Research Article: Comparative analysis of genetically-modified crops: Part 1. Conditional difference testing with a given genetic background

Date Published: January 16, 2019

Publisher: Public Library of Science

Author(s): Changjian Jiang, Chen Meng, Adam Schapaugh, Guangyuan He.

http://doi.org/10.1371/journal.pone.0210747

Abstract

The European Food Safety Authority (EFSA) mandates two sets of statistical tests in the comparative assessment of a genetically-modified (GM) crop: difference testing to demonstrate whether the GM crop is different from its appropriate non-traited control; and equivalence testing to demonstrate whether it is equivalent to conventional references with an history-of-safe-use. The equivalence testing method prescribed by EFSA confounds the so-called GM trait effect with genotypic differences between the reference varieties and non-traited control. Critically, these genotypic differences, which we define as a ‘control background effect’, are the result of conventional plant breeding. Thus, the result of EFSA equivalence testing often has little or nothing to do with the GM trait effect, which should be the sole focus of the comparative assessment. Here, an integrated method is introduced for both difference and equivalence testing that considers the differences of the three genotype groups (GM, control, and references) as a two-dimensional random variable. A novel statistical model is proposed, called the trait model, that treats the effects of the GM and control materials as fixed for their difference, and as random for their common background. For significance testing, the covariance structure of the three genotype groups is utilized to decompose the differences into the trait effect and the control background effect. The trait difference is then derived as a conditional mean, given the background effect. The comparative assessment can then focus on the conditional mean difference, which is independent of the control background effect. Furthermore, the trait model is flexible enough to include various types of genotype-by-environment (G×E) interactions inherent to the experimental design of the trial. Numerical evaluations and simulations show that this new method is substantially more efficient than the current EFSA method in reducing both Type I and Type II errors (protecting both the consumer and producer risk) after the background effect is removed from the test statistic, and successfully addresses two major criticisms (i.e. statistical model lack of G×E, and study-specific equivalence criterion) that have been raised.

Partial Text

In the European Union, the safety of a genetically-modified (GM) crop and derived food/feed is established, in part, using a proof-of-equivalence approach [1]. Under this paradigm, any phenotypic or compositional differences between the GM crop and its near-isogenic, non-traited comparator are evaluated in the context of natural variation, estimated from conventionally-bred varieties (hereafter, references) grown in the same field trials. Currently, the GMO Panel of the European Food Safety Authority (EFSA) mandates two sets of statistical tests in the comparative assessment of a GM crop [2]: difference testing to demonstrate whether the GM crop is different from its near-isogenic, non-traited comparator; and equivalence testing to demonstrate whether it is equivalent to conventional references with an history-of-safe-use. Per [2], this assessment should be based on data from a minimum of eight sites in one growing season (or two growing seasons each with four sites). Within each site, three genotype groups are assigned to plots under a randomized complete block design. The genotype groups include: a test (GM) variety, a control (the near-isogenic comparator), and a set of (minimum six) conventional references.

In this section, the asymptotic and sample performance of the conditional method is evaluated for a wide range of parameter values. Simulations are conducted on selected parameter sets to compare the conditional method and trait model with the EFSA method. Lastly, the EFSA example [2] on maize grain composition data is reanalyzed using the conditional method and results are compared. For the numerical evaluation and simulation, a typical TCR trial was assumed, i.e. ns = 8, nb = 4, ngs = 4. For illustration of the conditional method, all references were assumed different, i.e. ng = nsngs. The variation setting was defined as the ratio between the reference (or genetic) versus residual variances σg2:σe2. Zero G×E interactions were assumed in data generation for simulations.

The equivalence testing method prescribed by [2] assesses differences between the test and references. These differences, as the main topic of this paper, may be driven by a trait effect, a control background effect, or both. The mixed models proposed in the literature and adopted by EFSA confound the trait and control background effects. Two direct consequences are worth noting. The first is the mismatched outcome type (i.e. non-significance in the difference testing and non-equivalent to the conventional references) that may result in the two-step procedure—using ΔTC in difference testing and ΔTR in equivalence testing. Second, if the control background effect is large, the result of EFSA equivalence testing has little or nothing to do with the trait effect, which should be the sole focus of the comparative assessment. The proportion of background variation in the test-reference difference can be estimated as the variation reduction from the conditioning:
RDTR|DCR2=1−σDTR|DCR2σdTR2≈(1−σdTC22σdCR2)2

The equivalence testing method prescribed by [2] directly compares a GM crop with commercial references, and confounds the so-called GM trait effect with genotypic differences between the reference materials and control. Critically, these genotypic differences are the result of conventional plant breeding. Thus, the result of EFSA equivalence testing often has little or nothing to do with the GM trait effect, which should be the sole focus of the comparative assessment. A novel statistical model, called the trait model, was introduced that treats the effects of the GM and control materials as fixed for their difference, and as random for their common background. For significance testing, the covariance structure of the three genotype groups is utilized to decompose the differences into the trait effect and the control background effect. The trait difference is then derived as a conditional mean, assuming a background effect equal to zero; this is equivalent to statistically adjusting the control background to the reference mean. Consequently, the comparative assessment is independent of the natural genotypic differences between the control and references, which substantially reduces both Type I and Type II errors and corrects a fundamental flaw in the testing method mandated by the GMO Panel of the European Food Safety Authority.

 

Source:

http://doi.org/10.1371/journal.pone.0210747

 

Leave a Reply

Your email address will not be published.