europa    seventh

The Biostatistics work package performs statistical analyses of the results from the rodent feeding studies, reports on the results, and provides recommendations on the statistical methodology for long-term feeding trials.


The statistical power attainable for the primary parameters of toxicological interest when using sample sizes as specified in the standard OECD protocols is evaluated in a prospective power analysis.


The purpose of the statistical analysis is to derive point and interval estimates of the difference in parameters between the NK603-fed groups and the control group. These estimates are compared to a null hypothesis of equality (classical approach, differences tests) and to null hypotheses specifying a difference equal to external equivalence limits (G-TwYST approach, equivalence tests). Equivalence limits indicate minimum differences of potential toxicological relevance. For setting the equivalence limits use is made of data from non-GM groups in the feeding studies of the GRACE project.



Paper published describing the statistical approach for equivalence testing in G-TwYST

An open access paper describing the statistical approach for equivalence testing in G-TwYST has now been published in Food and Chemical Toxicology. The method was tested on data from the GRACE project, as described in this paper. We will now progress by applying the method in the statistical analyses of the G-TwYST feeding trials, to be reported first in the statistical reports as project deliverables.


  • An equivalence testing method is proposed to assess the safety of regulated products
  • We combine data from a current study with test and control, and historical studies with assumedly safe reference products
  • The method is illustrated with animal feeding studies using genetically modified and reference maize varieties
  • A high statistical power of the equivalence test is the basis for the equivalence criterion
  • Generalized fiducial inference is used to integrate uncertainties from the historical and the current data


van der Voet H, Goedhart PW, Schmidt K (2017). Equivalence testing using existing reference data:  an example with genetically modified and conventional crops in animal feeding studies. Food and Chemical Toxicology, 109: 472-485. https://doi.org/10.1016/j.fct.2017.09.044


Latest news

Go to news >

Comments on statistical methods for safety assessment published

March 6, 2018


In safety assessments a large number of non-target effects  has to be analysed statistically. A recent paper published in the Journal of Agricultural and Food Chemistry advocates an approach which leads to over-optimistic conclusions. In a recent Correspondence to this journal the concerns about the misapplication of methodology are explained. See: Safety Assessments and Multiplicity Adjustment: Comments on a Recent Paper, available on the Journal of Agricultural and Food Chemistry website at http://pubs.acs.org/doi/abs/10.1021/acs.jafc.7b03686.

Registration for Final Conference open

1 March, 2018


On April 16, 2018 G-TwYST organises a conference on general conclusions and recommendations from three scientific projects: G-TwYST, GRACE and GMO90+. Registration for this conference is compulsory and can be done by completing the registration form at the G-TwYST website. Registration is open until April 6, 2018.

More information, including a preliminary programme, is available on the G-TwYST Final Conference webpage.