Do animal studies matter for humans?

Research growth. PubMed, searched with microplastics[tiab] AND (human[tiab] OR health[tiab]) and counted by year of publication, all years retrieved on 2026-09-24: 11 papers in 2015, 25 in 2016, 36 in 2017, 97 in 2018, 192 in 2019, 383 in 2020, 522 in 2021, 825 in 2022, 1,103 in 2023, 1,416 in 2024 and 1,739 in 2025. From 2015 to 2025 that is 158 times as many a year, the figure used on this page and in the chart.^[3]

Studies in mice and rats. PubMed on 2026-09-24: (microplastic*[tiab] OR nanoplastic*[tiab]) AND (mice[mh] OR rats[mh]) NOT review[pt] gave 1,111 papers. Adding title and abstract words (mice, mouse, rats, rat) gave 1,339. These counts include every rodent study that mentions microplastics, not only health studies, so the page says "more than 1,100".^[3]

Dates. Each year on the timeline cards is the year the study was first published online. The Italian placenta study, for example, went online in December 2020 and appeared in print in January 2021; the Seoul lung study went online in September 2020; the Nanjing stool study went online in December 2021. "158x" describes research on microplastics and health (papers that mention humans or health); the full set of microplastics papers grew about 51 times over the same decade.^{[3, 47, 86, 87]}

Studies in people. No clean PubMed count exists for studies that measured plastic in people; the "Humans" label also covers human cell studies and reviews. The page uses the 2024 review's count instead: 28 rodent studies and 3 human studies met its criteria for fertility, the gut and the lungs.^[1]

Olson 2000's 71% is the share of human harms that animal tests had also shown (sensitivity). It does not say how often an animal finding comes true in people. The survey covered only medicines that went on to harm people, and it was an industry project; the first author worked at Pfizer. Only the abstract was read, because the paper is paywalled.^[2]

Monticello 2017 also reports that animal tests flagged 48% of the harms seen in first human trials, and that when animals showed harm to an organ, people showed the same harm 43% of the time. Part of the 86% reflects that most drugs do not harm most organs. The authors worked at drug companies and the IQ Consortium, an industry group. Abstract only.^[23]

Bailey, Thew and Balls (2014), writing from the British Union for the Abolition of Vivisection and FRAME, analyzed 2,366 drugs. They found that no harm in rats, mice or rabbits gave "little or virtually no evidential weight" that people would be spared, and that harm in these species "can add considerable evidential weight for human risk", though how much varied more than 100-fold between drug types. Both sides of this debate have interests: the supportive studies come from industry, the critique from anti-vivisection groups.^[78]

Clark and Steger-Hartmann (2018, authors at Elsevier and Bayer) studied 3,290 approved medicines and 1,637,449 reported adverse events over more than 70 years. They report that the study "confirmed the general predictivity of animal safety observations for humans", with heart-rhythm changes in dogs among the clearest matches. Abstract only.^[24]

The EPA says almost one-third of human cancer-causers were identified after animal testing, citing Huff 1993, which counted nearly 30 agents for which the animal evidence came first. Wilbourn 1986, the source of the 84% figure, and Huff 1993 were read as abstracts only. In the vinyl chloride case, the 1971 rats developed skin, lung and bone tumors, not the workers' liver cancer; the 1971 and 1974 papers were checked by title and date only.^{[10, 25, 26, 27, 28]}

Sun 2022 attributes the failure of about 9 in 10 drugs in clinical trials to lack of effect (40% to 50%), unmanageable side effects (30%), poor drug-like properties (10% to 15%) and business reasons (10%), from trial data for 2010 to 2017.^[72]

1937: a new US medicine, Elixir Sulfanilamide, killed more than 100 people. It had never been tested on animals or people, and it was mixed in a poison related to antifreeze. 1938: Congress passed a law requiring drug companies to send the FDA safety data before selling a new drug.^[6]

1961: thalidomide, a pill for morning sickness sold in 46 countries, was banned. In all, more than 10,000 children were born with severe birth defects. 1962: a new US law let the FDA ask for animal test results before any testing in people. By the 1970s, drugs taken for years were usually tested in 2 kinds of animal, over the animals' whole lives.^{[6, 73, 93]}

1997: the US, Europe and Japan agreed on one rulebook for safety-testing new medicines. Its 2009 update still calls for tests in 2 kinds of animal before a drug's first human dose. 2022: US law stopped naming animal tests and now lists lab and computer tests beside them. 2026: the FDA wants to rely less on animal tests and is adding tests on computers and human cells, yet it still calls animal safety tests "a critical method to identify potential risks to human health."^{[4, 17, 19, 20]}

ICH M3(R2) says the length of the animal studies "conducted in two mammalian species (one non-rodent) should be equal to or exceed the duration of the human clinical trials", up to set maximums, and its worked examples name the dog as the non-rodent. It allows less, or different, animal support for very small exploratory first-in-human studies, so "every trial needs two species" would be too strong. The FDA's 2025 roadmap names dogs and primates among the animals used.^{[4, 21]}

The first human dose: "In general, the No Observed Adverse Effect Level (NOAEL) determined in nonclinical safety studies performed in the most appropriate animal species gives the most important information."^[4]

1937 and 1938: the FDA's historian writes that Elixir Sulfanilamide "was not tested in animals or humans prior to marketing", that its solvent, diethylene glycol, "was a poison (chemically related to anti-freeze)", and that the tragedy "killed over 100 people". The 1938 Food, Drug, and Cosmetic Act added "a new provision ... requiring drug sponsors to submit safety data to FDA officials for evaluation prior to marketing".^[6]

The 2016 "forever chemicals" finding: "The NTP concludes that PFOA is presumed to be an immune hazard to humans based on a high level of evidence that PFOA suppressed the antibody response from animal studies and a moderate level of evidence from studies in humans." The monograph reaches the same conclusion for PFOS.^[92]

"For more than 50 years": the FDA's historian writes that within a few years of the 1962 drug amendments "a fairly standardized set of animal toxicology studies to precede and support human trials was in place", and that by the 1970s a drug for chronic use was generally tested in two animal species. The 1962 law allowed the FDA to require those reports; it did not itself order animal tests.^[6]

The Declaration of Helsinki, the world's code of ethics for medical research, says research on people must rest on "adequate laboratory and, as appropriate, animal experimentation."^[7]

The 2022 change (Pub. L. 117-328, §3209, 29 December 2022) substituted "nonclinical tests" for "preclinical tests (including tests on animals)" in 21 U.S.C. 355(i)(1)(A). It made animal tests one option among several; it did not ban them. The FDA's direct final rule of 22 September 2026 (FR Doc. 2026-19350) replaces "laboratory animals or in vitro" with "nonclinical" studies in 21 CFR 312.23(a)(8), effective 4 February 2027 unless significant objections arrive by 7 December 2026, and the FDA says it does not expect the change to affect industry practice.^{[17, 18, 19]}

The FDA's April 2025 roadmap aims to make animal studies "the exception rather than the norm" for drug safety testing within 3 to 5 years, starting with antibody drugs. In July 2025 the NIH said it will no longer write funding calls that support only animal research, while animal models "continue to be important".^{[21, 22]}

The EPA's default assumptions: "an agent that produces an adverse developmental effect in experimental animal studies will potentially pose a hazard to humans"; "An agent that produces an adverse reproductive effect in experimental animals is assumed to pose a potential threat to humans"; and "an agent that produces detectable adverse neurotoxic effects in experimental animal studies will pose a potential hazard to humans."^{[11, 12, 13]}

The BPA example is about a plastic chemical, not microplastic particles. The EU's 2024 regulation sets a tolerable daily intake 20,000 times lower than the 2015 temporary one, and also records earlier bans on BPA in baby bottles and in cups for infants. Some agencies, including the European Medicines Agency, formally disagreed with parts of the 2023 opinion.^{[14, 15]}

Europe's medicines agency adds that the animals chosen should respond to the drug the way people do, and that companies must explain their choice.^[5]

Access. Read as abstracts only: Jani 1990, Huang 2015, Amereh 2022, Ragusa 2021, Li 2020 (rats), Zhao 2023, Hong 2024, Pan 2026, Wang 2022 and Lee 2022. Leslie 2022's "17 of 22" comes from Aelo's full-text reading of 29 July 2026 (Results, Table 2); on 2026-09-22 and 2026-09-24 the publisher blocked automated reading, and the abstract gives only "22 volunteers" and a mean of 1.6 µg/mL. Read in full: Lee 2024, Fournier 2020, both 2022 Nanjing papers, Jeong 2025, Amato-Lourenço 2024, Wang 2023, Marfella 2024, Hu 2024 (with its published comment and reply) and the two Louisville studies.^{[43, 44]}

Placenta row. The Tehran study (Amereh 2022) was read as an abstract. Its 2 groups were picked by baby size: 13 pregnancies where the baby grew too slowly and 30 normal ones. Plastic turned up almost only in the first group, so its link between plastic and smaller newborns partly restates how the groups were chosen. The abstract says 3 of 30 normal placentas held plastic; the paper's own figure appears to show 4, so the story leaves that number out. The 2020 Rutgers rats had plastic placed in the windpipe, not swallowed.^{[46, 49]}

Testicles row. In 2021 the Nanjing lab reported that 4 and 10 µm particles reached mouse testicles within 24 hours and that sperm quality and testosterone fell after 28 days; its abstract does not say how the particles were given, so the story uses the lab's 2022 drinking-water study instead. The 2022 study reports its sperm results only as bar charts; read off Figure 2, about 9 in 10 sperm were alive in clean-water mice against about 4 to 6 in 10 in plastic-water mice, and misshapen sperm rose from about half to between about two-thirds and three-quarters. Those chart readings are approximate and are not printed in the story. Jeong 2025 fed 50 to 100 µm bits of worn PET in food at 5 mg a week; the authors call this an "annual intake relevant dose", an estimate this page does not repeat. Its motility result is shown in videos, with no number. Zhao 2023 ("male subfertility" in mice exposed before and after birth through drinking water) was read as an abstract.^{[50, 51, 52, 53]}

A size puzzle. In the 1990 rat study, particles larger than 300 nanometers were not found in the blood. Several human studies report much larger particles in blood and organs. Possible reasons include people taking up larger particles than rats do, and lab contamination.^{[42, 62]}

In the 2024 blood study, blank samples with no blood picked up 0 to 5 plastic bits on their own, and the team did not subtract them. The blood itself averaged 4.2 bits per milliliter. The link to the inflammation marker (hsCRP) is a correlation in 36 people.^[44]

The heat-based test used by the New Mexico lab (placenta, testicles, brain), in the artery study and in the 2026 heart study can mistake fat for polyethylene. Nine scientists from across Europe (Monikh and colleagues, 2025) challenged the brain study over limited contamination controls and lack of validation steps; the authors replied. In the brain study, the dementia figures came from a single digestion step, and the paper's table notes that values fell after extra fat removal, with no new number given.^{[54, 63]}

The 2023 artery-buildup study used mice bred to clog easily (ApoE-knockout), on a high-fat diet, and its clearest effect came at the top dose. The artery study's own authors note that most animal studies use far more plastic than they measured in people. The route in the 2020 rat heart study is not stated in its abstract.^{[64, 65, 66]}

In the 2025 Beijing-led study, the blood-vessel blockages and the behavior changes followed plastic injected into the blood; the tube-fed mice showed only that plastic reached brain vessels. The mice's balance and grip were back to normal by day 7 to day 28. The authors call it "premature" to apply their result to people.^[33]

Weight. The 2024 stool study grouped 121 young adults by how often they ate take-out food and found a positive link between plastic in stool and body mass index; it is a snapshot, not a study that followed people over time. A 2026 study of 1,080 ten-year-olds linked some plastic types to overweight and others to underweight. The larger particles in the 2024 Louisville study did not raise the mice's weight, and the 2018 Hangzhou mice given the higher amount lost weight.^{[69, 70, 71, 84]}

Liver, gut and lungs. The 2018 Hangzhou mouse study, the 2020 Seoul rat study, the 2021 São Paulo lung study, the 2022 England lung study and the 2022 Nanjing stool study were read as abstracts; the 2022 Hamburg liver study was read in full. The Hamburg study is a proof-of-concept case series of 11 livers, and its authors say plastic build-up could be a cause or a result of liver scarring. The Nanjing authors say bowel disease may make plastic stay longer in the gut. The Seoul rats also showed changes in other measures, but without a dose pattern.^{[84, 85, 86, 87, 88, 89]}

Their own words. The quotes in "Did animal studies point scientists to the human findings?" come from the opening of each human study: Marfella 2024 (author manuscript pp. 4 to 5), Liu 2024 (abstract), Amato-Lourenço 2024 (introduction), Montano 2023 (abstract) and Horvatits 2022 (abstract).^{[62, 66, 85, 90, 91]}

Heart. The 2026 heart study compared heart tissue from 33 surgery patients with high and low scarring; it also gave mice plastic by tube and found worse heart scarring under a stress drug. Abstract only.^[67]

Deng 2017, often cited for plastic moving from the gut to mouse organs, was left out: letters in 2019 questioned whether its uptake figures were believable.^{[82, 83]}

The 2022 Nanjing brain paper was corrected in December 2022. The erratum fixes the author affiliations and one abstract word ("higher" should read "lower"); this page uses the corrected paper.^[59]

These three rows are kept out of the story above. The liver row rests on a small case series, the lung studies are about breathed air rather than swallowed plastic, and in the gut study the authors say the disease itself may keep plastic in the gut longer.

Liver. The 1990 London rat study found the particles in the rats' livers too. In 2018, a Hangzhou team gave mice plastic in their drinking water for 5 weeks, and it disrupted how their livers handled fat. In 2022, doctors in Hamburg found plastic in 6 of 6 human livers scarred by cirrhosis; in livers without disease there was too little to count. Jani 1990 (rats, abstract): about 7% of the 50 nm dose and 4% of the 100 nm dose ended up in the liver, spleen, blood and bone marrow. Lu 2018 (mice, abstract): 0.5 and 50 µm polystyrene by mouth, in µg/L amounts, for 5 weeks; at 1,000 µg/L, lower liver triglyceride and cholesterol and lower expression of fat-making genes. Horvatits 2022 (full text): cirrhotic livers 4.6 to 11.9 particles per gram (mean 8.3) against 0.3 to 1.9 (mean 1.0) in livers without disease, which the authors count as negative given the limit of detection.^{[1, 42, 84, 85]}

Lungs. In 2020, scientists in Seoul had rats breathe plastic for 14 days, and signs of inflammation in their lungs rose with the amount they breathed. In 2021, a São Paulo team found plastic in 13 of 20 human lungs examined after death; in 2022, a team in Hull, England, found it in 11 of 13 lung tissue samples, with the most in the lower lungs. Lim 2021 (rats, whole-body inhalation of polystyrene for 14 days; published online September 2020): lung TGF-β and TNF-α rose with exposure concentration; other measures changed without a dose pattern. Amato-Lourenço 2021: 33 particles and 4 fibers in 13 of 20 autopsy lungs. Jenner 2022: 39 particles in 11 of 13 samples, 1.42 per gram before blank subtraction, lower lobes highest.^{[87, 88, 89]}

Gut. In 2018, a Hangzhou team found that mice given plastic in their drinking water made less of the mucus that protects the gut wall, and their gut bacteria changed. In late 2021, a Nanjing University team found that people with inflammatory bowel disease had about 50% more plastic in their stool than healthy people, and the more plastic, the worse the disease. The authors say the disease may also make plastic stay longer in the gut, and the number of people tested is not in the abstract. Lu 2018 (mice, abstract): gut mucus secretion fell with both particle sizes; richness and diversity of cecal bacteria changed. Yan 2022 (abstract): 41.8 against 28.0 particles per gram of dry stool.^{[1, 84, 86]}

Kosuth 2018 tested 159 tap water samples from 14 countries. 81% held man-made particles, 98.3% of them fibers. The US had the highest mean, 9.24 particles per liter. The team identified particles by microscope and a stain for natural fibers, and wrote that the particles are "most likely" microplastics but that confirming this needs spectroscopy.^[16]

Levine 2023 pooled 223 studies of semen samples collected from 1973 to 2018. Among men not selected for fertility, mean sperm concentration fell 51.6%, and the yearly decline rose from 1.16% after 1972 to 2.64% after 2000. The authors call for research on causes; they do not name one.^[56]

Lokeshwar 2021 used US national survey data on 4,045 adolescent and young adult men, from 1999 to 2016. Average testosterone fell over time, and the fall held among men with a normal body mass index.^[57]

Chartres and colleagues used the Navigation Guide method, with verdicts guided by the National Toxicology Program's OHAT approach. It was a rapid review done for California state lawmakers through the California State Policy Evidence Consortium, on a deadline, so it covered fertility, the gut and the lungs and did not look at weight or memory. After a shared start on 20% of the references, most studies were screened by one person, and a second person checked only the exclusions. Four databases were searched in July 2022 and April 2024. 22 of the 28 animal studies were in mice, 22 used polystyrene, and 22 were done in China. Exposures ran from 14 days to 32 weeks.^[1]

"Suspected" sits one step below "presumed" on the review's four-step scale (known, presumed, suspected, not classifiable). For sperm quality, the review rated the body of evidence "high" and wrote: "We downgraded the evidence from 'presumed' based on the sample size and number of studies." Across the 5 sperm studies, 9 of 9 measured effects pointed toward harm. The abstract mentions lung cancer, but the body and the Conclusion name only colon cancer, as a suspected link. The authors noted that some publications came from the same lab groups, so errors could spread across studies.^[1]

Six authors are at the University of California, San Francisco, and one at the Natural Resources Defense Council, an environmental advocacy group. The review ends with a call for government action. Funding: University of California Office of the President, the JPB Foundation and grant P30ES030284. The authors declared no competing financial interest.^[1]

The 2002 gene figures are counted from the mouse side: "Approximately 99% of mouse genes have a homologue in the human genome", about 80% have one clear partner, and fewer than 1% have none, "and vice versa". Only about 40% of the human genome lines up with the mouse genome letter by letter.^[29]

Europe's food-safety agency reports that in rodents 0.04 to 0.3% of swallowed test particles got past the gut wall, and that about 0.2% crossed human colon tissue in a lab chamber. More than 9 in 10 swallowed micro- and nanoplastics leave in stool.^[32]

The 2025 mouse study's authors give their own, larger vessel measurements, and call it "premature" to apply their result to people. Nobody has measured whether the higher share of neutrophils in human blood changes the risk of these blockages.^[33]

One review of placentas concludes that "the mouse model is often overvalued". Researchers also note that the type of placenta does not play a dominant role in how most drugs cross it.^{[36, 37]}

Scientists still argue about how closely mouse gene responses match human ones. A 2013 study found mouse responses to injury and infection close to random against human ones; a 2015 re-analysis of the same data found a strong match.^{[79, 80]}

Why Aelo does not turn mouse doses into human doses: there is no reliable way to do it for swallowed plastic. The formulas regulators use were built for drugs and chemicals that get into the blood, and more than 9 in 10 swallowed plastic bits pass straight through. So Aelo reports what the animals got and does not guess.^{[32, 40, 41]}

Two more shared parts. Mice and people have the same white blood cells that swallow what does not belong. And a mouse's smallest brain vessels are about as narrow as people's: about 2 to 5 thousandths of a millimeter in mice against about 6 in people.^{[34, 35, 38, 39]}

In 2006 in London, 6 healthy young men took a new immune drug, TGN1412, that had looked safe in monkeys. Within 12 to 16 hours all 6 were critically ill; all 6 survived. The drug flips a switch on certain immune cells. People's cells carry that switch; the monkeys' cells don't.^{[74, 75]}

Thalidomide: mice turned out to be less sensitive to it than rabbits and monkeys. That taught regulators that species can react differently. Since then, drug testing has used several kinds of animal, and there has not been another thalidomide.^[73]

Bailey, Thew and Balls (2014) wrote that the presence of harm in rats, mice or rabbits "can add considerable evidential weight for human risk", though how much varied more than 100-fold between drug types. The story above says "real weight" for that reason. A leading review that criticizes animal testing still opens by saying there is "no doubt" animal research has greatly benefited people.^{[77, 78]}

Animal tests can also flag harm that does not apply to people. The EPA's own example is a kidney tumor in male rats (the alpha-2u-globulin mechanism) that is not a hazard in people. Cancer agencies set such effects aside only with strong evidence that they do not apply to people.^{[8, 10]}

Van Norman 2019 opens by saying "There is no doubt that the use of animals in science and medicine has significantly benefitted human beings", then lists drugs that passed animal tests and harmed people, including fialuridine. The 6 men in the 2006 London trial were given a dose 160 times lower than the monkeys' no-harm dose, after adjusting for body size.^{[76, 77]}

When Aelo writes about an animal study, it names the species in every sentence about an animal result. It says "caused" only for a controlled experiment, and only about those animals. It says "linked" or "went with" for studies of people. It sets out what the studies do not prove in plain view, near the end, and it reports published challenges to a study alongside the study.

The earlier version of this page said 113 men gave a semen sample and plastic was found in all 113. The study recruited 113 men, but only 100 had both semen and urine tested for plastic, and those 100 of 100 had plastic in both.^[55]

It said more kinds of plastic were linked to lower sperm count, concentration and forward movement. In the adjusted results in the body of the paper, only sperm concentration held up.^[55]

It said the 2024 review "did not reach the same conclusion" for obesity or memory. The review never looked at them.^[1]

It said a 2023 heart-tissue study matched all its plastic levels to human levels. Only the lowest of 3 levels was matched. That study is no longer used here.^[81]

It said artery patients with plastic in their plaque were about 4.5 times as likely to have a heart attack, stroke or death. The study's 4.53 compares rates over time after adjusting for other risk factors, so this page says 4.5 times the rate.^[66]

The story was shortened and rewritten in plainer words for readers arriving from a single study. Only the animal findings with a clear match in people stay in the story. These matches were picked to show a pattern; they are not a full review. The 2024 finding of plastic in 23 of 23 human testicles (Hu 2024) moved to the statistics line of the sperm card. The history of animal testing, the liver and lung rows and the details of the two famous misses moved into this technical record, unchanged in substance. The placenta card now includes the 2022 Tehran study (Amereh 2022, read as an abstract), and the 2015 Taiwan mouse study moved to its statistics line. After a Brand Book 3.1 review, the gut row moved here too, and the cards now name the rats' route, the clinic population and the mouse strain. The 2022 Dutch blood study, the BPA example's own section and the tap-water paragraph were folded in or moved here. Every figure in the story is taken from the fact sheets named in the page's source file.^{[16, 43, 45, 49, 54]}