Experimental Design

Experimental design can contribute to reduction of animal use when animal using experiments are planned with consideration of the following aspects of experimental design:

Control of Variation

When choosing an experimental design consider:

  • which experimental unit will be used (e.g., individual animal, litter, cage)
  • what randomization process will be used
  • whether blinding to treatments should be used to avoid bias
  • using pilot studies to test the logistics of a proposed experiment and/or determine the likely response
  • the type of design that will be used (e.g. completely randomized, randomized block, factorial, Latin square, crossover, repeated measures, spilt-plot, incomplete block, or sequential designs)

Variation

Fixed effects variation can result from differences in species, sex, strain, age, experimental conditions, bedding and diet of experimental animals. These can be controlled and/or deliberately varied as part of the experiment.

Random effect variation results from inter-individual variation, non-systematic measurement error, and variation associated with time and location. This type of variation can be minimized with choice of experimental subject (e.g. use of inbred strains of animal) and experimental design. Using isogenic strains of rats and mice can decrease variation due to their genotypic and phenotypic uniformity.

Sample Size

Justification of sample size and number of animals to be used is an important component of animal care protocols. Using a too small sample size may not detect biologically important effects, while a sample size that is too large will waste animals. Consider using one of these statistical approaches to select sample size:

  • Power Analysis: requiring knowledge about the effect size of interest and standard deviation from previous experiments or a pilot study
  • Resource Equation Method: when there is no information about standard deviation or effect size

Some Canadian universities provide access to statistical consulting.

Typical Errors in Design of Animal-based Experiments

  • using an experimental design repeatedly without proper scientific justification (i.e. because it is common in the literature or the investigator is comfortable with it)
  • failing to correctly identify the experimental unit
  • lack of power (i.e. experiment is unable to detect clinical or biologically relevant responses)
  • uncontrolled variation (experimental units not allocated randomly, or treated and control animals kept separately and measured at different times and/or by different people)
  • potential for bias (un-blinded measurement of subjective measures; animal pain and/or distress)

Presentation of Results

Clear presentation of results is the final step of good experimental design and statistical analysis. It includes accounting for every experimental subject, appropriate measures of variance, and discussion of the biological relevance of any statistically significant and non-significant results.

Guidance on best reporting practices can be found in the ARRIVE Guidelines.

This section has been adapted from Festing and Altman (2002).

Quick Links

Publications