Selected issues of statistical evaluation of binary animal toxicity data

The statistical evaluation of binary data from animal toxicity studies is complex and lack of awareness of this complexity can lead to misinterpretations. The FSVO is therefore providing a document to raise authorities’ awareness of these difficulties.

One way in which authorities evaluate the harmful potential of substances is on the basis of studies in which groups of animals are exposed to different doses of the substances and the resulting effects are established.

The aim of these studies is:

  • to detect as many types of treatment-related effects as possible (type of hazard); and
  • to identify the lowest dose that triggers the first harmful effects (toxicological threshold dose).

Statistical significance and biological relevance

When evaluating an animal study, the first step is to use statistical methods to check whether observed effects may be treatment-related. The evidence is then examined to determine whether it is biologically plausible to attribute statistically significant effects to the treatment. The conclusion as to whether or not an effect is likely to be treatment-related is therefore always based on both a statistical and a biological analysis. Of special interest are binary data (data of the type: effect present, “yes” or “no”; e.g. tumours or malformations), on the one hand due to the severity of such effects and on the other hand because the interpretation of binary data is often difficult, both statistically and biologically.

Raising awareness of the complexity of statistical evaluation

The FSVO is providing a document to raise authorities’ awareness of the difficulties inherent in the statistical evaluation of binary data and the use of historical control data (see “More information”). The document summarises the available literature and internationally developed guidelines, accompanied by an FSVO commentary. It covers only certain aspects of data interpretation and does not claim to be comprehensive.

More information

Last modification 18.09.2019

Top of page

https://www.blv.admin.ch/content/blv/en/home/lebensmittel-und-ernaehrung/publikationen/wissenschaftliche-publikationen/aspekte-statistischen-beurteilung-tierstudien.html