Module IX Flashcards
- Describe the concepts of fidelity and discrimination, and discuss validation criteria for an induced and a spontaneous animal model.
Fidelity = general/overall similarity to what is being modelled.
Discrimination = specific similarity for one particular property to what is being modelled.
Use the seagull lookalike vs. the stick with red spots example. There are two models of a seagull mother, and they have a nest with seagull ducklings. In one group, they use a seagull lookalike, while the other group is just a stick with some red spots on it. When the ducklings see their mother they will make a sound and there are more responses from the stick than the lookalike. This is because the only clue that the ducklings have that this is their mother that’s the contrast between red and white from the spot on her beak, and that happens to have a much higher contrast on the stick than on the lookalike. So, the lookalike is a model of high fidelity, that means it looks in many ways like the seagull it is supposed to model, while the stick is a model of high discrimination, that means in one relevant phenomenon it it looks exactly like the seagull, while there are no similarities in other aspects. Most animal models are high discrimination.
Spontaneous animal models are when the phenomenon occurs spontaneously, such as the spontaneously hypertensive rat (SHR). Induced animal models are when the phenomenon is induced using chemicals, surgery, etc. There are also negative animal models where the phenomenon never occurs. Orphan models occur when animals develop diseases that look like human diseases (e.g. when cats develop obesity and diabetes).
External validity of an animal model refers to how well a model translates to the species it is modeling, typically humans. Construct validity is the extent to which both the model and the human phenomenon studies can be explained by the same theory (the cause of the disease is the same). Face validity is the similarity in appearance (the symptoms of FIV in cats are of high face validity to HIV in humans). Predictive validity is a measure of how much a drug has the same effect in humans and models tested.
Internal validity is the extent to which a study establishes a trust worthy cost and effect relationship between an intervention and the outcome.
It is dependent on procedures such as the experimental design, randomization, including proper controls, and the ability to reproduce the results.
- The size of our experiment is dictated by four elements: Variation, effect size, level of significance, and statistical power. Choose two of these elements, explain what they are, and how they influence how many animals we need to use.
The level of significance is a limit that is set before starting the study that determines how willing we are to accept false positives. The significance level is often set at 5% (p=0.05) meaning that 5% of the results will be a fluke. If the p-value is less than the significance level, then we can be relatively confident that our results are significantly different from our null hypothesis.
Statistical power is the ability to avoid false negatives. Statistical power is not reported in publications usually because to estimate the risk of a false negative would require us to know what the effect of our drug is. However, if we knew the effect of our drug, there would be no need to carry out a study. We often look at prospective power, which allows us to determine the chance of us discovering the effect of a treatment based on the smallest effect where we would still consider the treatment to be useful (e.g. how likely are we to discover the effect of a drug that can reduce the size of tumours by, say, 20%). We often term this the smallest effect size of interest. Typically, researchers will aim for having at least an 80% power in their studies – an 80% chance of finding this smallest effect size of interest. Studies that do not have a realistic chance at finding the effect that they are looking for are considered to be “underpowered.”
Variation. To estimate the variation ahead of time, you can review previously published literature or you can perform a small pilot study. The lower the variation, the more uniform the results will be and fewer animals will be needed to demonstrate an effect. Variation can be reduced by using more reliable equipment, etc. (see IX.3).
Effect size - The greater the difference we can expect between our treated animals and controls, the easier it will be to find. For example, if we want to use fewer animals in our study, we can let the tumors grow bigger, or use a stronger dose of our experimental drug.
- Describe relevant factors and possible sources of bias, when planning an animal experiment, and possible actions to prevent them.
Bias, in the scientific sense, is something that skews our results in a non-random way. This is different from variation. Researchers are the greatest source of bias. Blinding and randomization can help to reduce this. Additionally, whenever the animals of our experimental groups are inherently different, or we are unable to treat them equally, we run the risk of biasing our experiment. We can mitigate this through using different experimental designs.
- When planning an experiment, what do we mean by “experimental unit?” Give two examples of experimental units (of different “size”).
The experimental unit is the smallest element in our study for which we can draw any conclusions. This is often the same as an individual animal, but not always. For example, since mice are social animals, they are housed together. This means that readings of food or water consumption cannot be easily determined for each individual animal. In this case, the experimental unit would be one cage (e.g. one cage with three mice). An experimental unit can also be within an animal. For example, 2 tumours can be implanted, one on each side of a mouse, and only one will be given the treatment. This means that an experimental unit would be each tumour and the mouse acts as its own control.
- Why do we decide on the number of animals we will be using in an experiment before we start; and how can we do that in a sensible way?
The calculation starts with an analysis plan. We decide which measures we will be collecting in our study and then we ask ourselves what we want to do with those results. Once we have a plan – this often involves a statistical operation called a hypothesis test – we can find tools and equations to help us determine our number of experimental units – what is called sample size (n). See IX.2 for explanations on variation, effect size, level of significance, and statistical power.
- Sometimes when we plan experiments, we realize that there are benefits to using something other than a simple case-control study. Give an example of a more complex design, of when we might use it, and why.
Cross-over design: take the same group of animals and, in sequence, put them through different conditions or treatments. For a period of time we treat our animals and collect our measurements. For a period of similar length, we can then apply another treatment condition – for example a control period with no treatment. Washout periods are added between treatments.
Full crossover design: For two conditions, half our animals start with one condition and the other half with the control condition.
Latin square design: If we want to test more than two conditions, we employ what is called a Latin square design, where we create a pattern where every animal undergoes each treatment, and each treatment is equally represented in each time period.
With no dedicated control group, we can halve the number of animals we would use for a two-condition case-control study. Additionally, with every animal serving as its own control, we have minimized the biological variation in our study, meaning we can further reduce the sample size of our study. Note that crossover designs cannot be used if the effect of the treatment remains after the treatment and washout periods.
There are also multiple treatment case-control designs where there are multiple experimental groups instead of just a case-control. There are multifactorial designs where animals can be separated in a matrix (e.g. separate by sex, age, and treatment/control would create eight distinct groups.
- Sometimes when planning an experiment we may need outside help with our statistics. Give an example of when this might happen and of the elements of your study that you need to be able to present to, for example, a statistician.
When the statistics start to become more complex, it might be necessary to consult with a statistician. For example, in a multifactorial design, there could be many different tests between and within many different groups. This could get complicated and confusing quickly, especially when determining the sample sizes required for each group. A statistician can run a simulation to determine the required sample size. Note that you will still need to provide the variation, effect size, level of significance, and statistical power.
- How can pilot studies help us when planning our experiments? Can we use a pilot study to estimate an effect size?
A pilot study is a small-scale preliminary study conducted to evaluate feasibility, duration, cost, adverse events, and improve upon the study design prior to performance of a full-scale research project. A pilot study can be performed to get an estimation of the variation within groups (e.g. within the case and control groups). If there is not a study that describes the effect size published in the literature, a pilot study can be used to estimate effect size.
- Explain the principles of GLP in experimental animal research.
Good Laboratory Practice (GLP) is determined by law and it is a quality assurance system that ensures the reliability, integrity, and validity of non-clinical laboratory studies, including experimental animal research.
GLP helps to ensure safety and aims to ensure the consistency, reliability, uniformity, and quality of chemical non-clinical safety tests. It also aids with reproducibility.
Elements of GLP can be:
- Documents and records.
- Statistical procedures
- Instrumental validation
- Specimen/sample tracking
- Certification of lab facilities
- Certification of analysis
- Reagent/materials certification
- Explain key issues to be reported when publishing experimental animal studies.
Introduction - Provide sufficient scientific background to understand the motivation and context for the study and relevant references to previous work. Explain the experimental approach and rationale, why the animal species and model being used can address the scientific objectives, and the study’s relevance to human biology. The primary and any secondary objectives of the study, or specific hypotheses being tested, should also be noted.
Materials and Methods:
● Ethical statement - Ethical review permissions, relevant licenses and national or institutional guidelines for the care and use of animals
● Study design
○ a. The number of groups.
○ b. Any steps taken to minimize bias.
○ c. The experimental unit
● Experimental procedures (How, When, Where, Why)
● Sample size
○ Total number of animals and how this was decided
○ Independent replications of each experiment
● Allocation of animals to experimental groups
● Experimental outcomes (Primary and secondary)
● Statistical methods
● When describing the species, strain, age, sex, etc., use ILAR nomenclature
● Housing and Husbandry
○ Measures to protect microbiological status
○ Housing equipment
○ Number of animals per cage or housing unit
○ Bedding type, quality, pretreatment
○ Environmental temperature
○ Relative humidity
○ Lighting schedule
○ Ventilation and filtration
○ Measures to refine experimental techniques to benefit animal
○ Diet (type and composition, pretreatment, feeding schedule)
○ Water (type, quality, pretreatment, watering schedule)
Discussion:
● Interpret the results, taking into account the study objectives and hypotheses, current theory, and other relevant studies in the literature.
● Comment on the study limitations including any potential sources of bias, any limitations of the animal model, and the imprecision associated with the results
● Describe any implications of your experimental methods or findings for the 3Rs
● Comment on whether, and how, the findings of this study are likely to translate to other species or systems, including any relevance to human biology.
