Safety and nutritional assessment of GM plants and derived food and feed: The role of animal feeding trials Report of the EFSA GMO Panel Working Group on Animal Feeding Trials

In this report the various elements of the safety and nutritional assessment procedure for genetically modified (GM) plant derived food and feed are discussed, in particular the potential and limitations of animal feeding trials for the safety and nutritional testing of whole GM food and feed. The general principles for the risk assessment of GM plants and derived food and feed are followed, as described in the EFSA guidance document of the EFSA Scientific Panel on Genetically Modified Organisms. In Section 1 the mandate, scope and general principles for risk assessment of GM plant derived food and feed are discussed. Products under consideration are food and feed derived from GM plants, such as maize, soybeans, oilseed rape and cotton, modified through the introduction of one or more genes coding for agronomic input traits like herbicide tolerance and/or insect resistance. Furthermore GM plant derived food and feed, which have been obtained through extensive genetic modifications targeted at specific alterations of metabolic pathways leading to improved nutritional and/or health characteristics, such as rice containing b-carotene, soybeans with enhanced oleic acid content, or tomato with increased concentration of flavonoids, are considered. The safety assessment of GM plants and derived food and feed follows a comparative approach, i.e. the food and feed are compared with their non-GM counterparts in order to identify intended and unintended (unexpected) differences which subsequently are assessed with respect to their potential impact on the environment, safety for humans and animals, and nutritional quality. Key elements of the assessment procedure are the molecular, compositional, phenotypic and agronomic analysis in order to identify similarities and differences between the GM plant and its near isogenic counterpart. The safety assessment is focussed on (i) the presence and characteristics of newly expressed proteins and other new constituents and possible changes in the level of natural constituents beyond normal variation, and on the characteristics of the GM food and feed, and (ii) the possible occurrence of unintended (unexpected) effects in GM plants due to genetic modification. In order to identify these effects a comparative phenotypic and molecular analysis of the GM plant and its near isogenic counterpart is carried out, in parallel with a targeted analysis of single specific compounds, which represent important metabolic pathways in the plant like macro and micro nutrients, known anti-nutrients and toxins. Significant differences may be indicative of the occurrence of unintended effects, which require further investigation. Section 2 provides an overview of studies performed for the safety and nutritional assessment of whole food and feed. Extensive experience has been built up in recent decades from the safety and nutritional testing in animals of irradiated foods, novel foods and fruit and vegetables. These approaches are also relevant for the safety and nutritional testing of whole GM food and feed. Many feeding trials have been reported in which GM foods like maize, potatoes, rice, soybeans and tomatoes have been fed to rats or mice for prolonged periods, and parameters such as body weight, feed consumption, blood chemistry, organ weights, histopathology etc have been measured. The food and feed under investigation were derived from GM plants with improved agronomic characteristics like herbicide tolerance and/or insect resistance. The majority of these experiments did not indicate clinical effects or histopathological abnormalities in organs or tissues of exposed animals. In some cases adverse effects were noted, which were difficult to interpret due to shortcomings in the studies. Many studies have also been carried out with feed derived from GM plants with agronomic input traits in target animal species to assess the nutritive value of the feed and their performance potential. Studies in sheep, pigs, broilers, lactating dairy cows, and fish, comparing the in vivo bioavailability of nutrients from a range of GM plants with their near isogenic counterpart and commercial varieties, showed that they were comparable with those for near isogenic non-GM lines and commercial varieties. In Section 3 toxicological in vivo, in silico, and in vitro test methods are discussed which may be applied for the safety and nutritional assessment of specific compounds present in food and feed or of whole food and feed derived from GM plants. Moreover the purpose, potential and limitations of the 90-day rodent feeding trial for the safety and nutritional testing of whole food and feed have been examined. Methods for single and repeated dose toxicity testing, reproductive and developmental toxicity testing and immunotoxicity testing, as described in OECD guideline tests for single well-defined chemicals are discussed and considered to be adequate for the safety testing of single substances including new products in GM food and feed. Various in silico and in vitro methods may contribute to the safety assessment of GM plant derived food and feed and components thereof, like (i) in silico searches for sequence homology and/or structural similarity of novel proteins or their degradation products to known toxic or allergenic proteins, (ii) simulated gastric and intestinal fluids in order to study the digestive stability of newly expressed proteins and in vitro systems for analysis of the stability of the novel protein under heat or other processing conditions, and (iii) in vitro genotoxicity test methods that screen for point mutations, chromosomal aberrations and DNA damage/repair. The current performance of the safety assessment of whole foods is mainly based on the protocols for low-molecular-weight chemicals such as pharmaceuticals, industrial chemicals, pesticides, food additives and contaminants. However without adaptation, these protocols have limitations for testing of whole food and feed. This primarily results from the fact that defined single substances can be dosed to laboratory animals at very large multiples of the expected human exposure, thus giving a large margin of safety. In contrast foodstuffs are bulky, lead to satiation and can only be included in the diet at much lower multiples of expected human intakes. When testing whole foods, the possible highest concentration of the GM food and feed in the laboratory animal diet may be limited because of nutritional imbalance of the diet, or by the presence of compounds with a known toxicological profile. The aim of the 90-days rodent feeding study with the whole GM food and feed is to assess potential unintended effects of toxicological and/or nutritional relevance and to establish whether the GM food and feed is as safe and nutritious as its traditional comparator rather than determining qualitative and quantitative intrinsic toxicity of defined food constituents. The design of the study should be adapted from the OECD 90-day rodent toxicity study. The precise study design has to take into account the nature of the food and feed and the characteristics of the new trait(s) and their intended role in the GM food and feed. A 90-day animal feeding trial has a large capacity (sensitivity and specificity) to detect potential toxicological effects of single well defined compounds. This can be concluded from data reported on the toxicology of a wide range of industrial chemicals, pharmaceuticals, food substances, environmental, and agricultural chemicals. It is possible to model the sensitivity of the rat subchronic feeding S4 EFSA GMO Panel / Food and Chemical Toxicology 46 (2008) S2–S70 study for the detection of hypothetically increased amount of compounds such as anti-nutrients, toxicants or secondary metabolites. With respect to the detection of potential unintended effects in whole GM food and feed, it is unlikely that substances present in small amounts and with a low toxic potential will result in any observable (unintended) effects in a 90-day rodent feeding study, as they would be below the no-observed-effect-level and thus of unlikely impact to human health at normal intake levels. Laboratory animal feeding studies of 90-days duration appear to be sufficient to pick up adverse effects of diverse compounds that would also give adverse effects after chronic exposure. This conclusion is based on literature data from studies investigating whether toxicological effects are adequately identified in 3-month subchronic studies in rodents, by comparing findings at 3 and 24 months for a range of different chemicals. The 90-day rodent feeding study is not designed to detect effects on reproduction or development other than effects on adult reproductive organ weights and histopathology. Analyses of available data indicate that, for a wide range of substances, reproductive and developmental effects are not potentially more sensitive endpoints than those examined in subchronic toxicity tests. Should there be structural alerts for reproductive/developmental effects or other indications from data available on a GM food and feed, then these tests should be considered. By relating the estimated daily intake, or theoretical maximum daily intake per capita for a given whole food (or the sum of its individual commercial constituents) to that consumed on average per rat per day in the subchronic 90-day feeding study, it is possible to establish the margin of exposure (safety margin) for consumers. Results obtained from testing GM food and feed in rodents indicate that large (at least 100-fold) ‘safety’ margins exist between animal exposure levels without observed adverse effects and estimated human daily intake. Results of feeding studies with feed derived from GM plants with improved agronomic properties, carried out in a wide range of livestock species, are discussed. The studies did not show any biologically relevant differences in the parameters tested between control and test animals. The studies have shown that targeted compositional analysis is the cornerstone for the safety assessment of GM plants modified for agronomic input traits, and once compositional equivalence has been established, feeding studies with livestock species add little to their safety assessment. Examples of models for livestock feeding studies with GM plants with increased concentration of desirable nutrients are provided. Such studies should be conducted on a case-by-case basis to establish the nutritional benefits. Possible effects of the new feed resource on animal performance, animal health, efficacy, and acceptability of the new feed ingredient should be investigated, and time spans for such studies should be determined on a case-by-case basis. The feasibility and limitations of human studies with foods derived from GM plants are discussed, as well as the potential and limitations of post-market monitoring to detect unintended effects of these foods. Post-market monitoring is not a substitute for a thorough pre-market risk assessment. In Section 4 standards for test sample preparation, test materials, diet formulation and analysis are evaluated. Specific attention is paid to the choice of control diets and comparators, dietary stability, and nutritional balancing of diets. When testing whole foods, it is desirable to obtain the highest concentration possible of the GM food and feed in the laboratory animal diet without causing nutritional imbalance. Normal practice is to use a minimum of two test dose levels and negative control with which to create nutritionally equivalent balanced diets in a comparative protocol. It is recommended to include a relevant number of commercial varieties as control diets to demonstrate the biological range of the parameters which are measured in order to assess the biological relevance of statistically significant differences between the GM plant and its counterpart. The choice of the comparator for GM food and feed testing is crucial, and can be found in the parental (near isogenic) line. For modified macronutrients a comparator is the unmodified form of the macronutrient. For investigating GM food and feed with enhanced nutritional properties, choices for control diets should be made on a case-by-case basis. Section 5 provides information on the collection, analysis and interpretation of data and findings obtained from animal feeding studies. Data generation for the prediction of safety and nutritional value of GM plant derived food and feed must be of high quality in order to perform a proper hazard identification and risk assessment. This should be based on the use of standardised study designs conducted to the principles of Good Laboratory Practise, incorporating random quality assurance audits of all phases of the study. Expert data evaluation and analysis are critical for establishing any association between exposure and outcome. This involves specialists from a broad range of scientific disciplines such as toxicologists, haematologists, clinical biochemists, pathologists, human and animal nutritionists and also biostatisticians. One of the pivotal requirements in data analysis is to distinguish those effects which are potentially treatment related from spurious occurrences or the result of normal individual biological variation. If differences exist between test and control, comparison to historical control data from the same laboratory as well as published data for the strain, sex and age of the animal being investigated is helpful, as well as data obtained with commercial reference lines. In Section 6 strategies are outlined for the safety and nutritional assessment of GM plant derived food and feed. The generation of studies for pre-market assessment of the safety and nutritional properties of food and feed from GM plants should follow a structured approach with stepwise development and consideration of the data obtained at each step in order to formulate the questions to be asked and answered at the next step (see Fig. 3). Hazards related to the intended genetic modifications are evaluated applying in silico, in vitro and in vivo safety studies of newly expressed protein(s), newly formed metabolites, and of natural substances whose levels may have been altered as a result of gene insertion. Guidelines have been developed by OECD describing detailed protocols for the safety testing of these substances in food and feed. A detailed testing strategy should be designed based on the prior knowledge regarding the biology of these products, so that the relevant endpoints are measured in the individual test. Testing of the safety and nutritional value of the whole GM plant or derived food and feed should be considered where the molecular, compositional, phenotypic, agronomic and other analyses have demonstrated differences between the GM plant derived food and feed EFSA GMO Panel / Food and Chemical Toxicology 46 (2008) S2–S70 S5 and their conventional counterpart, apart from the inserted trait(s), or if there are any indications or remaining uncertainties for the potential occurrence of unintended effects. In such a case, the testing program should include at least a 90-day rodent feeding study. In the context of the safety and nutritional assessment of GM plant derived food and feed, the adapted 90-day rodent feeding study, if triggered by the outcome of the molecular, compositional, phenotypic or agronomic analysis, functions as a sentinel study designed to assess potential unintended effects of toxicological and/or nutritional relevance rather than determining qualitative and quantitative intrinsic toxicity of defined food constituents. In the situation where molecular, compositional, phenotypic, agronomic and other analyses have demonstrated equivalence between the GM plant derived food and feed and their near isogenic counterpart, except for the inserted trait(s), and do not indicate the occurrence of unintended effects, experiences with GM plants modified for agronomic input traits have demonstrated that the performance of 90-day feeding trials with rodents or feeding trials with target animal species have provided little if anything to the overall safety assessment (except for added confirmation of safety). The use of 90-days studies in rodents should be considered for the detection of possible unintended effects in food and feed derived from GM plants which have been more extensively modified in order to cope with environmental stress conditions like drought or high salt conditions, or GM plants with quality or output traits with the purpose to improve human or animal nutrition and/or health. Ninety-day studies with rodents are normally of sufficient duration for the identification of general toxicological effects of compounds that would also give adverse effects after chronic exposure. In general, long term, chronic toxicity testing of whole GM food and feed is not expected to generate information additional to what is already known from in silico/in vitro testing and from subchronic testing. In cases where structural alerts or other information is available about the possibly altered occurrence of food components in the GM food and feed compared to its counterpart, the performance of specific toxicological testing, e.g. chronic, reproductive, etc., should be considered case-by-case, but preferentially only for the single substance of concern. Livestock feeding studies with target animal species should be conducted on a case-by-case basis to establish the nutritional benefits that might be expected from GM plants with claimed nutritional/health benefits. Possible effects of the new feed resource on animal performance, animal health, efficacy, and acceptability of the new feed ingredient should be investigated, and time spans for such studies should be determined on a case-by-case basis. There is a need for a more uniform approach to the design and analysis of animal feeding trials, and in particular for appropriate statistical analysis of data. The process of data interpretation requires extensive professional experience of the field, together with a thorough understanding of the concept of causality. One of the pivotal requirements is to distinguish those effects which are potentially treatment related from spurious occurrences or result from normal individual biological variation. Post-market monitoring is not a substitute for a thorough pre-market risk assessment, neither should it be considered as a routine need. Knowledge gained through post-market monitoring might at best describe only broad patterns of human nutritional exposure. In general it cannot be relied upon as a technique for monitoring adverse events or other health outcomes related to the consumption of GM plant derived food and feed. It can be anticipated that in the future the predictive value of a 90-day rodent feeding studies used for the safety assessment of whole food and feed will be enhanced by the integration of new technologies like transcriptomics, proteomics and metabolomics into the experimental risk assessment approach. Moreover, the use of ‘profiling’ technologies may also facilitate a non-targeted approach in compositional analysis in order to aid the detection of unintended effects in GM plant derived food and feed due to the genetic modification. These technologies are still under development, and need validation before they can be used for routine safety assessment purposes. In Section 7 conclusions and recommendations are presented on: The comparative approach to safety and nutritional testing of food and feed derived from GM plants. In silico and in vitro tools available for safety and nutritional testing of GM plant derived food and feed. Testing of defined single substances from GM plant derived food and feed in in vivo studies. Testing of whole GM plant derived food and feed in animal feeding studies. Importance of a structured approach for development of data for the pre-market safety and nutritional testing of GM plant derived food and feed. Role of post-market monitoring. 2008 European Food Safety Authority. Published by Elsevier Ltd. All rights reserved.