Skip to main content
European Commission logo print header

GMP Two Year Safety Testing

Final Report Summary - G-TWYST (GMP Two Year Safety Testing)

Executive Summary:
In the project period from April 2014 to April 2018 G-TwYST performed three rat feeding studies with the GM maize NK603 without and with a Roundup application: two 90-day trials for subchronic toxicity testing, one with GM maize inclusion rates of 11 and 33% and one with GM maize inclusion rates of 11, 33 and 50%, as well as a combined chronic toxicity/carcinogenicity study with GM maize inclusion rates of 11 and 33%, all three of them based on OECD Test Guidelines for the testing of single chemicals and EFSA recommendations. No treatment-related effects were observed after feeding rats with the GM maize NK603 or GM maize NK603 plus Roundup for up to 2 years.

The trial data were analysed by statistical equivalence tests (depicting whether the data were within the historic control data range of the animal testing facility that performed the feeding trials), a procedure recommended by the European Food Safety Authority to be used for the analyses of animal feeding trial outcomes. G-TwYST took into account this recommendation by including historical reference data from the animal testing facility at the Slovak Medical University in Bratislava, Slovakia, to define equivalence criteria. The test allowed to clearly visualise whether relevant changes between test groups occurred. The methodology was compared with established difference tests and the results of the difference and equivalence tests very well matched.

Based on the experiences from the project and the discussion in an expert symposium, nine quality criteria were suggested to be taken into account when evaluating a rodent feeding trial in the course of a pre-market approval procedure as well as when planning a feeding trial on whole food/feed derived from GM plants in the course of research projects not related to a pre-market approval procedure.

The analysis of the plant material revealed no major deviations in nutrients between diets within each production batch analysed. Moreover, the low-level presence of glyphosate residues in all diets appears to be accounted for by contributions from other ingredients, particularly cereals, and does not affect the safety of the diets. Omics technologies may increase in the future the chances of identifying potential unintended impacts of transgenesis that are relevant for food, feed or environmental risk assessment. Further discussion is needed to decide whether omics analyses should be part of future risk assessment processes.

The project also explored the normative dimensions of controversies on animal feeding studies and how they relate to scientific aspects. In a controversy, science can become a strategic tool for stakeholders motivated by political and economic interests. In this context, regulatory scientists should avoid generalizations and clearly communicate, for example, the differences between regulatory and academic science norms or the existence of safety within politically established limits.

Project Context and Objectives:
The scope and general objectives of this research project were specified in the call FPE-KBBE-2013-FEEDTRIALS as follows:
“ The EU legal framework on GMOs ensures that genetically modified (GM) food and feed are when placed on the market safe as regards human and animal health and the environment. The European Commission has very recently adopted a Regulation concerning applications for the authorisation of GM food and feed, requesting applicants to carry out an obligatory 90-day feeding study with whole food/feed for each submitted GMO dossier. Depending on the outcome of previous studies, a two-year carcinogenicity study with rats may also be requested by the European Food Safety Authority (EFSA) on a case-by-case basis. EFSA has been requested to assist the Commission to provide supplementary guidance on key elements to consider for a 2-year carcinogenicity trial in rats with whole food/feed. In this context, and to address possible concerns, including those raised recently by a study on "Long-term toxicity of a Roundup herbicide and a Roundup-tolerant genetically modified maize" the following issues related to two-year feeding trials need to be addressed in a stepwise approach:

1. Execution of at least one rat feeding trial(s) with GM maize NK 603 (and additional GMOs where scientifically justified) applying the EFSA protocol. Participating institutions should strictly comply with all applicable international standards and norms concerning feeding trials in close collaboration with EFSA.
2. Analysing, reporting and providing recommendations, in particular as to the scientific justification and added value of such long-term feeding trials with regard to GMO risk assessment.”

In order to achieve these objectives, G-TwYST
• performed three rat feeding studies with the GM maize NK603 without and with a Roundup application: two 90-day trials for subchronic toxicity testing, one with GM maize inclusion rates of 11 and 33% and one with GM maize inclusion rates of 11, 33 and 50% as well as a combined chronic toxicity/carcinogenicity study with GM maize inclusion rates of 11 and 33%, all three of them based on OECD Test Guidelines for the testing of single chemicals and EFSA recommendations;
• reviewed recent and ongoing research relevant to the scope of G-TwYST;
• engaged with related research projects such as GRACE and GMO90+;
• developed criteria to evaluate the scientific quality of rodent feeding trials with whole food/feed;
• developed recommendations together with the research projects GRACE on the added value of long-term feeding trials in the context of the GM-plant risk assessment process (forthcoming);
• as a complementary activity, investigated into broader societal issues linked to the controversy on animal studies with whole food/feed in GMO risk assessment;
• allowed for stakeholder engagement in all key steps of the project in an inclusive and responsive manner;
• ensured utmost transparency of what was done, why it was done, and by whom it was done.

Project Results:
Feeding studies
G-TwYST performed three rat feeding studies with GM maize NK603 without and with a Roundup application: two 90-day trials for subchronic toxicity testing, one with GM maize inclusion rates of 11 and 33% and one with GM maize inclusion rates of 11, 33 and 50%, as well as a combined chronic toxicity/carcinogenicity study with GM maize inclusion rates of 11 and 33%, all three of them based on OECD Test Guidelines for the testing of single chemicals and EFSA recommendations. The trial with the increased inclusion rate of 50% maize in the diets was taken up following the discussion at the first stakeholder workshop (planning phase of the project activities). The diets were calculated in such a way that they were isocaloric and isoproteic; the diets containing 50% maize were modified regarding the other ingredients to obtain nutritionally balanced diets.

The study design of the 90-day feeding trials was based on the OECD Test Guideline 408 and EFSA recommendations on the performance of 90-day rodent feeding trials with whole food feed. Five test groups were established for each gender: two high-dose NK603 test groups with and without maize from field trials with and without Roundup applications, two low-dose test groups and the control group with the conventional counterpart variety. The rat body weight as well as the feed consumption during the 90-day feeding period was similar in all five experimental groups. The data on the clinical biochemical parameters and the relative organ weights underwent difference and equivalence testing. The statistical analyses revealed 25 significant differences (8% of 320 tests) and 301 significant equivalences (94% of 320 tests). In the case of the equivalence tests, the percentage lymphocytes (males/females) as well as for the growth rate, the hematocrit, the percentage eosinophils as well as the kidney and uterus weight (females only) were outside the equivalence interval. Taken together, no toxicologically relevant differences between the control group and the NK603 groups were observed. There were no relevant histopathology findings for the liver, kidneys, uterus, ovaries and adrenal cortices. Moreover, no histopathological alterations in the adrenal medullas, thyroid glands and mammary glands were observed. In summary, there were no treatment-related necropsy or histopathological findings following the administration of genetically modified NK603 maize or genetically modified NK603 maize plus Roundup to rats for 90 days.

The study design of the 90-day feeding trials with 50% inclusion rate of maize comprised eight test groups for each gender: two high-dose (50%) NK603 test groups with and without maize from field trials with and without Roundup applications, two low-dose (11%) test groups and the control group with the conventional counterpart variety (50%) and three test groups with a 33% inclusion rate to enable a comparison with the previous trial. The body weight as well as the feed consumption of male rats during the 90-day feeding period was similar in all eight experimental groups. The body weight as well as the feed consumption of female rats during the 90-day feeding period was similar in seven out of eight experimental groups, while the body weight and the feed consumption of female rats fed the GM maize NK603 treated with Roundup at an inclusion rate of 11% were higher than in the other seven experimental groups, including the one with an inclusion rate of 50%. The data on the clinical biochemical parameters and the relative organ weights underwent difference and equivalence testing. The statistical analyses revealed 46 significant differences (7% of 648 tests) and 643 significant equivalences (99% of 648 tests). In the case of the equivalence tests, the blood cholesterol and phosphorus levels as well as the relative kidney weight (females only) were outside the equivalence interval. Taken together, no toxicologically relevant differences between the control group and the NK603 groups were observed. The histopathology findings for the liver, kidneys, uterus, ovaries and vagina were toxicologically not relevant. Moreover, no histopathological alterations in the mammary glands were observed. In summary, there were no treatment-related necropsy or histopathological findings following the feeding of rats with genetically modified NK603 maize or genetically modified NK603 maize plus Roundup up to an inclusion rate of 50% in the diet for 90 days.

The combined chronic toxicity/carcinogenicity feeding trial with rats being fed the genetically modified (GM) maize NK603 cultivated in the absence or presence of Roundup at an inclusion rate of 11 and 33% in the diet were performed based on the OECD Test Guideline 453 and EFSA considerations on the applicability of OECD TG 453 to whole food/feed. Five test groups were established for each gender: two high-dose NK603 test groups with and without maize from field trials with and without Roundup applications, two low-dose test groups and the control group with the conventional counterpart variety.

Five NK603-fed groups had lower, one NK603-fed group had equal and two NK603-fed groups showed a higher mortality rate than the control group. The mortality rate of the male rats fed the NK603 maize at an inclusion rate of 33% in the diet was significantly higher than that of the corresponding control group. In contrast, the female rats fed the NK603 maize at an inclusion rate of 33% in the diet showed a much lower mortality rate than the corresponding control group.
The male rats fed the NK603 maize at an inclusion rate of 33% showed a significantly higher body weight when compared to the corresponding control group in the second half of the experimental period, while the body weight of the female rats was similar in all five experimental groups. The mean feed consumption of male and female rats surviving until the end of the 2-year feeding period was similar to the respective control groups. The data on the clinical biochemical parameters and the relative organ weights underwent difference and equivalence testing. Equivalence testing was only performed with the 3- and 6-month data, since no historical control data from the preceding GRACE project were available for the 12- and 24-month points in time. Ninety-five percent of the performed comparisons (i.e. 433 out of 456 tests) showed equivalences at 3 and 6 months. In the case of the equivalence tests, the mean red blood cell volume, the mean corpuscular hemoglobin and the percentage lymphocytes in males rats as well as the mean corpuscular hemoglobin concentration, the percentage monocytes and the cholesterol and phosphorus levels in female rats were outside the equivalence interval. Taken together, no toxicologically relevant differences between the control group and the NK603 groups were observed.

In all groups including controls, the most common cause of death was a pituitary pars anterior adenoma. The next most common was kidney chronic progressive nephropathy. The increased mortality in the male rats fed the NK603 maize treated with Roundup was statistically significant (see above), and was related to the increased number of deaths from pituitary neoplasia; however, there was no effect on the overall incidence of pituitary neoplasia. This increased mortality was considered to be due to the increased feed consumption in the male rats as frequently reported in the scientific literature and, hence, is considered unrelated to the test items NK603 and Roundup.

The necropsy findings in the chronic toxicity and in the carcinogenicity phase recorded in this study were considered to be within the normal range of background alterations seen in untreated animals of this age and strain. The non-neoplastic and neoplastic microscopic findings noted in the treated animals were considered incidental changes because they were consistent with spontaneously occurring findings described in the literature, the findings were distributed randomly among groups, or their appearance was similar to findings found in controls.

After examination of all female groups, the observed increase in the number of benign thymomas in the female group fed the GM maize NK603 treated with Roundup at an inclusion rate of 33% was not statistically significant when compared to the control group, or when benign and malignant thymomas were analyzed in combination. It was therefore considered an incidental finding. Moreover, this was also the case of pituitary gland and mammary gland tumours.
In conclusion, there were no treatment-related mortality, necropsy or histopathology findings following the administration of GM maize NK603 or GM maize NK603 plus Roundup to rats for up to 2 years.

Statistical methods
EFSA introduced in its guidance to the performance of 90-day feeding trials with whole food/feed the statistical approach of equivalence testing. This statistical methodology allows to define a “normal” range of values for variables recorded in a feeding trial. Moreover, it simplifies communication on test results as changes between the test groups became visually obvious even for a set of variables with heterogeneous dimensions. In the course of the EU-funded project GRACE the methodology was practically established. In the course of G-TwYST, it was further refined by incorporating historical control data from the animal testing lab to define equivalence limits. The reference data were provided by the GRACE project as freely available on the CADIMA website (www.cadima.info). G-TwYST partners coauthored a comparative paper on the GRACE data (open access; see Schmidt, K. et al. (2016) doi:10.1007/s00204-016-1857-x) describing the methodology. The improved statistical equivalence test design was further applied to the data generated in G-TwYST and compared to statistical difference tests routinely used to analyse feeding trials. Apart from consistency between the two methodologies, it became obvious that the graphical presentation of equivalence test results is easier to be understood e.g. by non-experts.

Quality criteria
G-TwYST reviewed and discussed initial 31 different aspects to be taken into account when evaluating the scientific quality of publications on feeding trials with genetically modified plants as feed constituents in rats. The criteria were further presented and discussed in an international expert symposium with 29 participants from G-TwYST and extern experts including the G-TwYST Advisory Board members. The Symposium included three sessions with the following topics: 1) Presentation and discussion of the proposed criteria; 2) Concepts in statistical methods applied to analyze feeding trials with whole food/feed derived from genetically modified plants as an example; 3) When does a statistically significant difference between two experimental groups become toxicologically relevant?
In the frame of the above-mentioned symposium, a concluding set of nine quality criteria was defined. It is important to note that the described set of key quality criteria applies to 90-day feeding trials as well as to long-term feeding trials such as those performed to test the potential chronic toxicity and/or carcinogenicity of whole food/feed with a duration of 12 to 24 months with the aim of providing data for the risk assessment in applications for the market release of a GM crop. The criteria are:
1. The design of the feeding trial is based on internationally recognized test guidelines, but adapted for specific needs of whole food/feed studies and non-targeted testing.
2. An analysis of the plant materials and diets including, among others, macro- and micronutrients, biological and chemical contaminants as well as the identification and quantification of the event, is performed.
3. The highest level of the plant material that can be incorporated in the animal diets without leading to a nutritional imbalance is tested.
4. A non-GM line with a comparable genetic background is used as a control.
5. Specific aspects regarding the choice and housing of the laboratory animals used in the feeding trials are considered.
6. Appropriate randomization techniques are applied.
7. A reliable and appropriate sample collection and processing strategy is implemented.
8. The staff performing the feeding trial and the analysis of the plant materials, diets and animal samples is “blind” with respect to the identity of the diets.
9. Appropriate statistical methods are applied to evaluate the power of the study and to analyze the obtained results.

Analyses of plant material
G-TwYST analysed the compounds of the plant material (maize kernels) and the diets to allow a quality control of the test material. Targeted analyses were performed on nutrient compounds such as vitamins, fatty acids, and amino acids, non-nutrients (e.g. trypsin inhibitors) and potential contaminants like heavy metals, pesticides and mycotoxins, since the basic feed quality might interfere with the NK603 effects elicited in the feeding trials. In addition, the material was examined for the presence of GMOs other than the intended NK603 event. Analyses of the plant and diet DNA confirmed that GM maize NK603 was indeed present in the diets in which it was expected. Whereas non-quantifiable traces of other GM crops were found in the diets, this is a common phenomenon in animal feeds and is not considered to impact on study outcomes. No major deviations in nutrients were found between diets within each production batch analysed. In addition, there was general consistency amongst batches, whilst the variability occasionally observed across groups in the case of secondary metabolites and vitamins can be accounted for by external factors (e.g. different analytical methods used by different laboratories). The low-level presence of glyphosate residues in all diets, whilst being low in maize on which the glyphosate herbicide was applied, appears to be accounted for by contributions from other ingredients, particularly cereals. Once again, the safety of the diets was not affected.

G-TwYST performed untargeted transcriptomic, proteomic and metabolomic analyses of plant materials (maize kernels/embryos). Different genetic backgrounds and growing conditions were taken into account by analysing NK603 varieties from Canada and USA and their conventional counterparts with omics technologies. Omics analyses provide comprehensive data on gene expression, protein and metabolite parameters, and may increase the chances of identifying potential unintended impacts of transgenesis that are relevant for the food, feed or environmental risk assessment. Omics analyses are not part of current risk assessment procedures. Therefore, in the frame of G-TwYST, only a pilot study with a limited number of samples was performed to explore the potential of the omics analyses. For now, all three omics strategies were applied; however, at the moment, transcriptomics still has the largest relative coverage. Further discussion would be needed regarding appropriate sample sizes if omics analyses were to be considered as part of future risk assessment processes. The experimental results were analysed in two ways: (i) by the direct comparison of the GM versus the non-GM materials, in line with the targeted compositional analysis that is currently part of the standard comparative compositional analysis, and (ii) using the Soft Independent Modelling of Class Analogy (SIMCA) one-class model approach. In the latter approach, the omics profiles of the GM maize varieties, the conventional counterparts and other maize varieties considered as safe were analysed in order to detect aberrant profiles, if any, rather than focusing on individual components. Based on the analyses included in the omics profiles, no indications have been obtained for changes in the physiology of the NK603 GM plant varieties, treated or not treated with the herbicide, that warrant further investigations. The present results obtained in the G-TwYST project are in agreement with those found in the previous GRACE project on the insect-resistant MON810 GM event.

In general terms, various key observations were made:
− The one-class model is able to tackle the large datasets usually obtained with omics and makes the outcomes tangible.
− No major impact of the genetic modification was observed in preliminary analyses, whereas there was an impact regarding variety, location and year.
− It proved a challenge to combine data from different field trials and generated with different techniques.
Based on these findings, it is recommended that in the frame of follow-up studies the design of field trials should be modified in such a way that it allows for correction (e.g. inclusion of varieties overlapping between trials).

Analysis of the normative dimensions and science communication
In order to gain a better understanding of the issues and the dynamics of the disputes on animal feeding studies and the possible interrelation between results from scientific experiments and values, world views or interests, G-TwYST explored two hypotheses developed following a literature review and with the help of extern advisors in an expert workshop: 1. Values are part of the regulatory science assessments performed as a response to controversial studies. 2. Actors engaged in the controversy often use, select, interpret and translate science in tactical strategic ways to try to serve their predetermined interests (two groups of actors: NGOs and industry).

Three representative case studies were selected as a basis for the analysis of the two hypotheses: 1) Seralini´s two year feeding study in rats fed GM maize, 2012; 2) Ramazzini Foundation study on the carcinogenicity of aspartame, 2005; 3) Stump study on functional and/or morphological effects in the offspring nervous system caused by BPA maternal exposure, 2010. Interviews were undertaken with various stakeholders directly involved in the events described in all three case studies. The highest number of interviews could be made in the GM maize case study. In the other two case studies, the contacted stakeholders were less responsive or willing to be interviewed.

The analysis identified that value-based decisions are embedded in regulatory science when producing, interpreting and reviewing scientific evidence. Yet the presence of values in regulatory science (apparent in science when compromises between competing views and interests are negotiated/achieved) are generally not acknowledged. The analysis further finds that conflicting values in risk assessment play a central role in regulatory science controversies. As a consequence of the presence of values, a full separation of science and policy is unrealistic both in practice and norms. Furthermore, during a controversy, science can become a strategic tool for stakeholders motivated by political and economic interests. Based on these findings, further policy recommendations were made to help address future controversies: (i) clear guidance on the scientific quality level required for a scientific study to be relevant for risk assessment should be devised; (ii) a lack of formal acknowledgement of different norms between regulatory and academic research contributes to the continuation of controversies in the public domain, as the public cannot easily assess the information presented in scientific studies; (iii) studies, which do not clearly accept or reject it, create confusion and ground for conspiracy theories. In case of challenges to the quality criteria/risk assessment, a structured, fair and transparent dialogue with the scientists making the challenges should be established.

Stakeholder involvement
G-TwYST implemented an RRI approach of stakeholder involvement not only for discussing project results but also in the research planning (upstream engagement). At the beginning of the project, some 700 stakeholders representing Competent Authorities, both at the EU and the national level of all 28 Member States, industry, farming and professional organisations, civil society organisations (CSOs) and scientists have been contacted. The planning stage workshop presented the detailed (technical) planning of the project. Forty-one stakeholder representatives from 14 Member States, USA and Norway participated: 22 represented Competent Authorities, 9 Industries, 2 CSO and 8 Academia. Based on the discussions, the feeding trials were modified in such a way that instead of a 2-year study with MON810 maize a subchronic toxicity study with an increased NK603 maize inclusion rate (50%) was performed. Moreover, a regular subchronic study and a combined chronic toxicity/carcinogenicity study with a 33% NK603 maize inclusion rate were performed. The stakeholder consultation at the result stage was split in a written consultation for the results of the first 90-day feeding study with a 33% inclusion rate and a workshop in the final month of the project presenting the results of the 90-day feeding study with a 50% inclusion rate and the combined chronic toxicity/carcinogenicity study. Thirty-five stakeholders signed a non-disclosure agreement and got access to the detailed research data. Seventeen written comments were received and answered by the project team and 29 stakeholders representatives of Competent Authorities, industry, civil society organisations and research organisations attended the workshop.
Though these detailed consultations were actually laborious for the project partners and the stakeholders because of the technical details provided and discussed, it was positively acknowledged, since the project performance was thereby updated and improved. Project team members experienced stakeholder contributions as helpful to improve the project both in terms of type of studies conducted, interpreting results, and clarity of plans and results. Challenges experienced include the limited flexibility in EU-funded projects to accommodate stakeholder suggestions resulting in modification of work plans and not enough time for an in-depth dialogue, particularly at the end of the project.

Potential Impact:
Regulatory impact, risk assessment and political dimensions
The legal framework for the risk assessment of GMOs in Europe is harmonized based on EC Directives and Regulations. In 2013, the Commission adopted the Implementing Regulation (EU) No 503/2013, in which it is specified that a 90-day feeding study with whole food/feed for each GMO event must have been performed to apply for the authorisation of GM food and feed. Nevertheless, a common understanding of the design, interpretation, scientific justification and added value of long-term feeding trials for GM food/feed risk assessment has not been achieved among different stakeholder groups and among Member States. One key objective of the call for proposals FPE-KBBE-2013-FEEDTRIALS was to explore the scientific justification of long-term feeding studies. The broad controversy on the necessity was fuelled by concerns regarding possible health impacts of GM maize, including those raised by a study on the "Long-term toxicity of a Roundup herbicide and a Roundup-tolerant genetically modified maize" published by G. Séralini and co-workers in 2012 (retraction by the journal, because of the criticised lack of quality, and its republishing in 2014).

Against this background, the G-TwYST project derived concrete recommendations as to the study design, data interpretation, quality assessment, scientific justification and added value of such long-term feeding trials with regard to the GMO risk assessment by performing feeding trials of different duration based on international accepted standards and norms.

The detailed data generated by G-TwYST will become freely accessible after peer review. Moreover, they can be combined with coherent existing data set provided by the EU-funded project GRACE (on MON810 maize) as well as the French project GMO90+ (on the same test material). This broad database provides detailed trackable information not only regarding the outcome of the studies but also regarding its performance. At the technical level, the project addressed the need for backing data for risk assessors and scientists to reconsider technical issues as well as the added value of animal feeding trials with whole feed e.g. at EFSA. Furthermore, this enables an outmost transparent discussion in the context of policy and decision making. Criteria to evaluate the scientific quality of 90-day and extended feeding trials on whole food/feed derived from genetically modified plants have been proposed in a publication by Schmidt et al. (2016).
Moreover, the results showed that that non-significant equivalences and significant differences occurred in no more than about 5% of cases across 1,424 equivalence tests and 3,472 difference tests performed in the trials, which is the expected percentage for statistical tests using a 5% error level. Moreover, the histopathological examinations did not reveal a negative impact by the GM maize. Therefore, the project concluded that there were no adverse effects related to the feeding of rats with the GM maize NK603 cultivated with or without Roundup.

Hence, the studies performed - neither the subchronic nor the combined chronic/carcinogenicity studies - did alter the existing risk assessment conducted in the course of the application for market release. This may strengthen the confidence in the regulatory risk assessment process.

The statistical analyses also showed that many parameters recorded in the course of such OECD Test Guideline-based feeding trials may vary considerably so that the doubling of the size of the test groups will hardly increase the sensitivity of the trial. Therefore, the project stressed the necessity to perform whole food/feed studies with clearly targeted hypotheses rather than use such studies for an untargeted screening of unanticipated toxicological effects. The lack of sensitivity also limits the use of 90-day feeding studies with whole feed to trigger the initiation of extended studies. In the Implementing Regulation (EU) No 503/2013, which mandatorily requests a 90-day feeding study for the authorisation of GM food and feed, also foresees a review of this obligation. The data provided by G-TwYST as well as by the other projects should be reconsidered when reviewing the above-mentioned regulation.

The considerable number of animals sacrificed in the course of such (extended) toxicity studies (720 in the 2 year study) and the before mentioned limited sensitivity of the tests collide with the three R approach and the aim to reduce animal trials promoted by the EC ( Directive 2010/63/EU "on the protection of animals used for scientific purposes). An animal welfare CSO frquently raised this issue during the stakeholder consultations.

Stakeholder involvement and science communication
G-TwYST implemented an RRI approach of stakeholder involvement not only to discuss project results but also in the research planning phase (upstream engagement). At the beginning of the project some 700 stakeholders representing Competent Authorities, both at the EU and the national level of all 28 Member States, industry, farming and professional organisations, civil society organisations (CSOs) and scientists were contacted. The planning stage workshop presented the detailed (technical) planning of the project. Forty-one stakeholder representatives from 14 Member States, USA and Norway participated. The stakeholder consultation at the result stage was split in a written consultation for the results of the first 90-day feeding study with a 33% GM maize inclusion rate and a workshop in the final month of the project presenting the results of the 90-day feeding study with a 50% GM maize inclusion rate and the combined chronic toxicity/carcinogenicity study. Thirty-five stakeholders signed a non-disclosure agreement and got access to the detailed research data. Seventeen written comments were received and answered by the project team and 29 stakeholders representatives. The stakeholder attending the workshops and consultations represented Competent Authorities, industry, civil society organisations and academia. Though these detailed consultations were actually laborious for the project partners and the stakeholders because of the technical details provided and discussed, it was broadly acknowledged, as the project performance was scrutinized, updated and improved up to a certain degree. Nevertheless, without a flexible timeframe for the project - a prolongation was not granted by the Commission – the project faced severe time pressure towards the end. This situation also limited the intensity of stakeholder involvement.

The project also explored the normative dimensions of controversies on animal feeding studies and how they relate to the scientific aspects. In a controversy, science can become a strategic tool for stakeholders motivated by political and economic interests. This is possible because value-based decisions are embedded in regulatory science when producing, interpreting and reviewing scientific evidence. Regulatory scientists should thus avoid sweeping generalizations and acknowledge values and limitations, as for example the differences between regulatory and academic science norms or the existence of safety within politically established limits. By claiming value-free and absolute risk assessment, risk assessors are put into an undefendable position. Acknowledgement of values and limitations in risk assessment could also potentially discourage the use of science as a strategic tool by different actors.
It is further concluded that clear guidance on the scientific quality level required for a scientific study to be relevant for risk assessment should be defined. For example, required validation for novel mechanisms of action could be clearly stated. Lack of formal acknowledgement of different norms between regulatory and academic research contributes to the continuation of controversies in the public domain, as the public cannot easily assess the information presented in scientific studies. When a study is not clearly accepted or rejected, it creates confusion and a good ground for conspiracy theories. In case of challenges to the quality criteria/risk assessment, a structured, fair and transparent dialogue with the scientists making the challenges should be established, the aim being a “forensic” investigation of the source and relevance of diverging science.

Scientific advancement and cooperation
As previously stated, a common understanding of the design, interpretation, scientific justification and added value of long-term feeding trials for GM food/feed risk assessment has not been achieved among expert and decision makers. The cooperation between the EU-funded projects G-TwYST, GRACE and the French project GMO90+ provides a unique data set based on equal test material (NK603 and MON810 maize) while at the same time overviewing different test durations and alternative and extended methods. Moreover, they gained a consistent set of results. Nevertheless, the open access will enable scientists to revisit and reanalyse the data.

A key question was the discussion on the quality of previously published toxicological studies with GM maize and, therefore, the project performed the trials by taking into account the OECD Test Guidelines as well as EFSA recommendations. Based on the experience gained in the course of the project and an additional expert workshop, a set of criteria for the evaluation of the scientific quality of mandatory rat and mouse feeding trials with whole food/feed derived from genetically modified plants was published in 2016. These criteria should be taken into account when evaluating a rodent feeding trial in the course of a pre-market approval procedure. Including an animal feeding study that does not (fully) comply with the proposed quality criteria in a risk assessment should be decided on a case-by-case basis.
The project practically and in detail adopted the EFSA recommendation to apply statistical equivalence tests (to determine whether the observed changes were within a “normal” range of safe parameter values) when analysing toxicological studies with rodents and whole GM-feed. The statistical method was further improved during the project to enable a quick judgement of many different test variable, i.e. to determine whether relevant changes between test groups occurred. The approach was compared with “classical“ statistical analyses and similar results were obtained. The equivalence test methods can be applied to any toxicological study to obtain a visual presentation of the (complex) results that is easily understood e.g. by non-experts. The methodology has been published in the meantime (open access publication).

The project also explored the application of omics approaches to evaluate unintended changes in plant material (maize kernels) in an untargeted approach. Statistical one-class models were used to support a decision on whether adverse changes might have occurred. For now, all three omics strategies (transcriptomics, proteomics and metabolomics) were applied; however, at the moment, transcriptomics still has the largest relative coverage. Further discussions will be needed regarding the appropriate test design and sample sizes if omics analyses will become part of future risk assessment processes.

Other dissemination activities
The project cooperated with the EU-funded project GRACE and the French project GMO90+ to provide a joint data pool from rat feeding trials with whole feed including NK603 and MON810 maize of different extentions and complementary approaches. It is jointly intended to freely share the detailed feeding trial data through the repositorium provided by the website www.cadima.info which is permanently hosted by JKI. Currently, the data generated by GRACE are accessible; the data from G-TwyST and GMO90+ will be added and become available by open access after peer review and publication.
A set of statistical papers and one on quality criteria evaluation of rodent feeding trials has already been published open access.
Results and activities were presented at international symposia on biosafety research in South Africa and Mexico (ISBGMO 13 and 14).
Conclusions and recommendations as well as reports from the project and related information is shared through the project website www.g-twyst.eu.
Press announcement and media invitations were circulated and journalists attended the final conference.
The joint key recommendations of GRACE and G-TwYST were published as policy briefs.

List of Websites:
www.g-twyst.eu
www.cadima.info
info@g-twyst.eu
ralf.wilhelm@julius-kuehn.de