How to account for animals with different strains, ages or sexes in the design and analysis of an in vivo experiment.
- The importance of animal characteristics
- Strain and substrain
- Genetic modifications
- Representing animal characteristics in the EDA
When planning an in vivo experiment you should always choose the animals that will best answer your specific scientific question. This includes not just the animal model but also the characteristics of the animals (e.g. strain and substrain, sex, age and baseline weight). If and how you take these into account in your experimental design, including analysis plans, affects the conclusions you can draw from your results. Failing to account for animal characteristics may mean you are unable to make robust conclusions from your experiment.
You may choose to use inbred or outbred animals depending on the aims of your experiment. Inbred strains do not necessarily have less variability in response to tests or treatments than outbred stocks (Tuttle et al, 2018), but they are generally better defined both genetically and phenotypically. Inbred strains are therefore often chosen for population genetic mapping and molecular genetics studies, for example.
You should also consider which outbred stock or inbred strain to use, again based on your specific scientific question. Each stock, strain, and in some cases substrain, has unique characteristics, such as fecundity, susceptibility to specific diseases or sensitivity to certain pharmaceuticals.
To make your results more generalisable you can use several different inbred strains or multiple different stocks. This can be done without increasing overall animal numbers by using strain or stock as a blocking factor.
Regulators, funders and publishers are encouraging researchers to study both sexes, across all stages of the research pipeline.
The misconception that female mice are intrinsically more variable due to the oestrous cycle is often used to justify studying only males. A meta-analysis comparing variance of male and female mice (without controlling for the oestrus cycle) found that variability was not significantly greater in females for any endpoint – but was in males for several traits.
Using male and female animals in an experiment allows you to generalise your findings to the whole population and so improves the external validity of your experiment. Sex differences exist in disease prevalence, symptoms and progression, as well as in treatment efficacy and side effects. Large scale studies with rodents frequently find that sex modifies the treatment effect. Using both sexes should always be the default experimental design, and if you plan to only examine one sex you will need to consider and justify this decision carefully.
You can ensure equal proportions of male/female animals in each group and prevent the introduction of sex bias in your experiment using randomisation. For example, by separately randomising males and females to the experimental groups.
How to calculate the sample size and the exact analysis method to use both depend on what conclusions you want to be able to draw about sex. In your analysis, sex can be used either as a blocking factor or a factor of interest, depending on the purpose of your experiment (see below).
If the purpose of your study is to determine the overall effect of an intervention, and sex differences are not of direct interest, use sex as a blocking factor in your randomisation and analysis.
For example, scenarios where the overall response may be different in males and females (e.g. the response is stronger in males) but you have no reason to believe that the intervention has different effects in males and females (e.g. you do not expect the direction of the response to be different, or the treatment to only have an effect in one sex). An example of this is a genetically altered mouse line that decreases the white blood cell count in both sexes, with a slightly larger decrease in males.
Including sex as a blocking factor in your experimental design and statistical analysis takes into account the variability due to sex differences. This makes it easier to detect the real effect of the intervention (or treatment) without increasing the total number of animals used in your experiment. If you use sex as a blocking factor, the sample size in each experimental group will be the same as if you were using a single sex, with each group consisting of both male and female animals (ideally half and half).
Sample size should be determined using a power calculation based on a biologically relevant effect size for the main independent variable of interest (e.g. drug treatment).
When sex is included as a blocking factor and the treatment/intervention is the only independent variable of interest, the statistical analysis performed should be a one-way ANOVA with a blocking factor. This will allow you to test if there is an overall difference between treated and control animals, but will not tell you if males and females respond differently to the intervention (i.e. it will not generate a p value for the interaction between treatment and sex). Summary/descriptive statistics for each sex can be plotted separately (e.g. a boxplot) to see if the sex of the animals leads to a visibly notable difference in the response to the intervention.
If you were to observe a difference in the treatment responses between sexes, you should consider including sex as an independent variable of interest in future experiments. This would require a suitable sample size to allow a comparison of treatments to be made separately for males and females (see below).
If the purpose of your study is to investigate whether the effect of the intervention depends on sex, use sex as a factor of interest in your experiment. If the intervention is the only independent variable of interest in your experiment other than sex, you can analyse your results using a two-way ANOVA with two factors of interest: sex and treatment/intervention.
If your experiment uses sex as a factor of interest, it is likely you are planning post-hoc comparisons to test the effect of the intervention or treatment in females and males separately. In this case males and females at different treatment levels are considered different experimental groups. For example, if the treatment is a drug there would be one male and one female group for each dose.
As usual, sample size should be determined using a power calculation based on a biologically relevant effect size for the main independent variable of interest. However, you may need a different power calculation to work out the sample size for a study where you plan to compare all groups using post-hoc tests vs one that does not (seek help from a statistician if unsure).
Using sex as a factor of interest may mean using more animals compared to a single-sex experiment or an experiment using sex as a blocking factor. This approach is more sensitive (i.e. the statistical power is higher) than using males and females in separate experiments and analysing the male and female data separately. Including both sexes and using sex as a factor of interest allows you to compare males and females and, if powered appropriately for post-hoc tests, to detect if the treatment effect depends on sex (which is not possible with a separate experiment for each sex).
If you are not planning post-hoc comparisons to test the effect of the treatment in males and in females separately, the sample size should be determined using a power calculation for a two-way ANOVA (seek help from a statistician if unsure). The experiment will be able to test the overall effect of the treatment/intervention and whether it depends on sex, but there may not be enough animals per group to reliably test the effect of the intervention/treatment in females and males separately.
Using animals at the age most relevant to the biology you are studying means your results are more likely to answer your research question. Different organs and systems reach developmental maturity at different times. Carefully consider the precise age that is most relevant for your research question.
For example, using juvenile animals of an Alzheimer’s model could limit the applicability of study results – depending on the Alzheimer’s model used and whether the study is focused on early- or late-onset Alzheimer’s.
It is important that genetically modified animals used to model human disease recapitulate the specific aspects of the disease relevant to your scientific question. When selecting a model also consider the approach used to create the model – different genetic modification methods will affect the chances of off-target effects or transgene copy number effects, for example. The genetic background the modifications are made in and what has been done to verify both the precise genetic change and the resulting phenotype, are also important factors when choosing a genetically modified animal model.
The animal characteristics node in an EDA diagram contains detailed information about the animals that will be used in your experiment. Details of species, strain and substrain, sex, age or developmental stage, and weight are important, as outlined in the ARRIVE guidelines item on experimental animals. Including them in your EDA diagram will ensure you have recorded key details making describing your study in compliance with the ARRIVE guidelines easier.
Distinct animal characteristics nodes are used for distinct combinations of animal characteristics, details of which can be entered in the properties of each node.