基因改造作物安全性評估之統計分析

Abstract

在基因改造作物及其衍生出的食物及飼料上市之前，安全性評估為一重要議題，其通常利用等價性檢定，即對於基因改造作物和一群隨機抽樣的已上市商用傳統作物來進行比較。等價性檢定的主要目的是在調查實行生物工程的過程中，基因改造作物的潛在非預期變化，是否在商用傳統作物的自然變化範圍之內。而歐洲食安組織 (EFSA) 也對於多重區域試驗提出一些規定來進行安全性評估。 在本篇研究中，針對歐洲食安組織 (EFSA) 所提出的規定，我們討論兩種試驗設計模式，包括均衡完全區域設計和均衡不完全區域設計。安全性評估方式則利用Kang and Vahl (2014, Crop Science, 51:2183-2200) 所發表的平均等價性檢定，其基礎建立於兩個線性模式，一為商用傳統品種，另一為基因改造作物和非基因改造對應作物。透過廣義樞紐量的概念，來建立等價性檢定，並藉由統計模擬和多筆資料分析來評估所提出的試驗設計模式表現和統計推論是否表現良好。模擬結果顯示，所提出的試驗設計模式及檢定方法對於日後實際上的應用確實有所助益。

Safety assessment has been an important issue before a genetically modified (GM) crop and its derived foods or feeds are released to the market. A compositional equivalence test is commonly conducted for the safety assessment of a genetically modified (GM) crop that is compared with a group of conventional reference varieties representing a population of crops that have a history of safe use. The main purpose of such equivalence tests is to investigate whether potential unintended changes in the GM crop caused from the bioengineering process are within the natural variation of conventional crops. The European Food Safety Authority (EFSA) provided some guidance for the multi-environment field trial to conduct the safety assessment. In this thesis, we discuss two classes of experimental designs to meet the requirements of the EFSA guidance, including BCSDs (balanced complete site designs) and BISDs (balanced incomplete site designs). We adopt the scaled average equivalence test presented by Kang and Vahl (2014, Crop Science, 51:2183-2200) based on two linear models for fitting the observed values of conventional reference varieties, and GM and non-GM counterpart, respectively. The concept of a generalized pivotal quantity is applied to develop the testing procedure for the equivalence. The performance of the proposed designs and statistical inference are evaluated through statistical simulation studies and numerical data analysis. It is shown that the proposed design and testing procedure can prove useful in practical use.