ARRIVE Guidelines: Reporting Animal Research

The ARRIVE guidelines set out what a publication describing animal research must contain, and the current version is ARRIVE 2.0, published in 2020. The name stands for Animal Research: Reporting of In Vivo Experiments. Its defining feature is prioritisation: 21 items split into the Essential 10, which are the minimum required for a reader to assess whether the findings are reliable, and an 11-item Recommended Set that supplies context.

That split is the substantive change from the 2010 original, which presented 20 items as a single flat list. The revision happened because compliance stayed low for a decade, and the diagnosis was that authors faced with 20 equally weighted items could not tell which ones actually determined whether their study could be believed or reused. ARRIVE 2.0 answers that directly.

Why preclinical reporting attracts particular scrutiny

Animal experiments occupy an awkward position. They are interventional, so the same threats to validity that clinical trials guard against apply in full, but they have historically been reported with far less of the accompanying machinery. Papers describing a treatment effect in mice frequently do not say how animals were allocated to groups, who was blinded, how many animals were excluded and why, or how the group size was arrived at.

The consequence is practical rather than theoretical. A preclinical finding that cannot be assessed cannot be reliably built on, and effort is spent either repeating work unnecessarily or pursuing effects that were artefacts of undisclosed choices. Since animal numbers are themselves an ethical cost, incomplete reporting has a reduction implication as well as a scientific one.

The Essential 10

These are the items ARRIVE 2.0 treats as non-negotiable. Described in our own words, they cover:

1. Study design. For each experiment, the groups being compared, including controls, and the experimental unit, meaning whether an individual animal, a cage, a litter or a tank is the thing being counted.
2. Sample size. The exact number of experimental units in each group, the total number used, and how the sample size was decided, including any a priori power calculation.
3. Inclusion and exclusion criteria. The criteria set in advance for including and excluding animals, experimental units or data points, any that were excluded and why, and confirmation that these were applied consistently.
4. Randomisation. Whether and how animals were allocated to groups, the method used to generate the sequence, and any steps taken to minimise confounding such as the order of treatment and measurement.
5. Blinding or masking. Who was aware of the group allocation at each stage: during allocation, the conduct of the experiment, outcome assessment, and data analysis.
6. Outcome measures. The primary and any secondary outcome measures, defined precisely, and for the primary outcome how and when it was assessed.
7. Statistical methods. Full detail of the methods used for each analysis, and whether the assumptions of the approach were assessed.
8. Experimental animals. Species, strain and substrain, sex, age or developmental stage, weight, and the source of the animals.
9. Experimental procedures. What was done to each group in enough detail for the work to be replicated, including what, how, when, where and why.
10. Results. For each experiment, a summary for each group with a measure of variability such as the standard deviation, and the effect size with a confidence interval where relevant.

The authoritative wording sits with the ARRIVE team and on the EQUATOR Network, and the official checklist should be downloaded from there rather than reconstructed from a summary.

The Recommended Set

The remaining 11 items add the context that makes a study interpretable and reusable: the abstract, the scientific background, the objectives, the ethical statement including the approving body and licence details, housing and husbandry, animal care and monitoring including welfare assessment and humane endpoints, the interpretation of results in the context of current evidence, generalisability to other species or to human biology, protocol registration, data access, and declarations of interest.

Calling these recommended does not make them optional in practice. Many journals require the ethical statement and data availability regardless, and reviewers routinely ask about housing and husbandry because environmental conditions are a recognised source of between-laboratory variation.

The item that causes the most statistical trouble

Item 1's experimental unit requirement deserves separate attention, because getting it wrong invalidates the analysis rather than merely obscuring it. If animals share a cage and the treatment is delivered through drinking water or diet, the cage is the unit that received the intervention, not the animal. Analysing individual animals in that situation treats correlated observations as independent, inflates the effective sample size, and produces confidence intervals that are too narrow and p-values that are too small.

The same applies to litters in developmental studies and tanks in aquatic work. The correct handling is either to analyse at the level of the unit that was randomised, or to model the clustering explicitly with a mixed effects approach. Stating the experimental unit plainly, as item 1 requires, is what makes this checkable, which is precisely why it sits in the Essential 10.

ARRIVE, PREPARE, and reporting versus appraisal

Two distinctions are worth keeping clear. PREPARE covers planning an animal study and ARRIVE covers reporting it, so the decisions PREPARE prompts are the ones ARRIVE later asks you to disclose. Using them in sequence is the same relationship that SPIRIT and CONSORT have in clinical trials, and that PRISMA-P and PRISMA 2020 have in evidence synthesis.

The second distinction is the familiar one. ARRIVE is a reporting guideline: it asks whether a study is adequately described. Whether it was well conducted is a separate judgement made with an appraisal instrument, and in preclinical systematic reviews that is typically a risk of bias tool adapted for animal studies rather than ARRIVE itself. Scoring ARRIVE items and presenting the total as a quality score is a methodological error, the same one that recurs with STROBE and COREQ.

Completing the checklist

Fill it in against the final manuscript, recording the page and section for each item rather than ticking it, and write not applicable with a reason where an item genuinely does not fit your design. Submit it as a supplementary file and name the version, ARRIVE 2.0, in the methods.

Read as a self-review, the checklist is unusually informative in preclinical work because the items most often unanswerable are the same four every time: how the sample size was decided, how animals were allocated, who was blinded at each stage, and how many animals were excluded and why. Those four are also exactly what a methodological reviewer will ask for, so an honest pass through the Essential 10 tells you the content of your first revision before it arrives.