2012年9月，法国研究人员在国际学术期刊《食品和化学毒物学》杂志（FoodandChemicalToxicology）上发表转“基因玉米致癌”论文，一度引发公众对转基因作物的恐慌，并且成为部分人士反对转基因食品的重要证据。但是，也遭到某些人的质疑。11月29日，《食品和化学毒物学》杂志出版方爱思唯尔集团（Elsevier）28日在美国宣布，由于研究方法和结论皆存在严重问题，决定撤除这篇论文。对此，塞拉利尼团队发表声明，称将坚持自己的结论。声明说，以往的研究发现凡是显示转基因农作物有负面效果的，都会被监管者从实验到统计方法做严格的重审，凡是声称转基因农作物安全的研究，都被照单接受。只要是没有报告负面效果的研究，都被接受为“安全”的证明，无论他们的研究方法有何种不足(被认为无关紧要)。

 “转基因玉米致癌”论文被撤稿 塞拉利尼发声明

　　顾秀林在其博客中发布了塞拉利尼团队声明的中英对照本（2013-11-28）。并希望严建兵教授袁越先生以及差一点乐得发疯的挺转派们来一个真正的科学批判。乌云遮天难持久，科学骗人难善终。转基因用于农业之邪恶，我们大家心里其实都明明白白。

　　以下是塞拉利尼团队的声明： 

　　我们是FCT一年多前发表的论文的作者，关于农达和耐受农达的转基因生物的事（塞拉利尼等2012）。对于同样的质疑，我们已经在同一个刊物上回应过（塞拉利尼等，2013），即：作为正常的科学辩论，仅仅由于实验鼠品系的选择和数量的原因，就判定研究结果“结论不完整”，这是不能接受的。我们坚持我们的结论。我们早已公布了对相同的质疑所做的回答，但至今没有见到对我们的任何回应（塞拉利尼等，2013）。

　　We, authors of the paper published in FCT more than one year ago on the effects of Roundup and a Roundup-tolerant GMO (Séralini et al., 2012), and having answered to critics in the same journal (Séralini et al., 2013), do not accept as scientifically sound the debate on the fact that these papers are inconclusive because of the rat strain or the number of rats used. We maintain our conclusions. We already published some answers to the same critics in your Journal, which have not been answered (Séralini et al., 2013).

　　关于实验大鼠品系

　　同一个大鼠品系，被用在研究致癌性和慢性化学毒理学的美国国家毒理学项目中(King-Herbert et al., 2010)。SD大鼠是常规性用于毒理和致癌效果实验中的动物，其中有孟山都公司的90天实验，被当做批准NK603转基因玉米应用的依据，其他转基因农作物也是这样做的(Sprague Dawley rats did not came from Harlan but from Charles-River) (Hammond et al., 2004; Hammond et al., 2006a; Hammond et al., 2006b).

　　Rat strain

　　The same strain is used by the US national toxicology program to study the carcinogenicity and the chronic toxicity of chemicals (King-Herbert et al., 2010). Sprague Dawley rats are used routinely in such studies for toxicological and tumour-inducing effects, including those 90-day studies by Monsanto as basis for the approval of NK603 maize and other GM crops (Sprague Dawley rats did not came from Harlan but from Charles-River) (Hammond et al., 2004; Hammond et al., 2006a; Hammond et al., 2006b).

　　这里有一个简明的初步的文献清单，表明在同行评审的杂志上SD大鼠被用在36个月的实验如(Voss et al., 2005) or in 24-month studies by (Hack et al., 1995), (Minardi et al., 2002), (Klimisch et al., 1997), (Gamez et al., 2007).，其中有一些文章就发表在FCT上。

　　A brief, quick and still preliminary literature search of peer-reviewed journals revealed that Sprague Dawley rats were used in 36-month studies by (Voss et al., 2005) or in 24-month studies by (Hack et al., 1995), (Minardi et al., 2002), (Klimisch et al., 1997), (Gamez et al., 2007).Some of these studies have been published in Food and Chemical Toxicology.

　　Number of rats, OECD guidelines

　　实验动物数量与OECD实验规范

　　OECD 实验规范：第408条，关于90天实验，第452条关于慢性毒性试验，第453条关于综合致癌性/慢性毒性试验，都要求用20只动物为一组(1981和2009的规定都这样要求)，尽管可以用10只动物的实验就能取得生物化学参数。我们做的是长期毒性研究而不是致癌性研究，从一开始就不是这样设想的。根据常规10只动物一组已经足够在生物化学水平上进行研究，我们测量的参数数量是非常大的。

　　OECD guidelines (408 for 90 day study, 452 chronic toxicity and 453 combined carcinogenicity/chronic toxicity study) always asked for 20 animals per group (both in 1981 and 2009 guidelines) although the measurement of biochemical parameters can be performed on 10 rats, as indicated. We did not perform a carcinogenesis study, which would not have been adapted at first, but a long-term chronic full study, 10 rats are sufficient for that at a biochemical level according to norms and we have measured such a number of parameters!

　　在我们的实验中，性激素干扰的参数以及其它参数对于解释一年之后的严重后果是充分的。我们采用的OPLS-DA统计方法是最适宜的。关于肿瘤和动物死亡，时间效果以及每只动物的平均肿瘤数量都必须被纳入分析。在风险研究中出现的每一个迹象，都必须被充分重视。孟山都公司的研究用了同样的大鼠品系，每组仅10只衡量20个参数，就得出同一种NK603转基因玉米“安全”的结论，而且他们的实验只做了3个月 (Hammond et al., 2004)

　　The disturbance of sexual hormones or other parameters are sufficient in themselves in our case to interpret a serious effect after one year. The OPLS-DA statistical method we published is one of the best adapted. For tumours and deaths, the chronology and number of tumours per animal have to be taken into account. Any sign should be regarded as important for a real risk study. Monsanto itself measured only 10 rats of the same strain per group on 20 to conclude that the same GM maize was safe after 3 months (Hammond et al., 2004).

　　The statistical analysis should not be done with historical data first, the comparison is falsified, thus 50 rats per group is useless

　　统计分析不应该先做历史数据，用这个方法做比较研究是错误的，用每组50只动物做研究是无意义。

　　采纳历史数据会把健康风险评估变成研究造假，因为食谱中的材料已经受到化学污染(by dibenzo-p-dioxins and dibenzofurans (Schecter et al., 1996)和汞污染(Weiss et al., 2005)，镉污染，铬污染等，污染的程度足以改变动物肝脏和肺脏的基因表达，足以扰乱基因分析(Kozul et al., 2008)。以往的食料中还发现农药和增塑剂污染，污染来自箱笼或者水(Howdeshell et al., 2003)。历史数据也有来自可能食用了转基因的动物，很多地方的鼠粮中的确发现了转基因成分。这一切都与污染水平相关，我们已经在实验大鼠和对照组大鼠中检测到这些问题。

　　The use of historical data falsifies health risk assessments because the diet is contaminated by dibenzo-p-dioxins and dibenzofurans (Schecter et al., 1996), mercury (Weiss et al., 2005), cadmium and chromium among other heavy metals in a range of doses that altered mouse liver and lung gene expression and confounds genomic analyses (Kozul et al., 2008). They also contained pesticides or plasticizers released by cages or from water sources (Howdeshell et al., 2003). Historical data also come from rats potentially fed on GMOs, some animal pellets in the world do indicate that. All that corresponds to the contamination levels for which we have detected some effects in our treated rats versus appropriate controls.

　　在历史数据中，2年SD雌性大鼠罹患乳腺纤维瘤的为13%~62%(Giknis, 2004)，但在我们的实验中对照组的发病率要低得多，这才是真正的对照，而我们的实验鼠发病率比对照组高很多，这使得我们的研究结果有显著性。动物的死亡率也是这样。

　　2-year historical data mammary fibroadenoma rate from Charles River SD females ranged from 13 to 62% (Giknis, 2004). We obtain a lot less in our controls, the real comparators, a lot more in treated rats. This makes our results significant, like for deaths.