What ARRIVE is and who maintains it
ARRIVE, Animal Research: Reporting of In Vivo Experiments, is a reporting guideline for studies that describe research involving live animals. ARRIVE 2.0, the current version, restructures the original 2010 guidelines into an Essential 10, the minimum items needed for a study to be reviewed and assessed at all, and a Recommended Set of further items that add important context, and was developed by an international working group (Percie du Sert et al., 2020).
The Essential 10 cover the study design, sample size, inclusion and exclusion criteria, the randomisation method, blinding, outcome measures, the statistical methods, the species and strain used, the experimental procedures, and the results including attrition. The Recommended Set adds items such as abstract content, housing and husbandry, and justification of the choice of animal model.
The two-tier structure is itself a response to how the original 2010 ARRIVE guidelines were used in practice: a single long list of items was often applied unevenly, with some items addressed in detail and others skipped. Separating out the ten items that are essential for any reader to judge the risk of bias in a study was meant to make the minimum expectation unambiguous, while still encouraging fuller reporting through the Recommended Set.
Why it exists
Preclinical animal studies have a long-recognised problem: many report promising effects that fail to replicate in later studies, and in some fields, fail to translate into effective treatments when tested in humans. Part of this problem is genuine biological difference between species, but part of it is a reporting and methodological problem: studies that do not randomise or blind assessment, use too few animals, or do not report which animals were excluded and why, tend to produce less reliable, often inflated, effect estimates. The same pattern, inadequate allocation concealment and blinding being associated with inflated effect estimates, has been documented in human trials as well (Moher et al., 1998), which is one reason the ARRIVE working group modelled parts of its structure on CONSORT rather than starting from nothing.
ARRIVE exists to make these methodological features visible, so that a reader, or a researcher planning to build on the work, can judge how much confidence the reported effect deserves before committing further animals, time and funding to extending it. This is an ethical as well as a scientific concern: using animals in a study that is then reported so incompletely that its findings cannot be trusted or built on is widely regarded as a poor use of those animals, independent of whether the study itself was conducted carefully.
Which studies it fits
ARRIVE applies to any study reporting research that involves live, non-human animals studied in vivo, across basic biology, preclinical pharmacology, toxicology, behavioural research and translational medicine. It applies regardless of species, from rodents and zebrafish to larger mammals, and regardless of whether the study is exploratory or confirmatory.
- A controlled experiment comparing a treated group of animals against a control group: ARRIVE's Essential 10 applies directly, including randomisation and blinding items.
- A study where the experimental unit is a cage, litter or tank rather than an individual animal: ARRIVE expects the sample size and statistical unit to be reported at that level, not as a simple animal count.
- An observational study of animals in their natural environment, with no researcher-controlled intervention: the core structure of ARRIVE still applies, though some items on allocation and blinding may be reported as not applicable, with a reason.
Where ARRIVE does not apply
ARRIVE is specifically for in vivo studies involving live animals, and does not apply to purely in vitro work, such as cell culture experiments, which have their own, separate reporting conventions. It also does not apply to studies in human participants; a human trial, even one testing a treatment first developed in animal models, should be reported with CONSORT, not ARRIVE (Hopewell et al., 2025).
ARRIVE does not cover purely observational studies of risk factors in animal populations in the sense STROBE covers human observational epidemiology (von Elm et al., 2007); where an animal study is genuinely observational rather than experimental, researchers should still use ARRIVE's structure but adapt which items on allocation and blinding are applicable, stating clearly which do not apply and why.
Finally, ARRIVE is a reporting guideline, not an ethical approval framework. Following it does not by itself satisfy a research ethics or animal welfare committee's requirements, which usually also cover justification of the choice to use animals at all, the three Rs of replacement, reduction and refinement, and local regulatory approval; these sit alongside ARRIVE rather than being replaced by it.
Why it matters
The translational value of preclinical research depends on being able to judge how reliable a given animal study's effect estimate is before relying on it to justify further research, including, eventually, research in humans. Reporting that lets readers see the sample size at the correct experimental unit, whether allocation was randomised and outcome assessment blinded, and which animals were excluded and why, is what allows this judgement to be made at all. Evidence synthesis frameworks such as GRADE explicitly weigh risk of bias, including these same features, when rating the overall certainty that can be placed in a body of evidence, preclinical or clinical (Guyatt et al., 2008).
How it is applied in practice
Many journals publishing animal research now require authors to complete an ARRIVE checklist at submission, and funders in several countries ask grant applicants to show that proposed animal studies will be reported in line with ARRIVE as a condition of funding, reflecting a wider policy aim of reducing the number of animals used in poorly designed or poorly reported studies. Many ethics committees and animal welfare bodies also use ARRIVE items, particularly sample size justification and the experimental unit, as part of their review of a proposed protocol before it is approved.
A simple habit that satisfies several ARRIVE items at once is to decide, before the experiment starts, what will count as an a priori exclusion criterion (such as an animal falling ill for reasons unrelated to the intervention) and to record every exclusion against that rule as it happens, rather than deciding afterwards which animals to leave out of the analysis.
Sharing the underlying data, and where relevant the analysis code, from an animal study follows the same FAIR principles that apply to human research data, and allows other researchers to re-analyse or include the study in a later systematic review of the preclinical literature (Wilkinson et al., 2016).
There is also a research-waste argument for ARRIVE that is specific to animal work: if a study is too poorly reported for other researchers to judge its reliability, a reasonable response is to repeat it rather than build on it, which uses more animals for no scientific gain. Several funders and institutions frame ARRIVE compliance partly in these terms, as a way of reducing unnecessary duplication of animal experiments rather than only as a transparency requirement.
An illustrative example
The passage below is a hypothetical illustration, not a quotation from a published study.
| Element | Without ARRIVE | With ARRIVE |
|---|---|---|
| Sample size | "Forty mice were used." | "Forty mice (20 per group) were used, housed five per cage; the cage was the experimental unit for the primary outcome, giving four cages per group, with the sample size based on a power calculation detailed in the methods." |
| Randomisation | "Mice were assigned to treatment or control." | "Mice were allocated to groups using a computer-generated random sequence, stratified by baseline weight, with allocation concealed from the researcher administering treatment." |
| Blinding | Not mentioned | "Outcome assessment was performed by an investigator blinded to group allocation; tissue samples were coded before histological scoring." |
| Exclusions | Not mentioned | "Two animals were excluded after unexpected illness unrelated to the intervention, identified before unblinding, and are reported with their reasons in the results." |
| Housing | Not mentioned | "Animals were housed under a 12-hour light/dark cycle at 22 degrees Celsius with free access to food and water, in accordance with the institutional animal welfare protocol." |
What to check before you submit
- Have you correctly identified the experimental unit, and reported sample size at that level?
- Is the method of randomisation and allocation concealment described?
- Was outcome assessment blinded, and is this stated explicitly?
- Are all animals entered into the study accounted for, including any excluded, with reasons?
- Is the statistical method for each outcome stated, matching how the data were actually structured (e.g. accounting for clustering within cages or litters)?
- Are species, strain, sex and housing conditions reported in enough detail for another lab to repeat the study?
- Official checklist and source: ARRIVE Guidelines (arriveguidelines.org) .
Key takeaways
- ARRIVE 2.0 sets an Essential 10 and a Recommended Set of items for reporting in vivo animal research (Percie du Sert et al., 2020).
- It exists because unrandomised allocation, unblinded assessment and misreported sample sizes are linked to less reliable preclinical findings.
- It applies to in vivo animal studies; human trials use CONSORT (Hopewell et al., 2025) and human observational studies use STROBE (von Elm et al., 2007).
- Correctly identifying the experimental unit, which is not always the individual animal, is one of the most commonly misreported items.
- How completely bias-related items are reported affects how much certainty can be placed in preclinical evidence, as in frameworks such as GRADE (Guyatt et al., 2008).
- ARRIVE's structure echoes lessons from human trial reporting, where inadequate blinding and allocation concealment are linked to inflated effect estimates (Moher et al., 1998).
