Good experimental design and statistics are essential to improving reproducibility & translation which in turn can help reduce animal use, reduce experimental cost, and speed drug development. Without proper experimental design, it might take 86 animal research projects to produce just one effective, safe human drug. However, if researchers use best experimental design and use of statistics that could be reduced to just 7 animal projects.
Good science is good business is good welfare is good public health.
What To Do (i.e., avoiding Beagley's red flags and more)
- Blinded, randomized experiments
- Both data collector & animal husbandry staff should be blinded, when possible
- A single treatment should not be in a single row
- Repeated experiments
- Present ALL data (no-cherry picking)
- Use positive & negative controls
- Validate reagents
- Use appropriate statistical tests
- Use both male & female animals
- Systematically introduce variation into your experiment to help increase translation
Additional general guidance on appropriate statistical tests
- T-tests & 2-way ANOVA's are generally insufficient considering natural biological variation
- Split-plot & factorial designs are much more powerful.
- Cage = subplot & blocking factor
- Cage = nested in sex
- This design is very useful because we know there is a lot of variability between cages within a sex such as due to location on a rack (different noise/light exposure) or simply interactions between animals.
- Mead's resource equation can be used calculate sample size for complex factorial designs. The magic number is 20.
- To reduce animal use, pilot projects are not recommended as they are typically underpowered & tend to overestimate effect sizes.
- Remember to analyze your interactions properly
- Do not confound treatment & cage. Instead, reduce animal numbers per cage & split treatments across cages.