Inclusion and exclusion
How to define criteria for including or excluding animals, experimental units, samples or data points from an experiment
- Deciding what to include and exclude
- Inclusion criteria
- Exclusion criteria
- Criteria for excluding data points from the analysis
Deciding what in include and exclude
Before you begin your experiment consider if there any conditions that experimental units, animals or samples have to meet in order for them to be used or continue to be used in the experiment. Conversely, are there any conditions under which animals, experimental units or samples would be disqualified from the experiment? What about criteria for excluding data points from your analysis?
These are your inclusion and exclusion criteria. It is important to decide these before your experiment begins to prevent you from unconsciously biasing your results by deciding as the experiment progresses.
Inclusion criteria
Inclusion criteria define the conditions animals, experimental units or samples need to meet in order to be used, or continue being used, in your experiment.
Examples of inclusion criteria include:
- A minimum weight for animals to be considered eligible for surgery.
- A minimum tumour volume before potential tumour-reducing drugs are tested.
- A specific behaviour training threshold before being tested with a behavioural task.
- A minimum deficit for an induced experimental model to adequately model a specific aspect of a disease.
Exclusion criteria
Exclusion criteria define the conditions under which animals, experimental units or samples are excluded from your experiment or data points excluded from your analysis.
Examples of exclusion criteria for animals, experimental units or samples include:
- If any animal develops a movement disorder that would affect behavioural measurements, they will be removed from the study.
- If any animal reaches a humane endpoint, they will be removed from the study and euthanised (e.g. a maximum tumour volume).
- If any animal develops complications as a result of surgery, they will be removed from the study.
- If any samples are damaged while being collected the sample will be excluded (e.g. histological section is badly damaged during processing and cannot be used).
- If any sample volumes are not large enough to be reliably measured the sample will be excluded from the study (e.g. a blood sample is too small to accurately measure a specific metabolite).
- Any samples that do not meet prespecified quality control standards or that have unacceptable levels of contamination will not be used.
If any animals (or experimental units or samples) do meet your exclusion criteria, or fail to meet your inclusion criteria, you may end up with data points missing from your dataset. Remember to think about how you will deal with any missing data and include this in your analysis plans, you may need to consult a statistician. You should also consider what criteria you will use to decide if any data points should be excluded from your analysis.
Examples of exclusion criteria for data points:
- Data values outside a biologically plausible range will not be included in the analysis (e.g. a negative value for body weight).
- Data values resulting from an error during data collection will not be included in the analysis (e.g. someone forgets to switch on equipment for a recording).
- Typos in data values will not be included in the analysis (e.g. if errors are transcribed when copying from original data source).
- Outliers will be removed using a built-in model in a statistical package (e.g. Grubbs’ method in GraphPad Prism).
Remember when you are calculating your sample size, the number of experimental units per group is the number you need for the analysis. If you are anticipating some attrition you will need to take this into account and start your experiment with a larger sample size. For example, if your experiment involves a surgical intervention that has a 10% attrition rate, add an extra 10% to your calculated sample size account for expected attrition.
