Statistical Analysis in the Safety Evaluation of Genetically‐Modified Crops: Equivalence Tests

Abstract

Risk assessment is an important step in the process of deregulating a genetically modified (GM) crop and its derived products for food and feed. The European Food Safety Authority (EFSA) evaluates safety by testing for the average equivalence between the GM crop and a group of commercial non‐GM reference varieties representing a population of crops with a history of safe use. This paper defines the equivalence tests described by EFSA in terms of parameters in the mixed model. A basic strategy for the equivalence tests is proposed. Generalized inference is employed to overcome two challenges in testing equivalences: the unconventional parameter of interest and the limited number of reference varieties. Balanced or partially balanced incomplete designs are recommended in equivalence testing. Simulation indicates this new method maintains the desired type I error rate and holds good power under practical designs. An illustrative data example is used to demonstrate the execution of the proposed method in SAS. Finally, issues associated with the equivalence tests and suggested topics for future research are discussed.

Equivalence criteria for the safety evaluation of a genetically modified crop: a statistical perspective

“…2·5 and 97·5) of the distribution of reference variety characteristics’ as the limits for GMO's relative deviation from the overall mean of references. Accordingly, Kang & Vahl (2014) proposed the following hypotheses under Model (2). …”

“… Note that because (1) and (2) the term under the square root of MEP is always greater than . Van der Voet et al (2011) refer to MEP 2 as the ‘outer’ confidence limit of the EL for ( μ T − μ R ) 2 . Kang & Vahl (2014) pointed out a series of issues associated with the two-step procedure. First, using the outer limit of the EL inflates the type I error rate of the testing procedure for a given set of hypotheses.…”

Equivalence analysis to support environmental safety assessment: Using nontarget organism count data from field trials with cisgenically modified potato

“…Rejection of this nonequivalence hypothesis implies that the difference is smaller than the LOC, and this can be considered as a proof of safety (Bross, ; Hothorn & Oberdoerfer, ; Millard, ; Perry et al, ). The advantages of using the equivalence concept for safety assessment have been described before (e.g., Perry et al, ; van der Voet, Perry, Amzal, & Paoletti, ; Meyners, ; Kang & Vahl, ; Goedhart, Voet, Baldacchino, & Arpaia, ; Vahl & Kang, ). A crucial argument in favor of equivalence testing is that the onus to do high‐quality, well‐replicated experiments with sufficient statistical power is placed on to those who wish to demonstrate the safety of GMOs (Perry et al, ).…”

G-TwYST harmonisation of statistical methods for use of omics data in food safety assessment

“…Consequently, it may be possible to include moderate amount of reference groups in the study. Univariate statistical tests that use these internal references (rather than historical data) to set equivalence limits have been proposed as well (van der Voet et al 2011, Kang andVahl 2014). The use of multiple internal references is not feasible in animal experiments.…”

“…In G-TwYST a method was proposed that allows to use data from previous (historical) studies to replace externally set equivalence limits (expert knowledge) [van der Voet et al 2017]. Tests that use internal references (rather than historical data) to set equivalence limits have been proposed as well [van der Voet et al 2011, Kang and Vahl 2014]. Figure 1 shows the performance of difference and equivalence tests for simulated data, for an example equivalence limit of 3.…”