Animal Subjects in Research : Statistical Section of the Animal Use Approval Form

Regarding the appropriateness of animal numbers, an excellent discussion on statistical design has been prepared by Farol Tomson, DVM at the University of Florida.

The IACUC sometimes has difficulty understanding the way researchers come up with the number of animals requested in their forms. Three questions are asked on the Animal Use Approval Form
  1. Approximately, how many animals will be needed for this project?
  2. How many experimental groups, replications, trials, etc. are required?
  3. How did you determine that the sample size, number of groups, replications, trials, etc. are appropriate, as they relate to the number of animals requested?
The following page is intended to help researchers prepare good and acceptable justification statements that explain why they need a certain number of animals. Included are suggested statements to put in the Animal Use Approval Form.
Indicate if protocol is for an exploratory or confirmatory study.
    Good things to include in protocol if a confirmatory study.:
  • Plan for animal use (table or chart) to be included.
  • Randomization method to be described (not required, but good to include.
  • Sample size justification (power analysis) must be included.
  • Technical support should be included.
Plan for animal use:Sample of table which summarizes plan for animal use.
  20 minutes post injection 40 minutes post injection 60 minutes post injection
Steroid Group n=4 n=4 n=4
Non-Steroid Group n=4 n=4 n=4
Total 8 8 8
Total number of animal requested = 24 + 5 [for training, or other purposes (please explain)] = 29
Randomization: Arguments for randomization include the following:
  • To avoid bias
  • To satisfy requirements of statistical methods used to analyze data
How to randomize
  • Software available
  • Use of random number tables
Suggested wording for the Animal Use Approval Form:
  • (Name here) will accomplish randomization by means of (software name or random number table).
  • There is no opportunity for randomization since (reason here).
Power Analysis:Arguments for doing a power analysis include the following:
  • Assure that probability of missing an important difference is small
  • Reduce cost of experiment
  • Reduce wastage/suffering of animals
Resources available for doing a power analysis
  • Software such as STPLAN (free). STPLAN is available by anonymous ftp from odin.mda.uth.tmc.edu.
  • Consult statistician: Biostatistics Consulting Lab (BCL), Paul Kubilis and Mike Ernst (not free).
Requisites for doing a power analysis
  • Selection of appropriate statistical method for analyzing data.
  • Specification of null and alternative hypotheses.
  • Specification of minimal effect size that should be detectable.
  • Estimation of ``nuisance'' parameters (often the variance, is not of interest and must be estimated by pilot data or using the crude relation
  • Significance level (chance of rejecting null hypothesis when null is correct) must be specified.
  • Power level must be specified (80% is usually acceptable).
How to Increase Power
  • Make significance level bigger.
  • Increase sample sizes.
  • Specify a larger effect size.
Suggested wording for the Animal Use Approval Form to describe your Technical Support: The above power analysis (description provided) was completed by (name here). We will use the resources of (name here) for management, analysis, interpretation, and presentation of our data. Technical support and guidance will be provided by (name here).