{"id":12392,"date":"2023-08-24T14:13:28","date_gmt":"2023-08-24T18:13:28","guid":{"rendered":"https:\/\/research.ncsu.edu\/administration\/?page_id=12392"},"modified":"2026-02-16T11:06:08","modified_gmt":"2026-02-16T16:06:08","slug":"justification-of-numbers-of-animals-used-for-research-teaching-and-outreach-activities","status":"publish","type":"page","link":"https:\/\/research.ncsu.edu\/administration\/compliance\/research-compliance\/iacuc\/iacuc-procedures-and-guidance\/justification-of-numbers-of-animals-used-for-research-teaching-and-outreach-activities\/","title":{"rendered":"Justification of Numbers of Animals Used for Research, Teaching, and Outreach Activities"},"content":{"rendered":"\n\n\n\n\n

Standard<\/h2>\n\n\n
\n
\n \n
\n \n \n <\/span>\n <\/span>\n

Hypothesis-Driven Research Studies<\/h2>\n <\/summary>\n

\n \n\n

Hypothesis-driven research studies typically are those for which cause-and-effect conclusions are the end goal. There typically are different treatment groups included in the experimental design, one of which is a control group that represents a normal or unaltered condition of the animal. With these types of studies, the assumption is made that there are previous data available, either conducted by the PI or published in peer-reviewed literature, from which a Power Analysis can be conducted. A Power Analysis is the calculation used to determine the smallest number of animals needed per treatment based upon the desired significance level (p-value); desired statistical power (1 \u2013 \u03b2); and the magnitude or size of the effect (difference among treatments). There are a number of different approaches for conducting Power Analyses for both continuous (analysis of variance, regression analyses, etc.) and categorical (chi-square, non-parametric analyses, etc.) data. Examples of how to conduct a Power Analysis for a variety of different types of studies can be found on-line or in most statistic\u2019s textbooks. The decision with regards to which one of these to use is at the discretion of the PIs and not dictated by the IACUC.<\/strong><\/p>\n\n\n\n

Most Power Analyses require PIs to select or choose the desired significance level (p-value); statistical power (1 \u2013 \u03b2); and effect size (difference among treatments) for their studies. These also are left up to the discretion of PIs and are not dictated by the IACUC <\/strong>However, for most animal-related experiments, significance levels of less than or equal to 0.1 and statistical powers of greater than or equal to 0.80 are considered highly powered studies and typically are required for publication in peer-reviewed journals.<\/p>\n\n\n\n

The other component that is required to conduct a Power Analysis is some estimate of the normal variation of the dependent (response) variable in the animal population being studied. The most common measure of this is the standard deviation. In many situations, PIs know or can calculate the standard deviation from their own data or obtain a reasonable estimate from peer-reviewed publications. Data in peer-reviewed publications typically are presented as the treatment mean plus\/minus the associated standard error and include the number of units used in these calculations. The mathematical relationship between the standard error and the standard deviation is as follows:<\/p>\n\n\n\n

Standard error = standard deviation \/ square root of the sample size.<\/em><\/p>\n\n\n\n

This equation can be modified mathematically to obtain an estimate of the standard deviation by multiplying each side by the square root of the sample size as follows:<\/p>\n\n\n\n

Standard deviation = standard error x square root of the sample size<\/em><\/p>\n\n\n\n

Inclusion of the results from the Power Analysis in the justification of animal numbers is a requirement for IACUC approval of hypothesis-driven research studies. The most common language that is used to do this is as follows:<\/p>\n\n\n

\n

\u201cResults of the power analysis indicates that (number) <\/em><\/strong>of animals per treatment will allow the detection of a (percentage or other units)<\/em><\/strong> difference among treatments at a significance level of (p-value)<\/em><\/strong> with a statistical power of (1 \u2013 \u03b2)<\/em><\/strong>.\u201d<\/p>\n <\/blockquote>\n\n\n\n

This statement provides only a justification for the number of animals per treatment. The number of treatments in the study and the number of times the study is going to be repeated need to be included in the justification, as well. In other words, the IACUC needs to be able to determine how PIs obtain the total number of animals requested from the results of the Power Analysis. This can be done with the following equation:<\/p>\n\n\n\n

Number of animals per treatment (from Power Analysis) x number of treatments in study x number of times study will be replicated<\/em>.<\/p>\n\n\n\n

For example, if the results of the power analysis determined that 10 animals were needed per treatment for an experiment with 5 different treatments and the experiment needed to be replicated once per year for 3 years, then the total number of animals requested should be as follows:<\/p>\n\n\n\n

10 animals\/treatments x 5 treatments\/study x 3 replicates = 150 animals<\/p>\n\n\n\n

Most studies typically involve collection of data associated with more than one dependent (response) variable and many of these may differ in terms of the normal variation within the population being studied. In this situation, PIs only need to conduct a single Power Analysis and should use the dependent (response) variable that has the largest variation within the population being studied. <\/strong>For example, if a proposed experiment involves monitoring both weight gain and circulating levels of various hormones as response variables and weight gain has a higher coefficient of variation ([standard deviation\/mean] x 100) <\/em>compared with the hormone levels, then the power analysis should be conducted using weight gain as the dependent variable.<\/p>\n\n\n\n

It is also important for PIs to anticipate and plan for any experimental losses of animals \u2013 animals that have to be removed before the study is completed. The most common way to do this is to add the anticipated experimental losses after the total number of animals needed has been calculated. For example, if it is anticipated that 10% of the animals might need to be removed from the experiment prior to completion of data collection and the total animals needed is 150 (based on the power analysis), then addition of 15 animals (150 x 0.1 = 15)<\/em> would be appropriate and PIs should increase their total number requested to 165.<\/p>\n\n\n\n

Careful consideration by PIs should be given to their experimental treatments when estimating potential experimental losses. For some studies, there may be one treatment that, for whatever reason, might result in a disproportionate number of experimental failures compared with other treatments. In these cases, it is advisable to calculate the anticipated experimental losses for each treatment separately. For example, a study might contain three experimental treatments: one that has increased calcium levels; one that has decreased calcium levels and one with calcium levels at the recommended (maintenance) levels. This could lead to a higher percentage of animals being removed in the reduced calcium treatment in a long-term study due to lameness, etc. If the normal experimental losses for this species of animal is 10%, but the PI expects this to increase to 20% in the reduced calcium treatment these adjustments need to made after the numbers of animals per treatment are determined from the Power Analysis. Assuming that results from the Power Analysis indicated that 10 animals were needed per treatment, then the number of animals needed in the control and high calcium treatments would 11 (10 x 0.1=1; 10 + 1)<\/em> and 12 for the low calcium diet (10 x 0.2 = 2; 10 + 2 = 12)<\/em>. If the experiment was to be replicated 3 times, then the total animal request would be 103 (11 + 11 + 12 = 34; 34 x 3)<\/em>.<\/p>\n\n\n\n

Summary<\/h3>\n\n\n\n