Keywords

1 Introduction

Data quality, reproducibility and reliability are a matter of concern in many scientific fields including biomedical research. Robust, reproducible data and scientific rigour form the foundation on which future studies are built and determine the pace of knowledge gain and the time needed to develop new and innovative drugs that provide benefit to patients (Freedman and Gibson 2015). In particular, research involving animals is essential for the progression of biomedical science, assuming that experiments are well designed, performed, analysed, interpreted as well as reported.

However, it has been described many times over the last few years that in preclinical research – particularly preclinical animal research – many findings presented in high-profile journals are not reliable and cannot be replicated (Begley and Ellis 2012; Peers et al. 2012; Prinz et al. 2011). This has led to the so-called reproducibility crisis which, according to some, may largely be due to the failure to adhere to good scientific and research practices and the neglect of rigorous and careful application of scientific methods (Begley and Ioannidis 2015; Collins and Tabak 2014). In this context, various reasons have been suggested to contribute to and perhaps explain the lack of reliability and reproducibility in preclinical research including inadequacies in the design, execution and statistical analysis of experiments as well as deficiencies in their reporting (Glasziou et al. 2014; Ioannidis et al. 2014; Jarvis and Williams 2016).

It has been reported that only a minority of animal studies described in the scientific literature use critical experimental design features such as randomisation and blinding despite these components being essential to the production of robust results with minimal risk of experimental bias (Hirst et al. 2014; Macleod et al. 2015). Furthermore, in a study by Bebarta et al., it was described that studies, which did not utilise randomisation and blinding, were more likely to display differences between control and treatment groups, leading to an overestimation of the magnitude of the treatment effects (Bebarta et al. 2003). Another kind of bias that may compromise the validity of preclinical research is reporting bias, consisting of publication bias as well as selective analysis and outcome reporting bias. In many cases, animal studies with negative, neutral or inconclusive results are not reported at all (publication bias), or only the analysis yielding the best statistically significant effect is selectively presented from a host of outcomes that were measured (selective analysis and outcome reporting bias) (Tsilidis et al. 2013). This under-representation of negative research findings can be misleading concerning the interpretation of presented data, often associated with falsely inflated efficacy estimates of an intervention (Korevaar et al. 2011). Furthermore, unnecessary repetitions of similar studies by investigators unaware of earlier efforts may result.

In 2005, Ioannidis stated that it can be proven that most published research findings are irreproducible or even false due to the incorrect and inadequate use of statistics for their quantification. Specifically, underlying factors such as flexible study designs, flexible statistical analyses and the conductance of small studies with low statistical power were described (Button et al. 2013; Ioannidis 2005). Along these lines, Marino expressed the view that poor understanding of statistical concepts is a main contributory factor to why so few research findings can be reproduced (Marino 2014). Thus, it is urgently required that best practices in statistical design and analysis are incorporated into the framework of the scientific purpose, thereby increasing confidence in research findings.

Additionally, transparent, clear and consistent reporting of research involving animals has become a further substantial issue. Systematic analysis has revealed that a significant proportion of publications reporting in vivo research lack information on study planning, study execution and/or statistical analysis (Avey et al. 2016; Kilkenny et al. 2009; Landis et al. 2012). This failure in reporting makes it difficult to identify potential drawbacks in the experimental design and/or data analysis of the underlying experiment, limiting the benefit and impact of the findings. Moreover, when many of these factors are intertwined, this can lead to negative consequences such as higher failure rates and poor translation between preclinical and clinical phases (Hooijmans and Ritskes-Hoitinga 2013).

Importantly, from an ethical perspective, laboratory animals should be used responsibly. In this context, it is of utmost importance to implement Russell and Burch’s 3Rs (reduction, refinement, replacement) principle in the planning and execution of animal studies (Carlsson et al. 2004; Tannenbaum and Bennett 2015; Wuerbel 2017) as well as more efficient study designs, improved research methods including experimental practice, animal husbandry and care. Also the availability of sufficient information and detailed descriptions of animal studies may help to improve animal welfare and to avoid unnecessary animal experiments and wasting animals on inconclusive research.

In the past decade, several guidelines and frameworks have been released in order to improve the scientific quality, transparency and reproducibility of animal experiments (Hooijmans et al. 2010; Kilkenny et al. 2010a; Nature 2013; NIH, Principles and Guidelines for Reporting Preclinical Research). The ARRIVE (Animal Research: Reporting In Vivo Experiments) guidelines focus on the clear and transparent reporting of the minimum information that all scientific publications reporting preclinical animal research should include such as study design, experimental procedures and specific characteristics of the animals used (Kilkenny et al. 2010b). Similarly, the Gold Standard Publication Checklist (GSPC) also aims at improving the planning, design and execution of animal experiments (Hooijmans et al. 2011). The ARRIVE guidelines were launched in 2010 by a team led by the UK National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs) and have steadily gained credence over the past years. Endorsed by more than 1,000 biomedical journals, the ARRIVE guidelines are now the most widely accepted key reporting recommendations for animal research (NC3Rs, ARRIVE: Animal Research: Reporting In Vivo Experiments). In addition, various leading scientific journals have begun to change their review practices and place greater emphasis on experimental details prompting authors to report all relevant information on how the study was designed, conducted and analysed (Curtis and Abernethy 2015; Curtis et al. 2015; McGrath and Lilley 2015; McNutt 2014a, b; Nature 2013). Such initiatives may help to ensure transparency and reproducibility of preclinical animal research, thereby improving its reliability and predictive value as well as maximising a successful translation into clinically-relevant applications. However, the compliance with these guidelines remains low several years later. An evaluation of papers published in Nature and PLOS journals in the 2 years before and after the ARRIVE guidelines were communicated suggests that there has been only little improvement in reporting standards and that authors, referees and editors generally are ignoring the guidelines (Baker et al. 2014). Quite recently, a similar analysis by Leung et al. has shown that the reporting quality in animal research continues to be low and that supporting the ARRIVE guidelines by several journals has not resulted in a considerable improvement of reporting standards (Leung et al. 2018). Obviously, despite the widespread endorsement of the guiding principles by multiple journals in various research areas, the impact of this endorsement on the quality of reporting standards of animal studies is only modest (Avey et al. 2016; Delgado-Ruiz et al. 2015; Liu et al. 2016; Schwarz et al. 2012; Ting et al. 2015). In part, this may be caused by the fact that the recommendations have limitations regarding feasibility and applicability across the diversity of scientific fields that comprise biomedical research making them impractical for some kind of studies. Moreover, researchers may not be convinced that it is necessary to apply effort in order to achieve maximum transparency and reproducibility of animal-based research. It is crucial to increase the awareness of the existence of animal research reporting guidelines as well as the importance of their implementation. A serious problem of guiding principles in general and the ARRIVE guidelines in particular is that most biomedical research journals endorse them but do not rigorously enforce them by urgently requiring comprehensive and detailed reporting of the performed research. A direct consequence of enforced compliance may be increased time and financial burdens making an balanced weighting between what is ideal and what is feasible and practical absolutely essential (Leung et al. 2018).

Nevertheless, the scientific community needs effective, practical and simple tools, maybe in the form of guidelines or checklists, to promote the quality of reporting preclinical animal research. Ideally, such guiding principles should be used as references earlier in the research process before performing the study, helping scientists to focus on key methodological and analytical principles and to avoid errors in the design, execution and analysis of the experiments.

A recent study by Han et al. showed that the mandatory application of a checklist improved the reporting of crucial methodological details, such as randomisation, blinding and sample size estimation, in preclinical in vivo animal studies (Han et al. 2017). Such positive examples support optimism that when reporting is distinctly required, important improvements will be achieved (Macleod 2017). Accordingly, the strict adherence to reporting guidelines will become useful to address the concerns about data reproducibility and reliability that are widely recognised in the scientific community.

In the present chapter, we discuss the minimum information that should be provided for an adequate description of in vivo experiments, in order to allow others to interpret, evaluate and eventually reproduce the study. The main part of the chapter will focus on the minimum information that is essential for the reporting in a scientific publication. In addition, a table will be presented distinguishing information necessary to be recorded in a laboratory notebook or another form of internal record versus information that should be reported in a paper. Examples of specific research areas such as behavioural experiments, anaesthesia and analgesia and their possible interference with experimental outcomes as well as ex vivo biochemical and histological analysis will be described.

2 General Aspects

Over the last decade, several guiding principles, such as the GSPC and the ARRIVE guidelines, have been developed in order to improve the quality of designing, conducting, analysing and particularly reporting preclinical animal research. These recommendations have in common that all major components of animal studies that can affect experimental outcomes, including conditions of animal housing, husbandry and care, have to be efficiently reported. In the following section, the most important aspects mentioned in these guidelines are summarised (Hooijmans et al. 2010; Kilkenny et al. 2010a). Finally, a table will be presented comparing information that is necessary to be recorded in a laboratory notebook or another form of internal protocols versus information that should be reported in a scientific publication (Table 1 ).

Table 1 Necessary information for including in a publication and recording in a laboratory notebook
Full size table

At the beginning of a preclinical animal study report, readers should be introduced to the research topic within the context of the scientific area as well as the motivation for performing the current study and the focus of the research question, specific aims and objectives. Primarily, it should be explained why the specific animal species and strain have been chosen and how this animal model can address the scientific hypotheses, particularly with regard to the clinical relevance of the project.

Any studies involving the use of laboratory animals must be formally approved by national regulatory authorities. Therefore, it is necessary to provide information indicating that the protocol used in the study has been ethically reviewed and approved. Additionally, any compliance to national or institutional guidelines and recommendations for the care and use of animals that cover the research should be stated (Jones-Bolin 2012).

In order to allow the replication of a reported study, a detailed description of the experimental animals has to be provided, including species, strain (exact genetic code/nomenclature), gender, age (at the beginning and the end of the experiment), weight (at the start of the experiment) and the origin and source of the animals. These biological variables are scientifically important since they often represent critical factors affecting health or disease of the animals and therefore may influence research outcomes (GV-SOLAS 1985; Oebrink and Rehbinder 2000). For the same reason, it is also essential to comment on the animals’ experience and to state if they are drug naïve or if they have received any previous procedures or treatments. Additionally, information about the health, microbiological and immune status of the animals can be of high relevance for study outcomes and the ability to replicate findings and therefore should be given (GV-SOLAS 1999). This means, e.g. to depict that the animals are kept under specific pathogen-free (SPF) conditions (accompanied by a list of pathogens excluded) and that their health and microbiological status is checked and monitored according to the FELASA recommendations (Nicklas et al. 2002). When using genetically modified animals, it is important to describe their genetic background, how these animals were generated and which control animals were selected.

There is increasing evidence that elements of the laboratory environment as well as housing and husbandry practices can significantly affect the animals’ biology and ultimately research outcomes (Hogan et al. 2018; Reardon 2016). This implicates an exact specification of the environmental conditions in which the animals were housed and where the experiments were conducted. The animal facility should be described concerning temperature, relative humidity, ventilation, lighting (light/dark cycle, light intensity) and noise (Baldwin et al. 2007; Speakman and Keijer 2012; Swoap et al. 2004; Van der Meer et al. 2004). In more detail, the specific housing conditions of the animals should be represented including type and size of the cages, bedding material, availability and type of environmental enrichment, number of animals per cage (and reasons for individual housing when applicable) as well as frequency of cage changes and handling procedures (Balcombe et al. 2004; Nicholson et al. 2009; Perez et al. 1997; Rock et al. 1997; van Praag et al. 2000). In addition, the reporting of nutrition and water regimes needs to be specified regarding the type (composition, special diets, purification) as well as access to food and water (ad libitum, restricted, amount of food/water and frequency and time of feeding or water supply).

When describing the procedures carried out in animal studies, several aspects require thorough consideration and need to be presented for each experiment and each experimental group, including controls. When has the experiment been performed (day and time of intervention and time interval between intervention and data sampling or processing)? Where has the experiment been performed (home cage, laboratory, special device/equipment for investigation)? What kind of intervention has been carried out? Here, details about the methodological techniques such as surgical procedures or sampling methods (including specialist equipment and suppliers) should be provided. Importantly, drugs and compounds used in the experiments need to be specified concerning name, manufacturer and concentration as well as the formulation protocol, dosage, application volume, frequency and route of administration. Additionally, when anaesthetics and analgesics are required for animal welfare reasons, it is crucial to include information about the name of these agents, administered doses, route and frequency of application as well as monitoring procedures of the animals’ physiological signs that are used to guarantee a sufficient level of anaesthesia and analgesia (Flecknell 2018; Gaertner et al. 2008). Similarly, the method of euthanasia applied at the end of the study should be described (Sivula and Suckow 2018).

To ensure the quality and validity of preclinical animal research, it is crucial to indicate if the performed study is a confirmatory or hypothesis-testing one and to implement appropriate experimental study designs (Johnson and Besselsen 2002). This comprises a clear definition of the experimental unit (individual animal or group of animals in one cage) as well as the number of treatment and control (positive, negative, vehicle) groups. In this context, the reporting of animal numbers (total number per experiment as well as per experimental group) is essential to assess biological and statistical significance of the results and to re-analyse the data. Additionally, any power and sample size calculations used for the determination of adequate animal numbers that allow the generation of statistically meaningful results should be reported (Button et al. 2013). Moreover, any actions undertaken to minimise the effects of subjective bias when allocating animals to experimental groups (e.g. randomisation) and when assessing results (e.g. blinding) should be stated (Bello et al. 2014; Hirst et al. 2014; Moser 2019). Randomisation is the best method to achieve balance between treatment and control groups, whereas blinded assessment of outcomes (assessing, measuring or quantifying) improves qualitative scoring of subjective experimental observations and promotes comparable handling of data. Both strategies enhance the rigour of the experimental procedure and the scientific robustness of the results.

When reporting the results of the experiments, statistics needs to be fully described including the statistical method/test used to analyse the primary and secondary outcomes of the study (Marino 2014). The exact number of analysed animals and a measure of precision (mean, median, standard deviation, standard error of the mean, confidence interval) should be presented. This is of high relevance for interpreting the results and for evaluating the reliability of the findings. Importantly, the number of excluded animals as well as reasons and criteria to exclude them from the experiment, and hence analysis, should be well documented. Furthermore, the description of outcomes should comprise the full spectrum of positive and negative results as well as whether there were attempts to repeat or confirm the data. Equally, all relevant adverse events and any modifications that were made to the experimental protocol in order to reduce these unwanted effects should be reported.

Finally, when discussing and interpreting the findings, it is important to take into account the objectives and hypotheses of the study as predetermined in the experimental study design. Additionally, a comment on the overall scientific relevance of the outcomes as well as their potential to translate into clinical significance should be included. In order to demonstrate how animal welfare issues have been addressed in the current study, any implications of the experimental methods or results for the replacement, refinement or reduction of the use of laboratory animals in research need to be described (Taylor 2010).

In conclusion, the meaningful and accurate reporting of preclinical animal studies encompasses a plethora of aspects, ranging from a detailed description of the experimental animal to a complete documentation of the statistical analysis. Creating transparency in this way can help to evaluate studies in terms of their planning, methodology, statistical verification and reproducibility. It is highly recommended to make all raw data, analyses and protocols available to the whole research community in order to provide insight into the full workflow of the scientific project.

3 Behavioural Experiments

Behavioural animal studies are of great importance to increase the scientific knowledge about the complex processes underlying animal behaviour in general as well as to investigate potential drug effects on behavioural outcomes. Furthermore, translational research aims to identify disease-relevant endpoints in behavioural animal studies that are robust, reliable and reproducible and ultimately can be used to assess the potential of novel therapeutic agents to treat human diseases (Sukoff Rizzo and Silverman 2016).

However, performing behavioural experiments in animals is largely challenging for scientists since studies of this nature are extremely sensitive to external and environmental factors (Crabbe et al. 1999). Specific housing conditions, e.g. the lack of environmental stimulation, can interfere with brain development and selectively alter brain functions, thereby affecting the expression of certain behaviour (Wuerbel 2001). Resulting stereotypies and other abnormal repetitive behaviours can be severely confounding in behavioural experiments and have an impact on the validity, reliability and reproducibility of scientific outcomes (Garner 2005).

Additionally, when measuring behaviour in animals, there are multiple other factors that may influence the generation of a behavioural response which can be classified as ‘trait’, ‘state’ and ‘technical’ factors (Sousa et al. 2006). ‘Trait’ factors include genetic (e.g. genetic background, gender) as well as developmental characteristics (e.g. stress experience, handling, housing conditions, social hierarchy) of the animals. ‘State’ factors comprise the time of the experiment, the experience and training status of the investigator, characteristics of the animal (e.g. age, health status, pharmacological treatment) as well as features of the experimental setup (e.g. construction, illumination, test environment, cleansing). ‘Technical’ factors encompass data acquisition (e.g. automated vs. manual observation, calibration, choice of behavioural parameters) as well as data analysis (e.g. distribution, normalisation of data).

In preclinical research settings, it is difficult to standardise all such factors, which may contribute to the poor reproducibility of behavioural observations in animals across different laboratories (Wahlsten 2001). Standardisation is assumed to minimise the variability of results and to increase sensitivity and precision of the experimental procedure. However, contrary to the assumption that rigorous standardisation of animal experiments may help to ensure their reproducibility, it has been proposed that rather, systematic variation of experimental conditions (heterogenisation) can lead to the generation of robust and generalisable results across behavioural animal studies since the external validity is enhanced, thereby improving reproducibility (Richter et al. 2010; Voelkl et al. 2018; Wuerbel 2000). Nevertheless, considering that a strict and universal standardisation of laboratory environmental and experimental conditions is exceptionally unlikely, it is of major importance to take into account any possible determinants that might exert an effect on animals’ performance when designing, conducting and analysing behavioural experiments and to report these factors accurately and transparently.

As mentioned above, there is increasing evidence that the laboratory environment and distinct husbandry and housing conditions may influence animal welfare and hence behaviour. Moreover, test outcomes of behavioural animal studies are highly dependent on small but important details regarding these conditions that are usually poorly reported. One such example is light conditions: light is a fundamental environmental factor regulating animal activity and physiology, and it has been found in rats that intense light conditions can lead to retinal damage, suppression of social play behaviour and locomotion as well as dissociation of circadian rhythms (Castelhano-Carlos and Baumans 2009). Similarly, environmental sounds that are inevitably present in animal research facilities also exert considerable effects on animals’ physiology and behaviour influencing sleeping patterns, locomotor activity, learning and anxiety reactions. Provision of a stable and controlled light and noise environment for the animals will contribute to their wellbeing and to the reproducibility of experimental outcomes, making a clear reporting of light and noise conditions obligatory.

Standard husbandry practices such as regularly performed cage-changing as well as commonly-used experimental procedures such as injections can significantly affect behavioural parameters in rodents, as measured by increased arousal behaviour and locomotor activity (Duke et al. 2001; Gerdin et al. 2012). These stress-related responses may have a considerable influence on the validity and quality of experimental outcomes and should be considered by researchers when designing study protocols and comparing data. Similarly, it has been shown that a change in housing conditions, including a combination of standard vs. individually ventilated cages and single vs. social housing, has a major impact on several physiological parameters and behavioural features of mice such as body weight, locomotor activity and anxiety-related behaviour (Pasquarelli et al. 2017). Thus, it is mandatory to clearly state as well as maintain a well-defined housing protocol during the experiment in order to ensure better comparison, reliability and reproducibility of experimental results across research facilities.

Environmental cage enrichment, which should be transparently reported when describing animals’ housing conditions, is strongly recommended by various guidelines regulating laboratory animal care and accommodation, as it is reported to enhance animal welfare, to protect against the development of stereotypies, to reduce anxiety and to positively influence brain development as well as learning and memory behaviour (Simpson and Kelly 2011). And indeed, it has been shown in rats and mice that environmental enrichment does not result in enhanced individual data variability nor generate inconsistent data in replicate studies between multiple laboratories, indicating that housing conditions can be improved without impacting the quality or reproducibility of behavioural results (Baumans et al. 2010; Wolfer et al. 2004).

Much evidence concerning the reproducibility of behavioural animal studies comes from the area of rodent phenotyping (Kafkafi et al. 2018). Some behavioural phenotypes, such as locomotor activity, can be highly reproducible across several laboratories, suggesting high stability and therefore better reproducibility (Wahlsten et al. 2006). In contrast, other behavioural phenotypes, such as anxiety-like behaviour, are more problematic to measure since they show increased susceptibility to a multitude of environmental factors that can affect the animals’ performance. Indeed, it has been reported that animal handling procedures, particularly the specific handling method itself, can elicit profound effects on animals’ anxiety levels and stress responses, indicating that the use of handling methods that will not induce strong anxiety responses will minimise confounding effects during experiments (Hurst and West 2010).

One of the most commonly used methods to investigate anxiety behaviour in rodents is the elevated plus maze (EPM) test (Lister 1987; Pellow et al. 1985). Besides strain, gender and age differences, it has been shown that the manipulation of the animals prior to the experiment (e.g. exposure to stressors, housing, handling procedures) and the averseness of the test conditions themselves (e.g. increased light levels) as well as repeated testing in the EPM can strongly influence the manifestation of anxiety behaviour (Bessa et al. 2005; File 2001; Hogg 1996). These crucial factors should not be excluded from experimental descriptions when reporting. Additionally, illumination of the EPM is a critical aspect that needs to be clearly specified. In fact, Pereira et al. concluded that it is not the absolute level of luminosity upon the arms, but the relative luminosity between the open and closed arms that predicts the behavioural performance of rats in the maze (Pereira et al. 2005).

Overall, it has been suggested that animal behaviour that is more closely linked to sensory input and motor output will probably be less affected by minimal modifications within the laboratory environment, whereas behaviour that is associated with emotional and social processes will be more sensitive (Wahlsten et al. 2006).

4 Anaesthesia and Analgesia

For numerous animal experiments such as surgeries or imaging studies, the use of anaesthetics and analgesics in order to reduce animal suffering from pain and distress is an ethical obligation and crucial to the 3Rs concept (Carbone 2011). However, it is known that these drugs (as well as untreated pain itself) can severely affect the animals’ biology and physiology, thereby influencing experimental data and introducing variability into research outcomes. Focusing on animal pain management means both an issue of generating high-quality, reproducible data and a substantial animal welfare concern. Dealing with this ethical and methodological conflict can pose a challenging task for scientists.

The ARRIVE guidelines recommend the reporting of anaesthesia and analgesia in order to achieve a full and detailed description of the experimental procedures performed in preclinical animal studies and to allow the critical evaluation and reproduction of published data. However, there is evidence that the current scientific literature lacks important details concerning the use of animal anaesthetics and analgesics, underestimating their potential interference with experimental results (Carbone and Austin 2016; Uhlig et al. 2015). In many cases, it is not clear whether scientists actively withhold treatment of animals with anaesthetic or analgesic drugs or just fail to include this information in the reporting, perhaps due to assumed insignificance to the experimental outcome. This creates the false impression that the selection of appropriate anaesthetic and analgesic regimens is not considered as a crucial methodological concern for generating high-quality research data. Furthermore, under-reporting of anaesthesia and pain management may also shape ongoing practice among researchers and encourage under-treatment of animals, which represents a serious problem concerning animal welfare.

Surgical pain and insufficient analgesia act as stressors and can elicit various effects on the animals’ immune system, food and water consumption, social behaviour, locomotor activity as well as metabolic and hormone state, among others, which may all influence the experimental outcomes of animal studies (Leach et al. 2012; Liles and Flecknell 1993). The use of anaesthetics and analgesics relieves surgical pain, thus contributing to the refinement of the experimental methods. Additionally, following the surgical procedure, an appropriate long-term pain management, which could last for several days, is required to ensure animal wellbeing. However, anaesthetic and analgesic drugs themselves may also confound experimental results, e.g. by regulating inflammatory pathways or exerting immunomodulatory effects (Al-Hashimi et al. 2013; Fuentes et al. 2006; Galley et al. 2000; Martucci et al. 2004). In cancer studies on tumour metastasis in rats, it has been shown that analgesic drugs such as tramadol are able to prevent the effect of experimental surgery on natural killer cell activity and on the enhancement of metastatic diffusion, which needs to be taken into account when using this kind of animal model (Gaspani et al. 2002). Furthermore, as demonstrated for inhalation anaesthesia using sevoflurane in rats, the expression of circadian genes may be severely influenced, which needs to be borne in mind in the design of animal studies analysing gene expression (Kobayashi et al. 2007).

As indicated in these few examples, the selection of appropriate anaesthetic and analgesic procedures is a key factor in preclinical animal studies and has to be carefully considered in the context of the specific research question and study protocol (Gargiulo et al. 2012). Scientists need to know which particular anaesthetic and analgesic drugs were used, including name, dose, application frequency and route of administration. Importantly, concerning long-term pain management after surgery, it is recommended to specify the duration of the analgesic treatment. Moreover, when it is decided to withhold analgesics because of interference with the research project, it is essential to include the reasons for this decision when reporting the study so that this information is available to those who may subsequently wish to replicate and extend such studies (Stokes et al. 2009).

Hypothermia, hypotension, hypoxemia and respiratory depression are frequently observed side effects during animal anaesthesia that can develop to serious health problems culminating in unexpected death (Davis 2008). These risks need to be incorporated when planning and performing experiments and highlight the importance of adequate animal monitoring procedures to eliminate the incidence of complications during anaesthesia. Additionally, the reporting of such events and their practical management (e.g. the use of warming pads) is crucial for scientists trying to reproduce and evaluate research data.

Animal imaging studies have specific requirements concerning anaesthesia that are related to the use of particular methodological techniques and the duration of the experiments. The primary reason for general anaesthesia in imaging studies is the need for the restraint and immobility of the animals in order to avoid movement artefacts and to obtain signals with maximal reproducibility (Gargiulo et al. 2012). However, anaesthetic agents can unintentionally affect physiological parameters of animals and confound the outcomes of different imaging modalities (Hildebrandt et al. 2008). As shown for positron emission tomography (PET) neuroimaging studies, the use of anaesthetics such as ketamine or isoflurane may alter neuromolecular mechanisms in animal brains, thereby leading to an incorrect representation of normal properties of the awake brain (Alstrup and Smith 2013). Moreover, repeated anaesthesia procedures and the preparation of the animals for the study may influence the processes under investigation. Physical restraint stress before the experiment can increase the anaesthetic induction doses and negatively influence the quality of some molecular imaging procedures such as PET due to altered kinetics and biodistribution of radiotracers (Hildebrandt et al. 2008). The latter effect has also been observed to be dependent on the choice of anaesthetics, the duration of fasting periods as well as to result from hypothermia observed as an adverse event from anaesthesia (Fueger et al. 2006).

As for surgical procedures, the careful selection of the most appropriate anaesthesia method addressing all the needs and goals of the specific research project and imaging modality is important (Vesce et al. 2017). Since anaesthetics can influence various physiological and pharmacological functions of the animals, monitoring of anaesthetic levels and of vital functions during imaging studies has proven useful. In order to achieve reproducible experimental conditions in imaging studies, a clear and consistent reporting of methodological details concerning the animals, fasting conditions, anaesthesia regimens and monitoring is absolutely essential.

5 Ex Vivo Biochemical and Histological Analysis

Numerous ex vivo methods, including biochemical and histological analyses, are used routinely to complement in vivo studies to add additional information or to address scientific questions which are difficult to address in an in vivo setting. The starting point for such studies is a living organism, and as such, many of the previously described considerations in, e.g. the ARRIVE guidelines are entirely applicable and should be included when reporting data from such studies. In the following section, we will highlight examples of studies where specific methodological details have been evinced to be important for outcome and as such should be included in any reporting of data from studies where similar ex vivo analyses have been carried out.

6 Histology

Histology is the microscopic study of animal and plant cells and tissues. It comprises a multistage process of cell or tissue collection and processing, sectioning, staining and examining under a microscope to finally quantification. Various methods are routinely applied in numerous cell and tissue types. The field of histology has been as affected as others by the lack of reproducibility of data across labs. In a recent report, Dukkipati et al. made the observation that conflicting data on the presence of pathological changes in cholinergic synaptic inputs (C-boutons) exists in the field of amyotrophic lateral sclerosis (ALS), thus making it difficult to assess roles of these synaptic inputs in the pathophysiology of the disease (Dukkipati et al. 2017). The authors sought to determine whether or not the reported changes described in the scientific literature are indeed truly statistically and biologically significant and to evaluate the possible reasons for why reproducibility has proven problematic. Thus, histological analyses were conducted using several variations on experimental design and data analysis and indeed, it was shown that factors including the grouping unit, sampling strategy and lack of blinding could all be contributors to the failure in replication of results. Furthermore, the lack of power analysis and effect size made the assessment of biological significance difficult. Experimental design has also been the focus of a report by Torlakovic et al. who have highlighted the importance of inclusion of appropriate and standardised controls in immunohistochemistry studies so that data can be reproduced from one test to another and indeed from one lab to another (Torlakovic et al. 2015). Lai et al. point to the difficulty in standardising complex methods in their report of the development of the OPTIClear method using fresh and archived human brain tissue (Lai et al. 2018).

A comparison of different quantification methods has been described by Wang et al. to determine hippocampal damage after cerebral ischemia (Wang et al. 2015). The authors start with the comment that multiple techniques are used to evaluate histological damage following ischemic insult although the sensitivity and reproducibility of these techniques is poorly characterised. Nonetheless, their output has a pivotal impact on results and conclusions drawn therefrom. In this study, two factors emerged as being important methodological aspects. Firstly, since neuronal cell death does not occur homogeneously within the CA1 region of the hippocampus, it is critical that the time post ischemic insult is accurately reported. Secondly, in terms of analysis regarding counting strategy, window size and position were both shown to have a major impact on study results and should therefore be clearly reported. Ward et al. make the point that in order to reproduce histopathological results from, e.g. the mouse, the pathology protocol, including necropsy methods and slide preparation, should be followed by interpretation of the slides by a pathologist familiar with reading mouse slides and familiar with the consensus medical nomenclature used in mouse pathology (Ward et al. 2017). Additionally, for the peer review of manuscripts where histopathology is a key part of the investigation, pathologists should be consulted.

The importance of such studies to the field is further acknowledged by the existence of numerous initiatives to improve reproducibility. For in situ hybridisation (ISH) and immunohistochemistry (IHC) biomarkers, the minimum information specification for ISH and IHC experiments (MISFISHIE) guidelines has been developed by the Stem Cell Genome Anatomy Projects consortium, and it is anticipated that compliance should enable researchers at different laboratories to fully evaluate data and reproduce experiments (Deutsch et al. 2008). The MISFISHIE checklist includes six aspects of information to be provided in the reporting of experiments ranging from experimental design, biomaterials and treatments, reporter (probe or antibody) information, staining protocols and parameters, imaging data and parameters and imaging characterisations. The use of statistics and any guidance on interpretation of results is, however, not included. The authors stress that the implementation of MISFISHIE should not remove variability in data but rather facilitate the identification of specific sources of variability. A similarly intended study describes a checklist of 20 items that are recommended to be included when reporting histopathology studies (Knijn et al. 2015). Thus, while reproducibility in histological analyses has been a problem and has perhaps hindered scientific progress, the field has adapted and adherence to new tools and guidelines that are now available offer hope that we are moving rapidly in a positive direction.

7 Ex Vivo Biochemical Analysis

Biochemical assessments can be performed in numerous ex vivo biological materials ranging from CSF to organoids and are routinely used to assess mRNA and proteins such as hormones.

Flow cytometry of ventricular myocytes is an emerging technology in cardiac research. Cellular variability and cytometer flow cell size are known to affect cytometer performance, and these two factors of variance are considered to limit assay validity and reproducibility across laboratories. In a study by Lopez et al., the authors hypothesised that washing and filtering create a bias towards sampling smaller cells than actually exist in the adult heart and they performed a study to test this (Lopez et al. 2017). The study results revealed that there was indeed a significant impact of washing and filtering on the experimental outcome and thus proposed a no-wash step in the protocol that could become part of a standard experimental design to minimise variability across labs.

Deckardt et al. have investigated the effect of a range of commonly used anaesthetics on clinical pathology measures including glucose, serum proteins, hormones and cholinesterase (Deckardt et al. 2007). The authors demonstrated differential effects of the different anaesthetics with regard to some of the measured parameters and differences across the sex and species used, thus demonstrating the importance of understanding the impact that an anaesthetic can have – even on ex vivo readouts – and to include appropriate controls. A similar study was conducted by Lelovas et al. which further highlights the importance of concise and accurate reporting of the use of anaesthetics in the collection of biological samples for biochemical readouts since their use can have a significant impact on outcome (Lelovas et al. 2017). Watters and Goodman published a comparison of basic methods in clinical studies and in in vitro tissue and cell culture studies reported in three anaesthesia journals (Watters and Goodman 1999). The authors identified 16 in vitro articles, and although they were not able to identify anything inherently wrong with the studies, they noted the small sample sizes and the lack of reporting on failures (only 2 of 53) and describe anecdotal evidence of experimenters only reporting on the experiments that work. The authors conclude with a call for all investigators to give reasons for sample size, to use randomisation and blinding wherever possible and to report exclusions and withdrawals, thus enabling an improvement in robustness and general applicability of the data published.

Antibodies are commonly used tools in research, particularly in ex vivo analyses. A common cause for the lack of reproducibility of data using antibodies could be due to the lack of thorough validation (Drucker 2016). The importance of standardised reagents has been highlighted by Venkataraman et al. who have described the establishment of a toolbox of immunoprecipitation-grade monoclonal antibodies to human transcription factors with the aim of improving quality and reproducibility across labs (Venkataraman et al. 2018). This work was conducted as part of the NIH protein capture reagents programme (PCRP) which has generated over 1,500 reagents that can be used by the scientific community.

8 Perspective

An improvement in quality in preclinical research and particularly where animals are used is urgently needed. To achieve this, it is of fundamental importance to change the way experimental results are reported in the scientific literature so that data can be more easily reproduced across labs. This should enable more rapid scientific progress and reduce waste. Scientists are encouraged to adopt the existing guidelines by defining all relevant information that has to be included in publications and study reports, with the aim of enhancing the transparency, reproducibility and reliability of scientific work. Ensuring that preclinical research proceeds along structured guidelines will strengthen the robustness, rigour and validity of scientific data and ultimately the suitability of animal studies for translation into clinical trials.

We have described several important factors relating to behavioural experiments that may influence the outcomes of some selected behavioural animal studies. Obviously, this represents only a small part of the various possible variations of the laboratory environment, equipment and methodological procedures that can affect animal behaviour. However, we have indicated the importance of considering and reporting all relevant details regarding behavioural experiments, which will help to resolve the common problem of poor reproducibility of certain findings across different laboratories and to ensure high quality of behaviour animal studies.

We have highlighted the use of anaesthesia and analgesia as factors that can have a significant impact on experimental data, and it is therefore of utmost importance that their use is reported comprehensively. High animal welfare standards require the use of anaesthetics and analgesics when performing painful and stress-inducing experiments. However, since these drugs may severely influence research outcomes, it is necessary to carefully select the most suitable procedures for the scientific question under investigation and to evaluate the importance of the scientific needs in the context of animal wellbeing and existing guidelines for the description of experimental animal research should be applied. The complete reporting of anaesthesia procedures as well as pain management could significantly improve the quality and reproducibility of preclinical animal studies and enhance animal welfare.

Ex vivo measures including histological analysis and biochemical readouts are seemingly just as prone to poor reproducibility as in vivo experiments. Clearly, the precise details of the in-life part of the study should not be overlooked in the reporting of such studies since this aspect can have a significant impact on overall experimental outcome and conclusions.

The field has reached a point where something needs to be done to improve standards, and indeed, to this end, numerous initiatives are ongoing. One such initiative is the Innovative Medicines Initiative consortium project “European Quality in Preclinical Research” (EQIPD). The EQIPD project aims to identify ways to enable a smoother, faster and safer transition from preclinical to clinical testing by establishing common guidelines to strengthen the robustness, rigour and validity of research data. Numerous academic and industrial partners are involved in this initiative, which should have a significant and positive impact in the next few years. Nevertheless, the output of EQIPD and similar efforts need to be embraced and for that, the entire scientific community has an important role to play.