Animal-study reports
The report of a preclinical animal experiment has to give a reader enough to judge the model, the conduct of the experiment and how far its result translates, and many published reports still leave out the animal, housing and test details that decide this.
Why animal studies have their own reporting guidelines
Rodents are commonly used as disease models, so they are inherently expected to be at least somewhat predictive of a human response to a drug. For that translational value to be real, animal studies must be designed, conducted, analysed and reported with the highest scientific rigour. They still lag behind the standards for reporting human clinical data, although both share the aim of generating unbiased data.
Reporting standards for human randomised controlled trials improved with the CONSORT (Consolidated Standards of Reporting Trials) statement of 1996, and randomisation and blinding are now regarded as the minimum requirements for performing and communicating science. Not all preclinical, or clinical, publications in peer-reviewed journals seem to meet these basic reporting standards. A 2009 survey by the UK National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs), covering publicly funded animal research in the United Kingdom and the United States, found that over 85% of the included publications did not report randomisation or blinding, and 41% lacked key information on the hypothesis and on the number and characteristics of the animals.
In response to those results, NC3Rs developed and funded the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines in 2010, following the example of CONSORT and aiming to improve the design of new experiments as well as the reporting of existing ones. ARRIVE is intended to improve the reporting of bioscience research in which laboratory animals are used, and more than a dozen journals had adopted it in 2010. Its 20-item checklist reminds authors to say how the study was designed, conducted and analysed, for example by giving the animal characteristics and the statistical approach. Its aim is to maximise the information published and minimise unnecessary studies. SYRCLE (the Systematic Review Centre for Laboratory Animal Experimentation) published a Gold Standard Publication Checklist that partly overlaps with ARRIVE. The EQUATOR Network provides more specific guidance for various types of preclinical research publication, and the Good Laboratory Practice regulations include study reporting. Despite initiatives such as ARRIVE and the Gold Standard Publication Checklist to standardise the reporting of animal studies in peer-reviewed journals, many publications still lack key methodological details, and animal studies are therefore often criticised for poor scientific quality and low translatability to the clinic.
What an incomplete report costs
By including or omitting methodological details, authors greatly influence the quality of the article and the ability to draw meaningful conclusions from its results. The value and clinical meaningfulness of an animal study hinge on thorough reporting of the experimental methods, and even seemingly small details may alter the reproducibility and generalisability of the outcomes. Publications that fail to provide even basic details on experimental design and analysis are rated as low-quality reports and excluded from systematic reviews and meta-analyses. Those reviews summarise primary research, represent the highest level of medical evidence and are used to guide clinical decision-making. An incomplete report is therefore a source of bias even where the study itself was conducted perfectly, and it may lead to faulty conclusions about a drug's safety and efficacy. A reader who meets an omission is left to assume the worst case, that the measure was neither considered nor performed, so the relevance and quality of the data may be judged poor whether or not they are. Unless research is adequately reported, the time, effort and resources invested in it are wasted, and in preclinical work wasted resources may mean animal lives lost for nothing.
What an animal-study report contains
The ARRIVE guidelines recommend, item by item, what an animal-study report gives. In the title, describe the content of the article as accurately and concisely as possible. In the abstract, accurately summarise the background and objectives, the species or strain used, the key methods, the principal findings and the conclusions. In the introduction, give enough scientific background, with references to previous work, to understand the motivation and context of the study, explain the experimental approach and rationale, and explain how and why the species and model can address the scientific objectives. State the primary and any secondary objectives or the hypotheses tested.
For each experiment, give the number of experimental and control groups, any steps taken to minimise subjective bias when allocating animals to treatment, such as randomisation, and when assessing results, such as who was blinded and when if blinding was done, and the experimental unit, which may be a single animal, a group or a cage. Give full details of how animals were allocated to the experimental groups, including randomisation or matching if done. Define the primary and secondary experimental outcomes. Give the total number of animals in each experiment and the number in each group, and explain how that number was arrived at, including any sample-size calculation. A timeline diagram or flowchart can be useful to illustrate how a complex design was carried out.
Report the species, strain, sex, developmental stage, such as the mean or median age with the age range, weight, as the mean or median with the range, source, international strain nomenclature, genetic modification status, such as knockout or transgenic, genotype, health or immune status, whether the animals are drug or test naïve, and any previous procedures. For each experimental group, including controls, give precise details of every procedure: the formulation, dose, site and route of administration, anaesthesia and analgesia with their monitoring, surgical procedure, method of euthanasia and any specialist equipment with its supplier, the time of day of each, and where it was done, such as the home cage, the laboratory or a water maze. Indicate the nature of the ethical review process, the relevant licences, and the national or institutional guidelines for the care and use of animals that cover the research.
Describe the housing, such as the type of facility, cage or housing, bedding material, number of cage companions and, for fish, tank shape and material, and the husbandry, such as the breeding programme, light/dark cycle, temperature, water quality for fish, type of food, access to food and water, and environmental enrichment. Report the welfare-related assessments and interventions carried out before, during or after the experiment. Give the statistical methods for each analysis and any methods used to assess whether the data met their assumptions. Report the characteristics and health status of each group before treatment or testing, the number of animals in each group included in each analysis as absolute numbers rather than percentages, any animals or data excluded and why, the results of each analysis with a measure of precision, and all important adverse events in each group.
Interpret the results against the objectives and hypotheses, current theory and other relevant studies, and comment on the limitations, including potential sources of bias, the limitations of the animal model and the imprecision of the results. Comment on whether and how the findings are likely to translate to other species or systems, including any relevance to human biology. List every funding source, with its grant number, and the funder's role.
As a general rule, alongside the guidelines specific to its field, a research report should include the protocol or outline of the study design, stating the specific aims and how they will be achieved; the raw data collected, as appropriate, and the analysable data; the data sharing plan; the statistical analysis methods; and an overall study report summarising the key findings and next steps. Analysable data is raw data put into a form that can be analysed statistically, and it is usually the most useful form in a study report and the summary data most readers refer to. The data sharing plan says how the researchers intend to distribute their findings and at what point in the research. It is important for readers to know how the data were processed and tested, so that the results can be replicated. Different publishers and publications have specific requirements of their own.
Animal, housing and test details most often left out
The methodological sections specific to animals are often under- or misreported.
Species, strain, substrain, vendor and vendor location. Most authors report species and strain, because both have long been known to affect behavioural and pharmacological responses. Less widely known is that substrains, and even the same strain obtained from different vendors, may exhibit distinct phenotypes. Wistar rats from two suppliers differ in their response to the same acute myocardial infarction model, those from one having a higher survival rate despite being more sensitive to cardiomyocyte damage. The vendor's location matters too: Wistar rats from one supplier's American and European operations are behaviourally distinct, and although both are Wistar rats they originated from different institutes, which shows the need to document strain, substrain, vendor and vendor location. The popular "Black 6" mouse strain is widely and wrongly assumed to be called C57/BL6, and its original breeder, The Jackson Laboratory, stated that there is no such thing as a C57/BL6 mouse. There are two distinct breeding colonies, C57BL/6J ("J" for Jackson Laboratory) and C57BL/6N ("N" for the National Institutes of Health), from which many substrains have emerged, such as C57BL/6NCrl from Charles River and C57BL/6JJcl from Clea Japan. Genetic and phenotypic differences within and among populations are well established, so a report should always give the complete substrain designation, which indicates the laboratory maintaining the colony. Genetically modified animals, such as knockout or transgenic animals, need additional information, and published guidelines and a checklist for reports on mutant studies exist.
Sex and age, as actual numbers. Failure to mention sex and age may skew data interpretation, because both are known to affect pharmacokinetics and pharmacodynamics in laboratory animals as in humans. The share of articles reporting them increased markedly in the two decades to 2017, yet in one study only 50% of the included articles published in 2014 reported both the sex and the age of their mice. There is hardly an excuse for the omission, since the variables are available to every researcher and take up little space. Giving the weight or a developmental stage such as "juvenile" or "adult" instead of the age does not suffice, because these vary greatly across laboratories. In one survey, the age at which rodents are considered adult spanned 6 to 20 weeks for mice and 8 to 16 weeks for rats, ranges that encompass distinct developmental events. A report should give the actual age, as the mean or median, variation and range, together with the weight and the health or immune status of all the animals included.
Housing and enrichment. Laboratory housing conditions significantly influence behaviour and pharmacological response, and should be reported in detail. Mice and rats are very social animals, so isolating them may induce stress and interfere with a drug's effects, which makes the distinction between single- and group-housed animals important. They are also sensitive to environmental complexity. In the 1940s the psychologist and neuroscientist Donald Hebb took some laboratory rats home for his children to play with, and those animals later performed better on cognitive tests than animals housed only in the laboratory. In the decades that followed, animals in enriched cages, containing sensory stimulation such as nesting material, tunnels, cardboard boxes or chewing toys, were found to have more synapses, more and longer dendrites, and protection against several types of brain injury. Because enrichment changes the brain this much, its presence or absence should be reported. Authors should not refer to "standard cages" without further clarification, because the standard in one laboratory may be a cage with bedding only while in another it also includes shelter and toys. A report should specify the number and sex of animals per cage, the model and size of cage and lid, the presence and type of any enrichment, the frequency of cage change, the room temperature and lighting, and the nutrition type and feeding regimen. One author recommends SYRCLE's Gold Standard Publication Checklist as a comprehensive guide to describing animal housing.
The behavioural test, step by step. Animal behaviour is central to evaluating drug efficacy, yet publications frequently lack a detailed description of the tests performed. Comprehensive guidance on reporting them appears to be absent: ARRIVE devotes two lines to the subject, prompting the author to define the experimental outcome clearly, and the Gold Standard checklist treats the frequency of handling as a relevant variable but offers no complete list. Given the number and complexity of available tests, a concise checklist is no easy task, but some general considerations apply to most of them and should be reported in every behavioural publication, for transparency and unbiased interpretation of the data. The most obvious is a precise description of the test. The novel object recognition test, a popular rodent cognition test, consists of habituation, training and testing, and takes one to four days in total depending on how long the animal is habituated to the arena; reported habituation lengths across laboratories span two minutes to two hours. That spread is not in itself a major concern as long as the duration of each phase and the intervals between them are reported. It becomes one when authors deviate from the original protocol without mentioning their modifications, or worse, give only the name of the test. For any test with multiple phases, all durations and intervals should be given, along with any time allowed for acclimatisation to the behavioural room immediately before testing. The rule of thumb is to provide as many details as possible: where water is involved, as in the Morris water maze or the forced swim test, its temperature and how often it is changed, and for any apparatus, its size and material, since these can affect the response to a drug.
Time of day, experimenter and test order. People probably perform better on a cognition test during the day than in the middle of the night, and the same has been observed in nocturnal rodents, whose behaviour is significantly affected by whether they are tested in their active or inactive phase. Testing should therefore preferably be conducted in the dark, active phase, unless a particular measure is known not to be affected by circadian rhythm. The same applies to collecting any tissue sensitive to the photoperiod. Besides the light cycle in the animal room, a report should give the time of day at which behavioural data or tissue were collected and, where applicable, the light intensity in lux. It should also state whether all tests were conducted by the same experimenter and how often the animals were handled before testing. Where more than one test was used, it should give the number of days between tests and their order, particularly if one of them is stressful for the animals. A timeline of experimental events is a good way to show a complex design or a battery of tests graphically.
The intervention and the tissue. A drug treatment must be described clearly: the dose per weight, the volume of injection, the route of administration, the frequency including the time of day it was given, the exact vehicle, and the method of preparation, such as sonication or multiple dilutions. Euthanasia and tissue collection must be documented too: the method, the time of day, whether all animals were euthanised on the same day, whether randomisation was applied, and how the tissue was collected. A report should say which lobe the tissue came from, not merely that "liver tissue was obtained", and how it was stored during and after collection. Papers frequently lack basic information on tissue processing, such as the dissection method, centrifugation speed or freezer temperature. Where decapitation is used, the report should include the measures taken to protect the rodents from the smell of blood, which can cause confounding hormonal effects, such as housing them in an adjacent room and having an experimenter free of blood scent bring them into the decapitation room.
Adherence to the checklists
Adhering to current guidelines such as ARRIVE or the Gold Standard checklist, to reduce the risk of bias and maintain a high scientific standard, is the author's responsibility. Journals, however, need to implement adherence rather than endorse the checklists passively, for example by requesting a completed checklist at submission, and they need to give authors enough space for all relevant details, in the manuscript body or in supplementary files. New medical writers and researchers could benefit greatly from education and training that address these reporting issues in greater detail.