Limits...
A statistical assessment of differences and equivalences between genetically modified and reference plant varieties.

van der Voet H, Perry JN, Amzal B, Paoletti C - BMC Biotechnol. (2011)

Bottom Line: It is proposed to present the results of both difference and equivalence testing for all relevant plant characteristics simultaneously in one or a few graphs, as an aid for further interpretation in safety assessment.A procedure is suggested to derive equivalence limits from the observed results for the reference plant varieties using a specific implementation of the linear mixed model.Three different equivalence tests are defined to classify any result in one of four equivalence classes.

View Article: PubMed Central - HTML - PubMed

Affiliation: Wageningen University and Research centre, Biometris, P,O, Box 100, NL-6700 AC Wageningen, Netherlands. hilko.vandervoet@wur.nl

ABSTRACT

Background: Safety assessment of genetically modified organisms is currently often performed by comparative evaluation. However, natural variation of plant characteristics between commercial varieties is usually not considered explicitly in the statistical computations underlying the assessment.

Results: Statistical methods are described for the assessment of the difference between a genetically modified (GM) plant variety and a conventional non-GM counterpart, and for the assessment of the equivalence between the GM variety and a group of reference plant varieties which have a history of safe use. It is proposed to present the results of both difference and equivalence testing for all relevant plant characteristics simultaneously in one or a few graphs, as an aid for further interpretation in safety assessment. A procedure is suggested to derive equivalence limits from the observed results for the reference plant varieties using a specific implementation of the linear mixed model. Three different equivalence tests are defined to classify any result in one of four equivalence classes. The performance of the proposed methods is investigated by a simulation study, and the methods are illustrated on compositional data from a field study on maize grain.

Conclusions: A clear distinction of practical relevance is shown between difference and equivalence testing. The proposed tests are shown to have appropriate performance characteristics by simulation, and the proposed simultaneous graphical representation of results was found to be helpful for the interpretation of results from a practical field trial data set.

Show MeSH
Simplified version of a graph for comparative assessment showing the seven outcome types possible for each characteristic analysed. After adjustment of the equivalence limits, a single confidence limit (for the difference) serves visually for assessing the outcome of both difference and equivalence tests. Here, only the upper adjusted equivalence limit is considered. Shown are: the mean of the GMO compared to the mean of the counterpart on an appropriate scale (square), its confidence interval (bar), a thick vertical line indicating zero difference (for proof of difference), and thinner vertical lines indicating adjusted equivalence limits on the same scale (for proof of equivalence). For outcome types 1, 3 and 5 the  hypothesis of no difference cannot be rejected: for outcomes 2, 4, 6 and 7 the GMO is different from its counterpart. Regarding interpretation of equivalence, four categories (i) - (iv) are identified: in categories (i) and (iv) there is a significant equivalence and non-equivalence, respectively, in categories (ii) and (iii) equivalence and non-equivalence, respectively, are more likely than not.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3050699&req=5

Figure 1: Simplified version of a graph for comparative assessment showing the seven outcome types possible for each characteristic analysed. After adjustment of the equivalence limits, a single confidence limit (for the difference) serves visually for assessing the outcome of both difference and equivalence tests. Here, only the upper adjusted equivalence limit is considered. Shown are: the mean of the GMO compared to the mean of the counterpart on an appropriate scale (square), its confidence interval (bar), a thick vertical line indicating zero difference (for proof of difference), and thinner vertical lines indicating adjusted equivalence limits on the same scale (for proof of equivalence). For outcome types 1, 3 and 5 the hypothesis of no difference cannot be rejected: for outcomes 2, 4, 6 and 7 the GMO is different from its counterpart. Regarding interpretation of equivalence, four categories (i) - (iv) are identified: in categories (i) and (iv) there is a significant equivalence and non-equivalence, respectively, in categories (ii) and (iii) equivalence and non-equivalence, respectively, are more likely than not.

Mentions: For the interpretation of results we recommend a graphical display, similar to those suggested by others [22,23]. However, certain adjustments are needed to account for the fact that equivalence limits are estimated values, and these are described in detail in the Methods section. Figure 1 presents a schematic simplified example of the display, showing the possible outcomes for a single characteristic. For any given characteristic there are then fundamentally seven possible types of outcome. Among these seven types there are four where the mean value of the GMO lies between the adjusted equivalence limits (types 1-4), and three where it lies outside the equivalence limits (types 5-7). It is assumed here that the line of no difference is in between the adjusted equivalence limits. If not, then the selected conventional counterpart is itself non-equivalent to the reference varieties and a separate, non-statistical discussion should consider the place and relative importance of difference and equivalence testing in the risk assessment.


A statistical assessment of differences and equivalences between genetically modified and reference plant varieties.

van der Voet H, Perry JN, Amzal B, Paoletti C - BMC Biotechnol. (2011)

Simplified version of a graph for comparative assessment showing the seven outcome types possible for each characteristic analysed. After adjustment of the equivalence limits, a single confidence limit (for the difference) serves visually for assessing the outcome of both difference and equivalence tests. Here, only the upper adjusted equivalence limit is considered. Shown are: the mean of the GMO compared to the mean of the counterpart on an appropriate scale (square), its confidence interval (bar), a thick vertical line indicating zero difference (for proof of difference), and thinner vertical lines indicating adjusted equivalence limits on the same scale (for proof of equivalence). For outcome types 1, 3 and 5 the  hypothesis of no difference cannot be rejected: for outcomes 2, 4, 6 and 7 the GMO is different from its counterpart. Regarding interpretation of equivalence, four categories (i) - (iv) are identified: in categories (i) and (iv) there is a significant equivalence and non-equivalence, respectively, in categories (ii) and (iii) equivalence and non-equivalence, respectively, are more likely than not.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3050699&req=5

Figure 1: Simplified version of a graph for comparative assessment showing the seven outcome types possible for each characteristic analysed. After adjustment of the equivalence limits, a single confidence limit (for the difference) serves visually for assessing the outcome of both difference and equivalence tests. Here, only the upper adjusted equivalence limit is considered. Shown are: the mean of the GMO compared to the mean of the counterpart on an appropriate scale (square), its confidence interval (bar), a thick vertical line indicating zero difference (for proof of difference), and thinner vertical lines indicating adjusted equivalence limits on the same scale (for proof of equivalence). For outcome types 1, 3 and 5 the hypothesis of no difference cannot be rejected: for outcomes 2, 4, 6 and 7 the GMO is different from its counterpart. Regarding interpretation of equivalence, four categories (i) - (iv) are identified: in categories (i) and (iv) there is a significant equivalence and non-equivalence, respectively, in categories (ii) and (iii) equivalence and non-equivalence, respectively, are more likely than not.
Mentions: For the interpretation of results we recommend a graphical display, similar to those suggested by others [22,23]. However, certain adjustments are needed to account for the fact that equivalence limits are estimated values, and these are described in detail in the Methods section. Figure 1 presents a schematic simplified example of the display, showing the possible outcomes for a single characteristic. For any given characteristic there are then fundamentally seven possible types of outcome. Among these seven types there are four where the mean value of the GMO lies between the adjusted equivalence limits (types 1-4), and three where it lies outside the equivalence limits (types 5-7). It is assumed here that the line of no difference is in between the adjusted equivalence limits. If not, then the selected conventional counterpart is itself non-equivalent to the reference varieties and a separate, non-statistical discussion should consider the place and relative importance of difference and equivalence testing in the risk assessment.

Bottom Line: It is proposed to present the results of both difference and equivalence testing for all relevant plant characteristics simultaneously in one or a few graphs, as an aid for further interpretation in safety assessment.A procedure is suggested to derive equivalence limits from the observed results for the reference plant varieties using a specific implementation of the linear mixed model.Three different equivalence tests are defined to classify any result in one of four equivalence classes.

View Article: PubMed Central - HTML - PubMed

Affiliation: Wageningen University and Research centre, Biometris, P,O, Box 100, NL-6700 AC Wageningen, Netherlands. hilko.vandervoet@wur.nl

ABSTRACT

Background: Safety assessment of genetically modified organisms is currently often performed by comparative evaluation. However, natural variation of plant characteristics between commercial varieties is usually not considered explicitly in the statistical computations underlying the assessment.

Results: Statistical methods are described for the assessment of the difference between a genetically modified (GM) plant variety and a conventional non-GM counterpart, and for the assessment of the equivalence between the GM variety and a group of reference plant varieties which have a history of safe use. It is proposed to present the results of both difference and equivalence testing for all relevant plant characteristics simultaneously in one or a few graphs, as an aid for further interpretation in safety assessment. A procedure is suggested to derive equivalence limits from the observed results for the reference plant varieties using a specific implementation of the linear mixed model. Three different equivalence tests are defined to classify any result in one of four equivalence classes. The performance of the proposed methods is investigated by a simulation study, and the methods are illustrated on compositional data from a field study on maize grain.

Conclusions: A clear distinction of practical relevance is shown between difference and equivalence testing. The proposed tests are shown to have appropriate performance characteristics by simulation, and the proposed simultaneous graphical representation of results was found to be helpful for the interpretation of results from a practical field trial data set.

Show MeSH