PLANET OF THE APES

Isabel Braga, MD phD

The public question is always binary: for or against the monkey in the laboratory. That is the question of someone who will not pay the bill. The question that counts has three clauses. What biological question that body answers. What harm it buys. And whether the number that comes out of there crosses the street and reaches the patient. Without the three, the white coat is liturgy.

The number industry does not put in the brochure

In 2024, PLOS Biology tracked therapeutic interventions that first appeared in animals. About 5% reached regulatory registration in humans. Animal–human concordance, in meta-analyses, swings from almost zero to one hundred percent: it is not predictable a priori. [13]

In stroke, about 69% of animal studies read “positive”; about 6% of phase 3 trials in people do. Mean power of the animal experiment, in that series, was 17%. [15] Alzheimer, 2002–2012: success rate of a disease-modifying drug in the region of 0.4%, with an amyloid model that “works” in the mouse and vanishes in the patient. [16] Pound and colleagues reopened six classes already reviewed. Less than 7% of the preclinical studies would pass Bateson’s cube — real human benefit times rigour times harm to the animal. Analgesia poorly reported. A third of the animals dead before the endpoint. [14] Freedman estimated that most biomedical preclinical work does not reproduce. [17]

That does not prove every animal is useless. It proves that most use is signal-fishing, not a bridge. The clinical trial exists precisely because the animal does not predict enough.

Where the model lies on purpose

The model does not “err.” The model is built to give the signal the protocol asked for. The animal is young, isogenic, male, challenged at hour zero, without hypertension, without tobacco, without three antihypertensives. The endpoint is infarct volume or time in a maze. In hospital the endpoint is death, a wheelchair, Mini-Mental. Translating one into the other is an act of faith with a spreadsheet.

There are structural lies, not accidents. In sepsis, the mouse withstands a bacterial load that kills people — and the drug that saves it does not save the patient. In stroke, door-to-needle time in the animal is minutes; in the human it is hours. In Alzheimer, the mouse carries a transgene the human does not have. In routine toxicology, the monkey enters because the agency dossier asks for a second species, not because that molecule needs a prefrontal cortex. Whoever designs the experiment knows this. That is why publication bias is so high: the negative does not travel; the inflated positive does.

The most honest sentence in the field came from someone who funded an HIV vaccine. After Merck’s STEP trial broke, Peggy Johnston, then at NIAID, said in Seattle: “Mice lie, monkeys sometimes lie, humans never lie. Some monkeys lied this time.”

The HIV vaccine the monkey approved and we did not get

For twenty years the rhesus was the poster for an HIV vaccine. Simian virus (SIV) and the SHIV chimera infect the monkey. One can measure load, CD4, rectal mucosa. One can publish “protection.” What did not come was a product. [24]

The teaching case is STEP, Merck’s adenovirus-5 vaccine. In the monkey challenged with SHIV89.6P — a “weak” strain, the field itself later admitted — the immunogen did not stop infection, but knocked down viral load and preserved CD4. That is what justified the phase 2b in three thousand volunteers. In humans, it did not stop infection and did not knock down load. There were more infections in the vaccine arm than in placebo; the difference did not close significance, but the trial stopped. [25][26] Watkins and colleagues, in Nature Medicine in 2008, were dry: the SHIV model did not predict STEP. The harder SIV model had already predicted failure — and was set aside.

HVTN 702, in South Africa in 2020, was stopped: 129 infections on vaccine versus 123 on placebo. It did not reproduce RV144. [27] RV144, in Thailand, had marginal efficacy of about 31% and no effect on viral load (RV152). [28] HVTN 705 / HPX2008 (mosaic Ad26 + gp140) was not significant. [29] BG505 SOSIP in rabbit and monkey produced autologous antibody with limited breadth in humans. [30]

The stock the brochure calls a model

In the United States, the USDA report for fiscal year 2021 — the one the NAS uses — showed a little over 70,000 non-human primates in active use, a number similar to 2008. That is not “existential scarcity.” It is inventory. [19] In FY2024 facilities reported 104,808 primates in total: 44,375 held only, with no protocol; 1,232 used in a procedure with pain the laboratory declared it could not relieve. [19] Almost 71% of NIH awards that declare a species use rhesus. In FY2021 the United States imported 31,844 primates; 30,649 — 96% — were cynomolgus macaques. [20]

Price gives the market away. Before 2020, a laboratory cynomolgus went for 4 to 7 thousand dollars. With the Chinese embargo in the pandemic, it went to 20 thousand. In 2022–23, Evercore analysts and the business press cited peaks of 55 to 60 thousand dollars a head. There was a “future” on an Asian foetus. [21] In Cambodia the hunter received about 175 dollars. The same animal left the country at more than 6 thousand. The U.S. Department of Justice began to treat part of that chain as wildlife sold with captive paperwork. [22]

The chimpanzee, which the United States no longer uses in invasive work, still costs: NIH spent 10.99 million dollars in 2024 to keep 343 animals — 87.78 dollars a head a day. [23] A protocol monkey is not cheap. A sick colony is expensive twice: it loses the biomodel and buys zoonotic risk.

What the American press documented

It is not activist legend. It is FOIA, USDA and magazines.

At NIMH, Elisabeth Murray’s laboratory — neuroscience of “fear” and attributing emotion to animated shapes — had the records of 73 monkeys opened by a freedom-of-information action. Eighty-five percent with hair loss; 74% with a metal pin screwed into the skull; 64% water-deprived in order to “cooperate”; teeth filed in more than half. Public funding in the tens of millions. [34]

In Louisiana, the University’s New Iberia Research Center holds up to 12,000 monkeys — the largest university breeding colony in the country. Internal video in 2026: cage, accumulated faeces, injury. In January 2025, 19 rhesus died in a 2 °F freeze. USDA cited an interstate shipment to Charles River with an expired health certificate, in a season of tuberculosis in American laboratory primates. [36]

The New Yorker retold Alpha Genesis, in South Carolina, and Morgan Island: dozens of escapes in a decade; more than 120 million dollars in federal contract since 2008. [35] Covance, in Vienna, Virginia, was filmed in 2004–05: nasal tube, slaps, a monkey with a broken arm for four days. [37] In 2026 NIH withdrew approval from Caucaseco, in Colombia, after authorities seized more than a hundred Aotus in rusted cages. [38] In a Chinese zoo, 84 rhesus with human tuberculosis; the skin test — the same one many colonies still use as the only filter — was negative in all of them. [39]

The pattern repeats. A fragile test. A colony open to humans. Euthanasia in place of a facility. Silence when the number appears.

What the law asks and what the Brazilian case file already holds

On Brazilian paper there is no shortage of rules. Arouca Law, Decree 6.899, DBCA, 3Rs, RN 60 (pair or group, enrichment), RN 37 (euthanasia), RN 38 (no animal in a merely demonstrative class). [1][2][3][4][5][6] In Europe, a primate only when another species will not serve. [7] FELASA requires tracking M. tuberculosis on arrival and at least once a year. Bushmitz already warned in 2009 that the tuberculin test lies. WOAH wants high-dose palpebral Old Tuberculin. CDC, on import, wants three negative tests. NIH tests again every six months. [8][9][10][11][12]

The file in case no. 5035958-79.2026.4.02.5101, before the 20th Federal Court of the Rio de Janeiro Judicial Section, already joined the deviation. [31] An outbreak of Mycobacterium tuberculosis in the rhesus colony in 2011–2012, with “death of a large number of animals.” A laboratory without a primer for salivary PCR. A serological kit described in the meeting as condemned for routine use. A container without biosafety level for the sick animal. A former staff member who started tuberculosis treatment and stopped because he “had improved.” Agenda item: public repercussions, “it is not yet the moment” to clarify, log the visit because of a lawsuit.

The same case covers 2020–2023. Handling stopped in the pandemic. In October 2020, a testicular granuloma, PCR and culture positive for the human bacillus. In the rhesus holding colony, 34 positive or suspect in 431 — 8%. Declared control: euthanasia of the positive, the suspect and group contacts, including a pregnant female with altered test, PCR and radiograph. The text in the file still asks the presidency for support “to contain the disease.” Rhesus stock: of the order of 550 in 2012, 484 in 2023, in a colony that had already passed 950 head. A drop compatible with the “large number” of deaths. [31] A colony with the human bacillus does not produce clean data. It produces bias. Vervenne described the window in which the monkey still has no granuloma and already sheds. [32]

In June 2026 the same agent closed Ibama’s Cetas in Rio: three dead capuchins, about fifty tests in people, one contact, no one ill. Another institution. The same bacillus. [33]

A ruler

Useful with a primate if all four are yes. The question is of a system and the inferior species failed in writing. N is small and the statistics were pre-registered. Biosafety is real, not a container. The endpoint in the animal is the endpoint that will be measured in the patient.

Useful in mouse, fish, organoid, chip and, when ethical, in human challenge: screening, dose, mechanism, toxicity of a known class. An alternative method CONCEA recognises has a deadline of up to five years to become obligatory. [40]

Useless: repeating the model that already failed in the same disease; a class; a bank of monkeys with no protocol; a maze endpoint sold as survival. Useless, above all, the colony that falls ill from the bacillus in the corridor and then presents itself as a heritage of science.

Arouca and Bateson’s cube coincide. A primate only after substitution was truly tried — not after the breeding colony needs to justify the payroll. The file already joined culture, minutes and euthanasia. It is for the reader to decide whether that is still called research or already called stock.

The question that counts

The public question is always binary: for or against the monkey in the laboratory. That is the question of someone who will not pay the bill. The question that counts has three clauses. What biological question that body answers. What harm it buys. And whether the number that comes out of there crosses the street and reaches the patient. Without the three, the white coat is liturgy.

Isabel de Fátima Alvim Braga — This article interprets published papers and a public case file. It is not a laboratory inspection report.

Numbered sources

1. Brazil. Law 11.794 of 8 October 2008 (Arouca Law).

2. Brazil. Decree 6.899 of 15 July 2009.

3. CONCEA. Normative Resolution 12/2013 (DBCA).

4. CONCEA. Normative Resolution 60 of 2 May 2023 (non-human primates).

5. CONCEA. RN 37 (euthanasia); RN 38/2018 (teaching).

6. Russell WMS, Burch RL. The Principles of Humane Experimental Technique. 1959.

7. European Union. Directive 2010/63/EU, art. 8.

8. Balansard I et al. FELASA Working Group Report, Laboratory Animals, 2019.

9. Bushmitz M et al. J. Medical Primatology, 2009.

10. WOAH. Manual — tuberculosis in primates; palpebral Old Tuberculin.

11. CDC. 42 CFR § 71.53.

12. NIH/OACU. D3-NHP-TB; ILAR Guide.

13. van der Naald M et al. PLOS Biology, 2024. Animal–human translation ≈ 5% to registration.

14. Pound P et al. PLOS ONE, 2018. Bateson cube: <7% permissible.

15. Schmidt-Pogoda A et al. Annals of Neurology, 2020. Stroke: 69% positive in animal, 6% in phase 3; power 17%.

16. Cummings J et al. Alzheimer’s Research & Therapy, 2014. Success 0.4% (2002–2012).

17. Freedman LP et al. PLOS Biology, 2015. Preclinical reproducibility.

18. National Academies of Sciences. Nonhuman Primate Models in Biomedical Research. Washington: NAP, 2023.

19. USDA/APHIS. Animal Welfare Act annual reports. FY2021: ~70,000 NHP in use; FY2024: 104,808 in facilities (44,375 held only; 1,232 in unrelieved pain).

20. NAS/NCBI. FY2021 imports 31,844 NHP, 30,649 (96%) Macaca fascicularis; rhesus is 71% of NIH NHP awards.

21. Nikkei Asia; Evercore ISI; Fortune. Lab monkey price: US$4–7k pre-2020; US$20k in shortage; peaks US$55–60k in 2022–23.

22. RFA / UN trade data. Cambodia 2019: local poacher ~US$175; later export ~US$6,660.

23. NIH ORIP. Chimpanzee Management Report, 1 Oct. 2024: 343 chimpanzees, US$10.99 m/year, US$87.78/animal/day.

24. Watkins DI et al. Nature Medicine, 2008. NHP models and the failure of the Merck HIV-1 vaccine in humans.

25. Cohen J. Science, 2007. STEP/Merck; Peggy Johnston (NIAID): “Mice lie, monkeys sometimes lie, and humans never lie.”

26. Merck / HVTN. STEP (HVTN 502): vaccination stopped; infections in vaccine arm ≥ placebo.

27. HVTN 702 (South Africa, 2020). Stopped: 129 infections on vaccine vs 123 on placebo.

28. RV144 (Thailand). Marginal efficacy ≈ 31%; no effect on viral load (RV152).

29. HVTN 705 / HPX2008 (mosaic Ad26 + gp140). Not significant.

30. Sanders RW, Moore JP et al. BG505 SOSIP in rabbit and monkey: autologous antibody, limited breadth in humans.

31. Case no. 5035958-79.2026.4.02.5101, 20th Federal Court, Rio de Janeiro. File: 2011/12 outbreak; 34/431 (8%) in 2020–21; M. tuberculosis culture; euthanasia of contacts and a pregnant female; no BSL-3; former staff with interrupted treatment.

32. Vervenne RA et al. Pathogenesis of TB in NHP — preclinical window of spread.

33. Nunes M. O Globo, 26 June 2026. Cetas-RJ/Ibama: three capuchins dead of TB.

34. PETA / FOIA. Elisabeth Murray lab, NIMH/NIH: 73 monkeys; 85% alopecia; 74% headpost; 64% water deprivation.

35. New Yorker, 2025. Alpha Genesis / Morgan Island: dozens of escapes; NIH contracts > US$120 m since 2008.

36. PETA / USDA, 2025–2026. New Iberia Research Center: ~12,000 monkeys; 19 rhesus dead in a freeze (Jan. 2025).

37. SourceWatch / PETA. Covance, Vienna/VA, 2004–05: undercover footage.

38. Science / STAT, 2026. NIH withdraws approval of Caucaseco (Colombia) after Aotus seizure.

39. Gong et al. M. tuberculosis outbreak in 84 zoo rhesus in China; TST negative in all.

40. CONCEA. Validated alternative method: up to five years to become obligatory substitution.