials between 2000 and 2015, only 13.8% ended in approval, according to an MIT analysis of more than 400,000 trial records (Wong, Siah & Lo, 2019) [2]. Drugs entering human trials have already cleared preclinical testing [3], which for a new medicine has routinely meant tests on animals [4][5][9]. The 2024 review by Ineichen and colleagues, led from the University of Zurich, pooled 122 systematic reviews covering 367 therapies for 54 diseases and followed them from the animal stage onwards: half reached any human study, 40% reached a randomised controlled trial and 5% were approved, a median of ten years after the animal studies [1].

Where the 2024 review could compare the two stages, positive results were common in both. Pooled across the 62 therapies with at least five animal studies each, the share of human studies with positive results was 86% of the share among animal studies, which the authors call an 86% concordance [1]. To explain why so few treatments then survive larger randomised trials, they lean towards weak design, such as missing randomisation and blinding, in both the animal studies and the mostly small early trials that follow, over the other possibility they raise, that strict requirements for randomised trials and regulatory approval leave useful treatments behind [1]. They also suggest that fields which largely bypass animal testing, such as medical devices, might turn out to have similarly modest translation rates, and note that animal studies can deepen the understanding of a disease even when no treatment comes of them [1]. Their own sample is likely to be biased, they warn, because reviews of translation may be more common in fields where some treatments have already reached patients [1].

Why do drugs that worked in animals fail in people?

Drugs that looked promising in animals mostly fail in people because they prove ineffective or unsafe. Summarising trial data from 2010 to 2017, pharmaceutical scientists led from the University of Michigan (Sun et al., 2022) attributed 40 to 50% of clinical failures to lack of efficacy and 30% to unmanageable toxicity, with 10 to 15% down to poor drug-like properties and 10% to commercial and strategic reasons [3]. Not all of that is the animal model’s fault, but part of it is built in. Sun and colleagues point to the biological differences between cell cultures, animal models of a disease and the human disease itself [3], and the 2024 review gives stroke as an example: treatments tested on young, healthy animals may not carry over to elderly patients with several other conditions [1].

Is animal testing still required for new medicines?

Animal testing is still routine for new medicines, though the legal requirement has started to go, beginning in the US. The FDA Modernization Act 2.0, signed in December 2022 (Han, 2023) [6], replaced “preclinical tests (including tests on animals)” with “nonclinical tests” in the US law on what a drug’s developer reports before it is first given to people, and defined nonclinical tests to include cell-based assays, organ chips and computer models as well as animal tests [7]. The US Food and Drug Administration’s own requirement outlived that change: in April 2025 the agency announced plans to phase it out, starting with monoclonal antibodies, and said thousands of animals a year, including dogs and primates, could eventually be spared [5]. In the UK, the government’s 2025 strategy for replacing animals in science aims to use validated alternatives to reduce the use of dogs and non-human primates in dedicated pharmacokinetic studies for human medicines, which track how the body absorbs and clears a drug, by at least 35% by 2030, and says it will keep supporting animal research where no viable alternative exists [9]. In the EU, the law sets the full replacement of animals used in science as its final goal while still regulating the animal studies done to develop new medicines, and the European Commission says alternatives need to be accepted by regulators internationally before animal use can truly fall [4].

What could replace animal tests?

Methods built on human biology are the main candidates: organoids grown from human cells, organ-on-a-chip systems and AI-based computer models of toxicity, which the FDA said in 2025 it would promote, adding that human-based tests can reveal toxic effects that could easily go undetected in animals [5]. They have limits, and Han (2023) cautions that for some uses, such as organ replacement therapies, non-animal testing may not be an adequate alternative for the foreseeable future [6].