For guidance on maintaining continuity of research and sponsored program activities during the COVID-19 crisis, please see ORSP’s Keep Discovering page.

Animal Number Justification

Guidance for PIs on Animal Number Justification

  • Purpose:  To help PIs better understand how to justify their requested number of animals.  Sample size calculation plays an important role at the planning stage of research projects to ensure sufficient subjects for answering the question(s) of interest.  If sample size is too large, study will waste resources.  If sample size is too small, study does not have enough power to detect treatment effects; if the study has to be repeated, it will waste resources.  While sample size calculation is important, going through the analytical process is useful on its own!
  • Background:
    • Sample size selected for animals should be “the minimum number required to obtain valid results.”  (PHS Policy US Government Principle III)
    • The Guide for the Care and Use of Laboratory Animals states that the requested number of animals be justified statistically whenever possible.
    • If conducting a study where statistical justification for animals is inappropriate (e.g., pilot studies, breeding protocols, teaching protocols, etc.), PIs must at a minimum provide information on how they arrived at the number of animals requested for the proposed activity.
    • 3 Rs (Reduce, Replace, Refine)
  • Determining Sample Size:
    • Use of Power Analysis:  When doing hypothesis testing, a sample size is calculated to achieve desired power (often .8 or .9) for detecting meaningful differences at a fixed Type 1 error rate.
    • The Power Analysis should match with the planned Statistical Analysis to be used in the study.
    • Estimates needed to determine sample size:
      • Target Power
      • Significance Level
      • Clinically important difference that we wish to detect
      • Additional nuisance parameters
    • Sample size estimates are approximate; parameters used in calculations are educated guesses.  Be conservative but realistic.
    • Guessing means and SDs is often the hardest part of power analysis.  Use of pilot studies, estimates from related studies, theoretical knowledge, or a range of minimum and maximum possible values can help.  A standardized effect size could also be used instead.
    • Statistical Justification on Protocol:
      • Hypothesis driven outline of the experimental design should be included in the protocol.
      • Number of animals proposed should include information on how many groups will be utilized, the number of animals in each group, and a statement on how the proposed research is related to any previous or published research when important to the justification of the number of animals proposed.
      • When appropriate, additional animals requested to account for potential experimental losses should be considered.
  • There are many statistical programs available to help PIs estimate sample sizes.  A free program called G*Power is available at   SPSS is available for Faculty Members through myolemiss (Tools & Resources, Software, SPSS).