14 of 15 studies on composition and 6 of 15 animal feeding studies were performed by biotech industry companies.

16 of 30 studies (53 %), used material actually sprayed with glyphosate herbicide during cultivation. No information on dosage was given.

Only 4 of 30 studies (13 %) address the issue of glyphosate residues. None of these 4 studies were funded by industry.

Only 1 of 30 studies (3 %) has performed analysis and quantification of glyphosate residues. 30 published reports from studies of compositional analysis glyphosate-tolerant GM plant varieties and from feeding studies using glyphosate-tolerant GM plant varieties have been reviewed. These studies were performed in the years 1996–2015. A simple synthesis of available information on study design, methods and results shows that:

These findings fundamentally challenge the basis for regulatory assumption of substantial equivalence between glyphosate-tolerant GM varieties and unmodified comparators. The findings are a strong argument for mandatory inclusion of pesticide analysis data in regulatory assessment of GM crop, notably in assessments of herbicide-tolerant crops. Two of the animal feeding studies performed by independent researchers [86, 87] used GM crop material as well as unmodified comparators supplied by industry. No analysis was performed to control the compositional quality of this material.

Thus, the literature review indicates that there are relatively few representative studies available for regulatory evaluation of scientific evidence on herbicide-tolerant crop quality and safety. Furthermore, it is found that the majority of available studies are presented as reports from compositional analyses and animal feeding studies, and predominantly performed either by biotech industry companies (with potentially conflicting interests in research outcome) or by subcontractors working for the biotech industry companies (Tables 2, 3). Society should expect the biotech industry companies to continue to conduct such studies to peer-review standard and the industry should continue to bear associated costs. However, it is evident that appropriate revisions of standards are needed, and supplementary studies by independent researchers should be encouraged. Published evidence on safety testing presented by the industry has generally been recognized by EFSA as sufficient for regulatory purpose, despite the fact that several potentially conflicting issues have been continuously raised by independent researchers [19, 20, 28, 33–35, 52, 86]. Such critique has also questioned both the principle of delegated self-control and the validity of industry methods. Some of this critique has led to temporary adjustments of protocols and changes in methodology. A review by independent scientists in 1999 [33] examined results of three initial industry tests that were published in 1996. These first industry tests claimed substantial equivalence of GM crops glyphosate-tolerant GTS 40-3-2 soybean [34–36] and seed from glyphosate-tolerant cotton [37] compared to unmodified isolines. However, the review noted that the industry reports were based on tests of glyphosate-tolerant material grown in artificial conditions without application of complimentary glyphosate herbicides. The GM crops thus produced, were seen to be “not representative” of the crops actually produced in agriculture [33]. Several industry researchers immediately acknowledged the necessity to change these specific approaches, and subsequent industry publications on quality of glyphosate-tolerant varieties of soy [38–41], maize [42], alfalfa [43] and cotton [44] specified that normal cultivation practice had been used in production, including prescription rate application of glyphosate via commercial glyphosate herbicides such as Roundup. One industry paper published immediately following the 1999 criticism even specifies in its title that glyphosate herbicides have been applied [45]. Despite this change of practice and the acknowledged need for realistic field conditions to produce material for evaluation, numerous subsequent tests have been published where again biomass from glyphosate-tolerant cultivars is used for comparison despite having been grown in artificial conditions without application of complimentary herbicides. Recently, 10 studies presented by industry applicants as evidence for regulatory approval of glyphosate-tolerant cultivars were reviewed [34, 35], and the author concludes that lack of relevant herbicide application is still a discrediting flaw in such studies. However, although this highlights one systematic flaw in studies currently accepted for regulatory purpose documentation, the unknown magnitude of herbicide residues must be recognized as a subsequent and not least important aspect.

30 1 46 47 30 65 1 The relevance of testing for herbicide residues is highlighted by the findings of a recent study on composition of plant material [] performed by independent researchers. The study reports high levels of glyphosate residues (Fig.) in glyphosate-tolerant soybean (Roundup Ready soy GTS 40-3-2). The study also finds that residues of glyphosate and the primary metabolite aminomethylphosphonic acid (AMPA) are correlated to differences in crop composition. In 2003 and 2004, independent research demonstrated that residues of glyphosate herbicides will accumulate in glyphosate-tolerant plant material [] but found lower quantities than the subsequent findings reported in 2014 []. Another recent report from tests performed in Argentina (by independent scientists working for the German NGO Test-Biotech) have reported findings of even higher levels of glyphosate residues in harvests of glyphosate-tolerant soy [] (Fig.). These latest results indicate very high glyphosate residue levels up to 100 mg/kg in soybean and stand as an important indication which necessitates further sampling and analysis.

The results indicate a rise in glyphosate residue levels in recent decades. In 1999, a major producer of both glyphosate and GM crops declared that glyphosate residue levels of 5.6 mg/kg in glyphosate-tolerant soybean were considered to be extreme high values [30]. It seems apparent from Fig. 1 that such levels at present would be considered moderate or even low. To explain tendencies of rising residue levels it would be relevant to investigate actual application rates. Global production figures support the notion that very large quantities of glyphosate are being sold and dispersed.

It is interesting to note that several independent researchers have mentioned the specific question of glyphosate-residues in glyphosate-tolerant crops, asking for more data to clarify this issue [46, 47, 62]. The question has also been addressed in a review of concepts and controversies in EFSA environmental risk assessment of GM-crops; it was found that even in an environmental context more data on glyphosate residues is needed, as post-harvest biomass is potentially affecting soil biota and adjacent environments [19].

Studies of glyphosate-tolerant cultivar composition have identified differences in essential plant constituents, which have been attributed to in-plantae metabolic effects of glyphosate residues [48–50]. Such research indicates that glyphosate residues have negative effects on composition. Contrary to this, a recent review by authors from the United States Department of Agriculture [51] conclude that there is not sufficient evidence for claiming that glyphosate in glyphosate-tolerant crops a) significantly affects mineral composition or b) changes rhizosphere microbial community or c) increases susceptibility to disease from plant pathogens.

As a direct critique of the regulatory policies enforced by the European Food safety Authority EFSA, independent scientists have claimed that the present regime of industry self-control (autoregulation) is insufficient to provide necessary evidence and ensure the long-term interests of society. Industry studies therefore must be supplemented with additional, independent, research [19, 20, 52]. This, however, is not a view shared by researchers representing interests of biotech companies, who often participate in systematic opposition to any results questioning industry studies. It has been described as highly regrettable that independent scientific work is often attacked and discredited by concerted efforts of industry proponents and journal editors loyal to biotech sector interests [28]. A recent study [52] found clear evidence of double-standards in criteria for evaluation of safety studies on GMO cultivars such as herbicide-resistant plants. The authors document that evidence confirming safety is not exposed to the same intensive scrutiny as evidence indicating possible harm. This is paradoxical, as it should be evident that faulty findings in the first of these categories has potential for inflicting negative effects on consumer health. Faulty findings in the second category will not have the same implications, but may lead to exaggerated precaution, which can be conflicting in relation to commercial interests.

Evidence has emerged during the compilation of this review, which to a certain degree confirms the claims of double-standards: One of the industry studies reviewed here serves as a noteworthy example of malpractice. The study [31] was published by journal Food and Chemical Toxicology. The scientists authoring the study were employees of commercial companies Pioneer Hi-bred and DuPont. They conducted a safety study on DP-356Ø43-5 glyphosate-tolerant soybean by testing cultivated material in a feeding study using rats. According to the methods chapter of the study, the tested DP-356Ø43-5 glyphosate-tolerant soybean was sprayed with glyphosate herbicide. Glyphosate herbicide was the only pesticide used in the strictly controlled production on parallel fields of; (a) glyphosate-tolerant soy (sprayed) and (b) unmodified soy (not sprayed). The irregular aspect relates to the fact that a wide array of subsequent tests for pesticides was performed in the produced soy materials, screening these for a variety of active ingredient chemicals. And, although glyphosate-herbicide was specified to be the only pesticide applied in the strictly controlled test-plot cultivation, an analysis for glyphosate residues was omitted. Instead the cultivated material was analyzed for numerous herbicide ingredients that were fundamentally irrelevant. This published study should be seen as an example supporting the arguments demanding revision of the regulatory framework mandating self-control of biotech industry products. Furthermore, given the recent heightening of qualitative requirements for such studies, which in its utmost consequence is seen as retractions of publications, I nominate the mentioned study [31] as a prime candidate for editorial re-evaluation.

Other reviews of published testing Four recent reviews of data on GM crops in agriculture [53–56] present evidence confirming herbicide-tolerant cultivar equivalence, as compared to non-modified comparators. None of these reviews mention herbicide residues or their potentially conflicting nature in relation to concept of substantial equivalence. Contrary to this, three reviews by independent scientists approach the role of herbicide residues in GM crops or present indications of toxicity. In one of these [57] the authors review several contested safety assessment studies and speculate whether adverse effects reported in animal testing in 2002 [58], in 2004 [59] and in 2009 [60] could be attributable to pesticide residues contained in the tested GM-crop material. Another recent review [61] concludes that parts of published evidence in assessments of health risks of GMO foods are general indications of toxicity. The regulatory challenges relating to oversight and development of standards for investigating herbicide residues in herbicide-resistant crops are largely ignored by industry, by most independent researchers and by regulatory authorities. However, a few specific aspects have been reviewed and important recommendations have been presented [62]. Such recommendations include specific measures, notably the concept of supervised field trials, which is seen as an important potential improvement of the current system of industry self-control and scientific autonomy.

The future of herbicide-tolerant crops Commercially advanced herbicide-tolerant cultivars are popular amongst stakeholder investors and farmers engaged in agro-industrial production of maize and soybean in countries of North- and South America. From a database listing GM crop varieties pending regulatory approval [63] it seems that a majority of these GM crops are either herbicide-tolerant varieties or varieties with stacked events which include herbicide tolerance. Some new varieties have herbicide-tolerance traits which are selected from microorganisms systematically bred in environments with high glyphosate concentrations [64]. Traditional first-generation glyphosate-tolerant crops, such as the GTS-40-3-2 soybean which still dominates global production, are only 45–50 times more tolerant than unmodified varieties (the glyphosate dose inducing LC 50 -outcome in GTS-40-3-2 is about 50× that of unmodified soy). Obviously, this physiological vulnerability can be perceived as a deficiency, subsequently limiting continuous spiraling increase of dosage as main strategy against hard weed. By using new sources of transgenes and gene-stacks with combinations of several transgenes conveying multi-pathway tolerance to specific active ingredients, second-generation cultivars are seen as having significantly improved tolerance to specific herbicides or combinations of herbicides. This development should be seen primarily as a method paving the way for escalating application of herbicide. It seems that in the on-going struggle to eradicate resistant weeds, farmers rely heavily on solutions offered by commercial producers of herbicides. A main strategy seems to be developments that allow for higher dosage of herbicides such as glyphosate. It is recommended that such developments should be met by regulatory initiative to ensure necessary regulatory oversight of inevitable and expectable secondary consequences, such as compositional changes and combinatorial effects with other plant- or pesticidal compounds. These potential changes must be monitored in analysis of representative material, which can be taken as samples from the actual agro-ecological production systems. The present maximal residue limits (MRLs) allow for relative high concentrations of herbicide residues. In Brazil in 2004 the MRL in soybean was increased from 0.2 to 10 mg/kg: a 50-fold increase, but only for glyphosate tolerant soy. In Europe, the MRL for glyphosate in soybean was raised by a factor 200 from 0.1 to 20 mg/kg in 1999 [66] and the same MRL of 20 mg/kg was adopted by the US based on recommendations of the Codex Alimentarius Commission. In 2013, the MRL tolerance levels for glyphosate residues in US soybean were raised from 20 to 40 mg/kg (Fig. 1). The increases coincide with industry development of new GM varieties with stronger tolerance to glyphosate. In these cases the MRL values appear to have been adjusted pragmatically in response to actual observed, or expected, increases in the content of residues in glyphosate-tolerant soybeans. In this context it would be appropriate to collect and review more of the existing data on glyphosate residues in glyphosate-tolerant crops. However relevant such a question may be, it cannot be satisfactorily answered here due to the fact that only sparse published information exists on this issue. Despite the limited number of analyses for glyphosate residues in glyphosate tolerant crops, the few recent tests reported [30, 65] indicate surprisingly high levels of glyphosate residues. Such findings should fundamentally challenge regulatory assumption of substantial equivalence between glyphosate-tolerant varieties and their unmodified comparators.

Substantial equivalence The principle of substantial equivalence is fundamental for assessment of genetically modified plants, justifying some explanation. Substantial equivalence is a concept developed by OECD in 1991–1993 [23], establishing that a novel food, for example one derived from genetic modification or engineering, should be considered the same as (and as safe as) a conventional food, if it demonstrates the same characteristics and composition as the conventional food [67]. In 1997, the European Commission regulated its policy on novel foods (from GM plants) stating that food and feed from such plants are expected not to “present a danger for the consumer”, or “mislead the consumer”, or “differ from foods or food ingredients which they are intended to replace to such an extent that their normal consumption would be nutritionally disadvantageous for the consumer” [68] The regulation goes on to state that “[this policy…] shall apply to foods or food ingredients […] which, on the basis of the scientific evidence available […] are substantially equivalent to existing foods or food ingredients as regards their composition, nutritional value, metabolism, intended use and the level of undesirable substances contained therein” [68]. This allows a discussion on the qualitative evaluation of substances which vary from benign to harmful. It should be evident that pesticide residues belong in the category “undesirable substances”.