Volume 20, Issue 5 p. 5226-5257
Open Access

Residues of glyphosate in food and dietary exposure

John L. Vicini

Corresponding Author

John L. Vicini

Regulatory Sciences, Bayer Crop Science, Chesterfield, Missouri, USA


John L. Vicini, Regulatory Sciences, Bayer Crop Science, 700 Chesterfield Parkway West, Chesterfield, MO 63017, USA.

Email: [email protected]

Contribution: Conceptualization, Writing - original draft, Writing - review & editing

Search for more papers by this author
Pamela K. Jensen

Pamela K. Jensen

Regulatory Sciences, Bayer Crop Science, Chesterfield, Missouri, USA

Contribution: Writing - original draft, Writing - review & editing

Search for more papers by this author
Bruce M. Young

Bruce M. Young

Regulatory Sciences, Bayer Crop Science, Chesterfield, Missouri, USA

Contribution: Writing - original draft, Writing - review & editing

Search for more papers by this author
John T. Swarthout

John T. Swarthout

Regulatory Sciences, Bayer Crop Science, Chesterfield, Missouri, USA

Contribution: Writing - original draft, Writing - review & editing

Search for more papers by this author
First published: 16 August 2021
Citations: 20


Glyphosate is the active ingredient in Roundup® brand nonselective herbicides, and residue testing for food has been conducted as part of the normal regulatory processes. Additional testing has been conducted by university researchers and nongovernmental agencies. Presence of residues needs to be put into the context of safety standards. Furthermore, to appropriately interpret residue data, analytical assays must be validated for each food sample matrix. Regulatory agency surveys indicate that 99% of glyphosate residues in food are below the European maximum residue limits (MRLs) or U.S. Environmental Protection Agency tolerances. These data support the conclusion that overall residues are not elevated above MRLs/tolerances due to agricultural practices or usage on genetically modified (GM) crops. However, it is important to understand that MRLs and tolerances are limits for legal pesticide usage. MRLs only provide health information when the sum of MRLs of all foods is compared to limits established by toxicology studies, such as the acceptable daily intake (ADI). Conclusions from dietary modeling that use actual food residues, or MRLs themselves, combined with consumption data indicate that dietary exposures to glyphosate are within established safe limits. Measurements of glyphosate in urine can also be used to estimate ingested glyphosate exposure, and studies indicate that exposure is <3% of the current European ADI for glyphosate, which is 0.5 mg glyphosate/kg body weight. Conclusions of risk assessments, based on dietary modeling or urine data, are that exposures to glyphosate from food are well below the amount that can be ingested daily over a lifetime with a reasonable certainty of no harm.


Modern agricultural practices provide farmers with tools to maximize the production of food for a growing world population. Herbicidal pesticides help farmers grow more food on less land by protecting crops from weeds competing for essential nutrients, water, and sunlight. Oerke (2006) concluded that 34% of potential crop losses were attributable to weeds.

Glyphosate is a herbicide that is a phosphonomethyl derivative of the amino acid glycine (Figure 1). Glyphosate inhibits Class I 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS), an enzyme in the shikimate pathway that is required for de novo synthesis of aromatic amino acids in plants. Because these amino acids are needed for synthesis of proteins and lignin in all eukaryotic plants (Tzin & Galili, 2010), glyphosate is a broad spectrum herbicide. Cells from humans and animals do not have this pathway, which is why phenylalanine, tryptophan, and tyrosine are essential amino acids that need to be supplied in diets for mammals and birds (Giesy et al., 2000). Glyphosate has global regulatory approvals for its use as a nonselective herbicide in agriculture, mostly preplanting, as well as around roadways and railroads.

Details are in the caption following the image
Glyphosate chemical structure and microbial degradation pathways in soil. Abbreviations: AMPA, aminomethylphosphonic acid; GOX, glyphosate oxidoreductase. Adapted from Duke (2021) and Giesy et al. (2000)

Transgenic varieties of corn, soybeans, canola, and sugar beets were made tolerant to glyphosate by the insertion of a gene that encodes for a Class II microbial EPSPS that is not inhibited by glyphosate. This variant was discovered from the CP4 strain of Agrobacterium tumefaciens that was isolated from wastewater at a glyphosate manufacturing facility (Barry et al., 1997). Aminomethylphosphonic acid (AMPA) is a major microbial degradation product of glyphosate (Figure 1) but AMPA does not compete with the EPSPS substrate phosphoenolpyruvic acid for enzyme binding (Duke et al., 2012; Reddy et al., 2004). In rats and humans, metabolism of absorbed glyphosate has not been shown to occur (BfR, 2015; Niemann et al., 2015).

Development of glyphosate tolerant (GT) crops required multiple global regulatory agencies to authorize the cultivation of the GT seed along with additional approvals for the application of glyphosate over-the-top on GT crops. Glyphosate is considered by these regulatory agencies to be safe when used according to each specific label use. More specifically, regulatory agencies have concluded that glyphosate does not pose a cancer risk when used according to the label. Recently, one organization, that lacks regulatory oversight, listed glyphosate as a “probable carcinogen” (IARC, 2015); however, multiple regulatory agencies and competent authorities have assessed that listing and all of these reviews concluded that glyphosate is not a carcinogen. Assessments by these regulatory agencies and competent authorities about the carcinogenicity of glyphosate are listed in Table S1.

Global regulatory authorities ensure food safety by establishing realistic levels of human exposure to a given pesticide that enable calculating exposure levels that provide a “reasonable certainty of no harm” for the consumer (Winter et al., 2019). However, the increased public awareness of what is in food has resulted in a surge in reports on residues of glyphosate, not only in peer-reviewed scientific journals but also in nonscientific media articles and stories posted to nonrefereed social media sites. A literature search was conducted on topics related to glyphosate, including residues, from the last 10 years of publications following EFSA guidelines (EFSA, 2011). Relevant literature identified for this search for all of these topics is available (GRG, 2021). Additional papers were identified by checking references from these sources for additional manuscripts and government reports that were not identified by the original search. In order to provide a robust discussion of the scientific findings and limitations for a representative, albeit not fully comprehensive, set of reports, articles were selected that (1) applied specifically to food that is commonly consumed, (2) for field experiments that appeared to follow product labels, and (3) provided some evidence of type of the method for quantifying residues. A second search was conducted using the terms “Glyphosate” or “Glifosato” from 2015 to 2020 using Buzzsumo, which searches media content on the internet across popular social media. This search generated 70,414 hits. The public impact of the media reports is remarkable and, therefore, necessitates some inclusion in this review, when related to reports of glyphosate detection in food. Many of these media reports have interest-grabbing headlines about the detection of pesticide residues, but do not provide interpretable data, or when they do, the values are rarely provided in a context to help understand if these residues could result in a consumer health issue. Moreover, laypeople typically lack formal training in the scientific principles underlying dose–response relationships and, consequently, often regard the detection of a chemical as synonymous with that presence being unsafe (Siegrist & Bearth, 2019).

The media reports and misunderstandings of dose–response principles have created an expectation or desire that food and beverages should have zero residues. The people that consumers look to for food safety advice (e.g., health care professionals, nutritionists, and dietitians) should have scientifically balanced information needed to discuss the topics of residues. However, current training curricula for these professions have limited, if any, coursework on relevant topics. For instance, medical students receive little training on issues of environmental chemical exposure (Temte & McCall, 2001). As a start, it is important to convey that detectable residues, including naturally occurring toxicants, are in many foods regardless of agriculture practice, including both organic and conventional agriculture, and that the presence of residues does not directly equate with harm (Winter et al., 2019). A next step is to provide the necessary context around what a reported detection means in terms of current safety guidelines. The science-based risk assessment of residues relies on the reporting of accurate concentrations of residues of interest in commodities/foods and comparison of these values to a regulatory-derived safety standard. The maximum residue limits (MRLs) are derived from regulatory studies of residues obtained when pesticides are applied according to the proposed label to agricultural products following good agricultural practices (European Commision, 2020b; Winter et al., 2019). These data are used to statistically determine the legal limits for residues when individual crops are grown in adherence to a regulatory agency's approved product label. By themselves, they are not a safety standard, per se. Crops with residues that are less than the MRL would be considered safe, assuming all other crops treated with the same pesticide had residues that are at or below the MRL; however, a crop that experiences an exceedance of the MRL is not by definition unsafe. One way to provide the proper context regarding dietary residues and potential risk is by comparing the sum of these residues for all crops to health-based guidance values, such as the acceptable daily intake (ADI). The ADI is a health-based guidance value (Herrman & Younes, 1999) that, according to the Chemicals Regulation Directorate (2013), is the amount of a pesticide that can be ingested over a lifetime without appreciable risk. In the United States, an equivalent health-based guidance is the reference dose (RfD) and tolerances are the legal limits for residues in crops. The most recent EU ADI for glyphosate (0.5 mg glyphosate/kg body weight) is used throughout this review in calculations, partly because it is a more conservative value than the RfD used by the U.S. Environmental Protection Agency (EPA) (1 mg/kg).

Advancements in analytical chemistry technology have contributed to increased reports of detection of pesticide residues because increasingly sensitive detection methods have resulted in many assays now accurately measuring pesticides in parts per billion (ppb, µg/L, or µg/kg). As a result, it is becoming increasingly more likely that residues, that were previously too low to measure, will be detected. Comparing the sum of residue concentrations of foods in the diet to health-based guidance enables an assessment of potential health impacts. Moreover, the increasing availability of assay methods that require less complicated and less expensive instrumentation has increased the reporting of residue detects. However, like all analytical methods, these less complicated assays must be validated for the sample matrix being assayed. If the validation is lacking, this results in the reporting of residue values that are not quantifiable or are reported in the absence of the laboratory-derived limits of detection. Risk assessment involves a comparison between the ADI or RfD, derived from animal studies, and estimated exposure of the pesticide from all sources. The objectives of this review are to: (1) review the assays available for measuring glyphosate in food, water, beverages, and urine; (2) review reports of testing glyphosate in food or urine, and convert these values to estimates of exposure; and (3) put these exposure estimates into context by comparing them to health-based guidance values used to assess risk.


There are many different types of analytical technologies that can and have been used to detect glyphosate and its metabolite AMPA. Both molecules are highly polar and ionic and have extremely limited solubility in organic solvents, which makes their analysis more difficult than most pesticides on the market today. As a result, glyphosate and AMPA are not typically included in multiresidue pesticide monitoring methods. A review by Koskinen et al. (2016) on methods for analysis of glyphosate and AMPA in crops, water, and soil published between 2000 and 2015 highlights the vast combinations of derivatization, separation, and detection techniques that have been attempted over the years. The majority of published methods were based on liquid chromatography with fluorescence detection (LC-FLD) or liquid chromatography coupled to mass spectrometry (LC-MS/MS). However, in recent years, an ELISA (enzyme-linked immunosorbent assay) kit has been used increasingly for glyphosate analysis. Thus, this review will focus on discussion of these three techniques and their advantages and disadvantages for glyphosate analysis.

2.1 Liquid chromatography with fluorescence detection

LC-FLD methods were developed before mass spectrometers became common in analytical labs. While these methods have drawbacks, as discussed below, they currently provide a reasonable alternative for labs without access to more sophisticated LC-MS/MS equipment. Since glyphosate and AMPA do not have a chromophore group, typical LC detectors cannot be used for residue analysis without first derivatizing the analytes. Although several derivatization reagents have been tried (Arkan & Molnár-Perl, 2015), the most commonly used are o-phthalaldehyde (OPA) and 9-fluorenyl methoxycarbonyl chloride (FMOC-Cl), both of which result in products that can be detected using fluorescence. Since any compound in the sample having a primary amine will be derivatized, the resulting chromatograms can have many peaks. Therefore, the only means for identifying glyphosate and AMPA is by comparison of the retention time to that of pure reference standards. This lack of specificity due to the occurrence of many peaks makes the analysis more susceptible to matrix interferences, which can result in mistaken identification (false positives) or overestimation in quantitation from coeluting matrix components. For this reason, complex matrices, such as food and feed commodities, require extensive sample cleanup and concentration steps prior to analysis to reduce matrix interferences and to obtain the desired method sensitivity. This makes these methods very tedious and labor-intensive and limit the number of samples that can be analyzed in a day.

The method developed by Cowell et al. (1986) is an example of a method that requires extensive cleanup to obtain reliable detection at a limit of quantitation (LOQ) of 0.05 ppm (mg/kg) for crop matrices using OPA derivatization. A majority of LC-FLD methods, however, have favored the use of precolumn derivatization using FMOC-Cl (Druart et al., 2011; Fitri et al., 2017; Hogendoorn et al., 1999; Kaczyński & Lozowicka, 2015; Le Bot et al., 2002; Nedelkoska & Low, 2004; Wang et al., 2016). Although these methods use less extensive sample cleanup compared to OPA, they require careful optimization of the derivatization reaction conditions to minimize matrix effects. An advantage of FMOC-Cl methods is that the derivatives of glyphosate and AMPA become less polar, making them more readily separated using traditional reverse phase columns and thus easier to implement in most labs. This simpler chromatography is one reason why FMOC-Cl derivatization is also used in several LC-MS/MS-based methods, such as for analysis of milk and nutritional ingredients (Ehling & Reddy, 2015), crops (Bernal et al., 2012; Goscinny et al., 2012), and various foods (Liao et al., 2018; Thompson et al., 2019).

2.2 Liquid chromatography coupled to mass spectrometry

In general, complicated, multistep derivatization and cleanup procedures are not desirable as they are time-intensive and often more prone to errors. The introduction of LC-MS/MS provided the ability to analyze glyphosate and AMPA without derivatization and reduced the degree of cleanup required. Several LC-MS/MS direct analysis methods have been developed for different matrices, including various crops (Botero-Coy et al., 2013; Chamkasem & Harmon, 2016; Kaczyński & Lozowicka, 2015; Marek & Koskinen, 2014; Martins-Júnior et al., 2011), food matrices (Chamkasem & Vargo, 2017; Chen et al., 2013; Jensen et al., 2016; Nagatomi et al., 2013; Steinborn et al., 2016; Zoller et al., 2018), and urine (Jensen et al., 2016). LC-MS/MS provides added selectivity over LC-FLD by using mass measurement for identification in addition to retention time comparison to reference standards. The increased specificity provided by MS/MS detection results in greater sensitivity and accuracy, compared to LC-FLD methods. The ability to use stable-isotope-labeled (e.g., 13C) analogs of glyphosate and AMPA as internal standards adds another degree of confidence in confirming the analyte identity as well as improving accuracy and precision. For these reasons, LC-MS/MS methods are typically preferred for analysis of glyphosate and AMPA.

2.3 Enzyme-linked immunosorbent assay-based methods

ELISA has increasingly been used in the last 5 years as a simple and quick method for glyphosate analysis and a commercial kit is available (Abraxis, now owned by Eurofins Scientific, Luxembourg). Early application of the ELISA was aimed at development, optimization, and validation for analysis of glyphosate in water samples (Clegg et al., 1999; Rubio et al., 2003) and was primarily intended as a rapid screening tool, where any positive results would be confirmed by a more quantitative method, such as LC-MS/MS (Byer et al., 2008; Lee et al., 2002). Used in this manner, ELISA can be a useful and effective tool, especially for analysis of water samples where interferences from matrix components are limited. However, since there are no sample clean-up or separation steps involved in the ELISA, it is easy to see how application to more complex matrices could present issues in accurate quantitation.

According to the ELISA kit's instructions, calibration standards are prepared in water and the resulting standard curve is used to extrapolate the concentration of glyphosate in the sample. This assay uses an indirect competitive ELISA technique, which means glyphosate in the sample competes with a glyphosate-enzyme conjugate for antibody binding sites. Therefore, the intensity of the absorbance signal generated by the enzyme reaction product is inversely proportional to the amount of glyphosate that is present in the test sample. Because the assay uses indirect detection, components present in more complex matrices, like milk, may interfere with multiple aspects of the ELISA, including the glyphosate-enzyme conjugate binding to the antibody, that would result in a decrease in the amount of enzyme product generated. The resulting lower absorbance signal would then be interpreted incorrectly as the presence of glyphosate in the sample. Even with 100- or 1000-fold dilution of samples prior to analysis, macroconstituents, such as sugars and lipids, could still have an impact on the assay. Preparing the calibration standards in the matrix being analyzed is an easy step that could help correct for these matrix effects (Schmidt & Alarcon, 2011; Singh et al., 2018), yet none of the reports cited in this review using this ELISA have indicated using this approach.

Even in the relatively simple matrices of surface and ground water, some researchers found the assay tended to overestimate glyphosate concentrations or generate false positives (Byer et al., 2008; Mörtl et al., 2013), possibly due to interferences from a range of components in samples, such as metal ions, organic matter, or high acid content. Unfortunately, the glyphosate ELISA is often applied to complex matrices without validating performance in the target matrix, such as with reports on breakfast foods (ANH-USA, 2016; John & Liu, 2018), beer and wine (Cook, 2019; Fagan, 2016), and breast milk (Honeycutt et al., 2014). In other studies, which used the ELISA to analyze human and animal tissue samples (Krüger et al., 2014) or honey, corn, and soy product samples (Rubio et al., 2014), reported results included recovery tests in the matrices being analyzed, but did not evaluate potential matrix effects on the limit of detection (LOD) or generation of false positives.

The glyphosate immunoassay approach has also been extended to other formats, such as fluorescence covalent microbead immunosorbent assays (FCMIA) (Biagini et al., 2004) and immunostrip tests (Eurofins Abraxis, Warminster, PA). The FCMIA method, validated in water and urine (Biagini et al., 2004), offers a promising way to analyze for multiple pesticides simultaneously, but no similar validation has been published for the glyphosate immunostrip tests

2.4 Assay validation

Assay validation is a requisite step no matter the type of analytical method used. The method should be assessed for its ability to generate accurate and reproducible data at the glyphosate concentrations being measured and for the sample matrix being analyzed. Method validation is a rigorous approach in which several key parameters that define the capabilities and limitations of the method are evaluated. Validation criteria include defining the linear range, LOD, LOQ, specificity, accuracy, and precision. Detailed discussion of each of these parameters and how they are applied and evaluated under varying analytical scenarios can be found elsewhere (Hill & Reynolds, 1999; SANTE, 2017). The focus of this discussion of studies that used the ELISA is on how particular aspects of these parameters can impact confidence in the reliability and accuracy of the glyphosate data produced.

Estimation of accuracy and precision are central components of defining a method's capabilities and are determined using spike and recovery experiments. A spiked concentration at or near the LOQ (FAO/WHO, 2017a) should be tested to ensure assay accuracy since it is reasonable to expect that these low levels of analyte are most affected by interference from matrix components. When glyphosate assays are conducted without examining the accuracy of the ELISA in the relevant matrix (Cook, 2019; Honeycutt et al., 2014; John & Liu, 2018) or by using concentrations spiked at much higher levels than the LOQ (Krüger et al., 2014), the reported data cannot be accurately interpreted.

Critical to reporting assay values at the low range is properly establishing the LOD and LOQ of the method used. There are multiple techniques for determining LOD and LOQ (Bernal, 2014; Corley, 2003), but simply defined, the LOD is the smallest concentration of analyte in a sample that can be reliably differentiated from the background (e.g., blank sample), whereas the LOQ is the smallest concentration that provides a quantitative result with suitable accuracy and precision. It is not sufficient to define the LOD and/or LOQ simply based on the lowest concentration in a standard curve, as this practice does not account for the effects of matrix components and thus does not reflect the true capability of the method. Instead, the LOD and LOQ are specific to the matrix being analyzed and must be determined experimentally through the use of blank matrix samples. Analysis of blank samples is also important in demonstrating the specificity of the method. If blank samples are not included, it is hard to know if the glyphosate detected is real or a possible artifact caused by components from the matrix or other reagents that are interfering with the assay (ANH-USA, 2016; Cook, 2019; Honeycutt et al., 2014; John & Liu, 2018). Confirming results with a secondary method can be a useful tool for verifying the presence (or not) of glyphosate in the sample. Simply comparing results between two methods should not, however, be the primary basis for determining a method is valid, as was done by Krüger et al. (2014) when comparing an ELISA assay to GC-MS/MS, as this does not verify the LOD/LOQ, accuracy, precision, or matrix effects of the method being used.

2.5 Use of reported data

Pesticide residue reports should include clear descriptions about the quality of the method being used since this information is critical to evaluating the reliability of the data. No analytical method is perfect or universally applicable to all sample types, so it is important to know details of the method used and to what extent, if any, the method was validated for its stated purpose.

When interpreting data, treatment of values around the LOD and LOQ can be a source of error or confusion. By definition, values below the LOD are not reliably distinguished from background noise and thus often are better reported as “not detected” or “<LOD.” Results above the LOD but below LOQ, are by definition, not quantitative and, therefore, are inconsistent with being mathematically analyzed. If included in the report, regulatory agencies suggest that they be regarded as qualitative results, often designated as <LOQ (EPA, 2000; SANTE, 2017), and their treatment in any statistical analysis be clearly stated and not be included as its numeric value. Various resources (Corley, 2003; EFSA, 2010; EPA, 2000) provide guidance on how to handle values below LOD or LOQ when performing calculations.

Using a method that has not been properly validated may be suitable for qualitative assessments of the data, such as comparing the frequency of detects between different sample groups. However, if the data are being used to make further calculations, statistical comparisons, or for use in a risk assessment, then it is critical that there is confidence in the numerical results. Studies that provide the information needed to determine whether the method was validated according to generally accepted standards (FAO, 2010; Hill & Reynolds, 1999), including determination of LOD and LOQ, and if samples were analyzed using appropriate quality control measures (FAO/WHO, 2017a; FDA, 2018a) can be more reliably used in risk assessments.

2.6 Summary about glyphosate assays

All methods have their pros and cons and a summary of the relative merits of the three method types discussed is summarized in Table 1. For instance, LC-MS/MS methods to detect glyphosate and AMPA residues offer the most sensitivity, specificity, and reliability across a wide variety of complex matrices but are more expensive and require greater expertise than other methods. Although the ELISA, when compared to LC-MS/MS and LC-FLD, is quicker and less expensive for analysis of glyphosate, reports to date using this method to quantify glyphosate in complex matrices have typically provided, as summarized below in this review, little to no supporting validation data that allow the results to be interpretable and repeatable by other scientists (ANH-USA, 2016; Byer et al., 2008; Cook, 2019; Fagan, 2016; Honeycutt et al., 2014; John & Liu, 2018; Krüger et al., 2014; Mörtl et al., 2013; Rubio et al., 2014). Regardless of technique, to interpret data, knowledge of the capabilities and limitations of the method used are needed. Thus, when reading about reports of glyphosate residues where the data are presented without inclusion or reference to method details, validation details, or appropriate validation of assay capabilities (e.g., LOQ, accuracy, and specificity) in the relevant matrix, the results should be regarded cautiously.

TABLE 1. Relative differences in selected characteristics of types of assays used to quantify glyphosate in different matricesa
Sample matrix LC-MS/MSb LC-FLD ELISA
LOD/LOQ Water + + +
Corn/soy + +++ ++
Milk + +++ ++
Urine + + +
Accuracy Water +++ +++ +++
Corn/soy +++ +++ ++
Milk +++ +++ +
Urine +++ +++ ++
Specificity (lack of false positives) Water +++ +++ ++
Corn/soy +++ ++ +
Milk +++ ++ +
Urine +++ +++ +
Difficulty/time (incl. sample prep) Water + + +
Corn/soy ++ +++ +
Milk ++ +++ +
Urine + + +
Cost NA +++ ++ +
  • a Relative comparison of assay types and not of specific assays used in studies cited in this paper.
  • b Ranking: + Low; ++ Medium; +++ High.
  • Abbreviations: ELISA, enzyme-linked immunosorbent assay; LC-FLD, liquid chromatography with fluorescence detection; LC-MS/MS, liquid chromatography coupled to mass spectrometry.


Global regulatory authorities, such as the U.S. FDA, U.S. EPA, and the European Food Safety Authority (EFSA), have required extensive data both prior to, and often after, commercialization of pesticides, such as glyphosate. The goal of monitoring food for pesticide residues by regulatory agencies is to ensure the public that pesticide applicators are following approved label instructions and that residues are not at unexpectedly high amounts (Winter et al., 2019). The U.S. FDA and EFSA compile yearly monitoring reports of pesticides using validated assays that are capable of measuring multiple pesticide residues in a single analysis. Access to comprehensive data from these surveys is important to understand the average concentrations, validation limits of the assays, and percent of the samples that have detectable residue levels.

A new glyphosate assay, separate from the multiresidue method, was developed and was first used with samples from the 2010 survey in the EU (EFSA, 2013). In the United States, an assay exclusively for measuring glyphosate was first used for market surveys in samples obtained in 2016 for a limited number of foods (FDA, 2018b). A total of 16,069 samples are tested and are reported in Tables 2–4, and >99% of these had values that were less than the MRL. The limited numbers of samples that had glyphosate greater than the MRL suggest that there was a high degree of compliance in use of glyphosate in the production of these crops.

TABLE 2. Selected examples of market survey of unprocessed food items for detectable glyphosate in Europea
Commodity # Samples # Detects (%) # Detects >Tol. (%) Min Detect (mg/kg) Max Detect (mg/kg) Avg. Detect (mg/kg) Min LOQb (mg/kg) Max LOQ (mg/kg) EU Tolerance (mg/kg)
Small grains
Barley 152 18 (12%) 0 0.02 8.9 1.47 0.01 0.1 20
Buckwheat and other pseudo-cereals 152 16 (11%) 12 (8%) 0.02 12.7 1.42 0.01 0.1 0.1
Lentils (dried) 238 82 (34%) 1 (0.4%) 0.01 11 1.46 0.01 0.1 10
Oat 113 2 (2%) 0 0.68 0.82 0.75 0.002 0.1 20
Rye 759 31 (4%) 0 0.01 1.8 0.38 0.002 2 10
Wheat 1,234 50 (4%) 0 0.01 2.9 0.4 0.002 0.5 10
Animal-derived commodities
Animal fat (all) 54 0 0 0.01 0.1 None
Eggs 15 0 0 0.01 0.1 None
Milk (all) 161 0 0 0.0001 0.1 0.05
Row crop commodities
Maize/corn 96 0 0 0.002 0.1 1
Sweet corn 83 0 0 0.01 0.1 3
Rapeseed/canola 66 0 0 –  0.01 0.1 10
Soybeans 61 4 (7%) 0 0.5 0.89 0.69 0.002 0.1 20
Other commodities
Honey 563 51 (9%) 19 (3%) 0.01 0.61 0.1 0.01 0.14 0.05
  • a Data sources: (EFSA, 2018, 2019, 2020).
  • b Multiple LOQs are reported by different countries in the survey; therefore, minimum and maximum LOQs are provided.
TABLE 3. Combined results for glyphosate in samples of corn, soybean, milk, and eggs in the United States (FDA and USDA)a
Commodity Source # Samples # Detects (%) # Detects > U.S. Tol. (%) Min Detect (mg/kg) Max Detect (mg/kg) Avg. Detect (mg/kg) LOQ (mg/kg) U.S. Tolerance (mg/kg)
Soybean grains USDA 300



0 0.265 18.53 1.94 0.25 20
Soybean grains FDA 337



0 0.003b 10 1.08 0.01 20
Corn (maize) FDA 313



0 0.002b 4.5 0.07 0.01 5
Milk FDA 121 0 0 0.01 NTc
Eggs FDA 108 0 0 0.01 0.05
  • a Data Sources:(FDA, 2018b, 2019; USDA, 2020).
  • b FDA designates residue levels detected above the LOD but below the LOQ as “trace.”
  • c No tolerance for milk in the U.S. Codex and EU has MRL = 0.05 mg/kg.
TABLE 4. Summary offoods tested for glyphosate by CFIA in Canada and just needs a space separating of and foodsa
Program Food type # Samples tested % Samples with glyphosate residues detected % Samples with glyphosate residues above MRLs
National Chemical Residue Monitoring Programa Fresh fruits and vegetables 317 7.3% 0%
Processed fruits and vegetables 165 12.1% 0%
Targeted Surveysa Grain products 869 36.6% 3.9%
Juice and other beverages 496 16.3% 0.2%
Bean/pea/lentil products 869 47.4% 0.6%
Soy products 263 11.0% 0%
Children's Food Projecta Infant cereal 82 31.7% 0%
Infant food 127 30.7% 0%
Total 3188 29.7% 1.3%
Kolakowski et al. (2020) Other grains 626 30.5% 5.6%
Corn and corn products 501 23.0% 0%
Pulses and pulse products 770 43.1% 0.6%
Wheat and wheat products 807 76.3% 0%
Barley and barley products 103 49.5% 0%
Oats and oat products 310 74.5% 0%
Soy and soy products 204 9.8% 0%
Dairy and/or meat 22 0% 0%
Fresh or processed fruits and vegetables 1473 11.9% 0.4%
Infant foods 927 31.3% 0%
Manufactured foods 2212 60.8% 0%
Total 7955 42.3% 0.6%
  • a Data Sources: (CFIA, 2017).
  • Abbreviation: MRL, maximum residue limit.

The paucity of peer-reviewed reports of glyphosate residues in cultivated crops might be due to the complexity of obtaining samples with adequate knowledge of specific crop cultivation practices. Data from cultivation studies are variable, especially when the variability of environmental effects are not captured by a limited number of trial sites (Fleming et al., 2017). Instead, it is easier, and more common, to do market surveys of the foods that consumers purchase, recognizing that these will likely be blended commodity crops with various cultivation practices.

3.1 Soybean

Other than the surveys conducted by regulatory agencies (Tables 2–4) or residue studies submitted to regulators, there are only two academic studies that measured glyphosate residues in commodity crops and both are with soybeans (Arregui et al., 2004; Bøhn et al., 2014). Arregui et al. (2004) conducted a study in Argentina at five farms over three growing periods. Glyphosate was measured using extraction and high performance liquid chromatography (HPLC) analysis with an LOD = 0.02 mg/kg. Glyphosate was reported in soybeans at 0.1, 1.6, and 1.8 mg/kg for two, two, and three applications, respectively. AMPA values were lower (not detectable, 0.4 and 0.9 mg/kg, respectively) than glyphosate concentrations. The authors concluded that concentrations of glyphosate and AMPA were greater when sprayings were applied later in the growing season, near flowering.

Bøhn et al. (2014) compared concentrations of glyphosate between organic, conventional, and GT soybeans. Samples were collected during one growing season from 31 individual fields in Iowa. The samples were analyzed by a commercial lab but information on the type of assay and the LOD/LOQ was not provided. The authors reported for the genetically modified soybean samples that the average concentrations of glyphosate and AMPA were 3.26 (range = 0.4–8.8) mg/kg and 5.74 (0.7–10) mg/kg, respectively. There were no residues of glyphosate or AMPA in the conventional or organic soybean samples.

The glyphosate residues from both studies (Arregui et al., 2004; Bøhn et al., 2014) are less than the current European MRL and U.S. Tolerances for glyphosate in soybeans (20 mg/kg)1 and are within the values reported by EFSA in multiple years of market basket surveys (Table 2).

Soybeans have been tested using validated analytical methods in surveys conducted by several regulatory agencies from around the world. Glyphosate was detectable on soybeans in 7% of the samples collected in the surveys conducted by EFSA (EFSA, 2018, 2019, 2020) (Table 2). Other commodity grains that are approved for import as GT crops into Europe (corn and rapeseed) did not have detectable glyphosate (Table 2). The U.S. EPA survey specifically tested crops due to public interest in GT corn and soybeans and there were no residues greater than the tolerances established by the EPA (FDA, 2018b, 2019). Glyphosate was detected in 61% of soybean samples for the 2 years of testing (Table 3). The USDA Pesticide Data Program (USDA, 2020) also conducts pesticide testing of foods related to infants and they tested 300 soybean samples grown in Missouri from 2010 to 2011. Glyphosate was detected in 90% of the samples and the average amount of glyphosate was 1.94 mg/kg, with a maximum of 18.53 mg/kg. The EPA established tolerance for glyphosate on soybeans is 20 mg/kg. A survey conducted by regulators in Canada tested 28 samples of fresh or frozen soybeans and found one sample with glyphosate at a level less than the MRL; however, there is no indication as to whether any of these samples were conventional or GT varieties (Kolakowski et al., 2020).

3.2 Corn (maize)

Glyphosate residues in corn have not been quantified in regulatory surveys conducted by EFSA and CFIA. Detection of glyphosate occurred in 58% of samples from the surveys conducted in the United States where GT corn is grown (FDA, 2018b, 2019); however, the average detectable amount was 0.07 mg/kg, which is 70 times less than the EPA tolerance.

3.3 Other crops

In addition to the glyphosate residue levels summarized above for two major crops, soybean and maize, residue levels are available for other crops, including fruits, vegetables, and other grains, including regulatory agency surveys (EFSA, 2018, 2019, 2020; Kolakowski et al., 2020; Zoller et al., 2018). An early government agency report about analyzing foods for glyphosate was conducted on samples collected in 2010 and 2012 in Germany (Scherbaum et al., 2012). Information about the assay method or LOD and LOQ is not provided. They analyzed 124 grain samples (not including corn) reported that two samples (millet and rye crispbread) had detectable residues, both of which were below the MRL. In general, fruits and vegetables had the lowest detections of glyphosate compared to other food categories, and if detected, values were far less than the MRLs. The crops with the highest percentages of detections were lentils, beans, and wheat. Lentils had the most detections (value > LOQ) in the European surveys, with glyphosate detected in 34% of samples with 0.4% exceeding the MRL. Exceedances for buckwheat and honey are calculated from MRLs that are not derived by field data but instead from the LODs of the assays, which result in a significantly lower MRL. For instance, the field-derived tolerance for cereals in the United States is 30 mg/kg compared to the assay-derived value in the EU, which is 0.1 mg/kg. In Canada (Kolakowski et al., 2020), glyphosate was most prevalent in wheat, pulses, barley, and oat products and was of low prevalence in fresh fruits, vegetables, and soy products. In general, glyphosate was less prevalent in organic food than conventional.

3.4 Animal products (milk, meat, and eggs)

Animal products, with the exception of kidney and liver due to their physiological functions, are not expected to contain meaningful residues of glyphosate when animals are fed crops cultivated with glyphosate at labeled application rates. This is because glyphosate has a high water solubility (10.5 g/L), a low octanol-water partition constant (log POW = −3.2), and is rapidly excreted via the kidney (BfR, 2015; Bus, 2015). Three reports warrant detailed review because they generated considerable online discussion (Honeycutt et al., 2014; John & Liu, 2018; Samsel & Seneff, 2017). The first, an internet report, was about human milk, but it is reported here because the physiology of milk synthesis is similar across mammals and because milk is a considerable source of nutrients for infants, children, and neonatal mammalian animals. That report said that glyphosate was found in three samples of human milk (Honeycutt et al., 2014) based on the ELISA kit that was not validated for a complex matrix like milk. A subsequent study demonstrated that glyphosate was not detected in human milk (McGuire et al., 2016) and it was accompanied by a report (Jensen et al., 2016) of a validated assay that was significantly more precise for milk than the assay used by Honeycutt et al. (2014) (LOD = 1.0 vs. 75 µg/L), respectively. In order to not confuse whether their subjects actually had been exposed to glyphosate, McGuire et al. also confirmed that nearly all of their subjects were exposed to glyphosate at the time that milk samples were collected by sampling urine and milk from each subject simultaneously. Subsequent reports have confirmed that glyphosate is not detectable in human and bovine milk (EFSA, 2018; Ehling & Reddy, 2015; FDA, 2018b; NZ Ministry for Primary Industries, 2012; Steinborn et al., 2016; von Soosten et al., 2016; Zoller et al., 2018). Ehling and Reddy (2015) also demonstrated that glyphosate and AMPA were not detected in bovine whole milk powder.

John and Liu (2018) used the same ELISA kit to test several foods, including locally purchased milk and beef. Glyphosate was detected in organic milk, conventional milk, beef, and fish at 0.442 µg/L, 0.533 µg/kg, 0.553 µg/kg, and 1.495 µg/kg, respectively. The absence of data on ELISA validation in this report (John & Liu, 2018) for these varied and complex samples makes the results from this study difficult to interpret. Although this report has only marginal information on how the analyzed samples were collected, the reported data are also inconsistent with results reported by other studies with meat, milk, and egg samples in which validated assays and more complete information on sample collection conditions are reported (EFSA, 2019; FDA, 2018b). Lastly, it is noteworthy that the ELISA kit has been shown to generate false positives in milk (Bus, 2015; Chamkasem & Vargo, 2017; McGuire et al., 2016).

The third report that needs detailed review relative to the topic of glyphosate in meat or other animal tissues is by Samsel and Seneff (2017). This report was apparently based on their interpretation of metabolism studies that utilized radiolabeled glyphosate. They speculated that glyphosate was synthetically incorporated into proteins and suggested that it is “hidden” in protein molecules resulting in numerous maladies. For the hypothesis presented by Samsel and Seneff (2017) to be consistent with the mechanism through which mammalian cells synthesize proteins, glyphosate would need to substitute for glycine. However, it has been shown by Antoniou et al. (2019) that glyphosate does not substitute for glycine in metabolically active mammalian cells. Perhaps, the speculative interpretation of the metabolism study results by Samsel and Seneff (2017) comes from their interpretation of a study (EPA, 1994) in which a small percentage of the 14C-radiolabel was measured in bone and muscle fractions. However, detections of 14C in these samples were not molecularly characterized to confirm the author's speculations. Additionally, given the very low detected levels of 14C-radiolabel and the timeframe, the data most likely are the result of transient binding of 14C-glyphosate to calcium in the bone as glyphosate is known to weakly bind minerals, like calcium (Williams et al., 2000). This is not an indication that glyphosate was incorporated into a protein.

Glyphosate has not been found in milk, eggs, or fish from seven regulatory agency surveys, all of which used validated assays (EFSA, 2018, 2019, 2020; FDA, 2018b, 2019; Kolakowski et al., 2020; Zoller et al., 2018). One exception for the lack of glyphosate in animal products is that glyphosate was above the LOQ in three samples of sausage or meat loaf in the study of Zoller et al. (2018), but there was no information about whether these samples were 100% meat or if they contained plant-based fillers.

3.5 Summary about raw agricultural commodities

Regulatory surveys are an important tool to ensure postmarketing phytosantitary vigilance and play an important role in regulatory risk assessment. Since approved pesticide labels establish the legal limits that prevent unacceptably high residue levels (Winter et al., 2019), these survey studies monitor foods for pesticide residues to assess whether pesticide applicators are following the label requirements. In the vast majority of commodity analyses, the observed residue levels are below the legal threshold suggesting that most farmers are not exceeding product label rates and that most harvested crops are legal for trade (EFSA, 2018, 2019, 2020; FDA, 2018b, 2019). Assays conducted as part of the regulatory agency surveys are validated in order to reliably and accurately measure the target analyte(s), given the variety and complexity of the many raw agricultural commodities included in these surveys. The MRL is the highest level of a pesticide residue that is legally tolerated in or on food or feed when pesticides are applied correctly. The MRL is based on residue levels obtained in supervised field trials under GLP where the product is applied according to the proposed label instructions. Prior to formally setting the MRL, it is subject to a risk assessment in which it is compared along with other exposures with health-based guidance values. The ongoing surveys of glyphosate residues beginning with samples collected in 2010 (EFSA, 2013) and most recently with samples from 2018 (EFSA, 2020) suggest a high degree of compliance and provide no indication that residues have exceeded regulatory levels. Residue studies by nonregulatory agency investigators are helpful additional data beyond findings of regulatory agencies; however, the pesticide data need to be collected with detailed, validated methodology that generate reliable results.


This section reviews foods that are derived from products that were derived from crops sprayed with glyphosate during cultivation. Although it might be reasonable to expect glyphosate residues in processed foods to reflect the levels in their raw agricultural commodities, the washing and removal of the outer coating of grains during food processing would be expected to reduce residues. Isolation of portions of foods/feeds that contain glyphosate, such as hulls, or by drying might increase the residue concentration.

4.1 Soy products

Most soybeans are not consumed directly by people, instead they undergo various processes to produce different soy fractions, such as soy protein concentrate or isolated soy protein. Rubio et al. (2014) used the ELISA kit to test soy fractions and they reported that the assay was validated for each matrix, but details of the validation information were not provided. Also, the reported LOQ for several substances is the same as the LOQ reported for qualitative measurement of glyphosate in water according to the kit instructions, suggesting that the LOD was not established independently for the matrices tested. Glyphosate was detected in eight out of 28 (36%) soy sauce samples. Concentrations were greater than the method LOQ (75 µg/L) with a range between 88 and 564 µg/L and a mean of 242 µg/L. They also claim that glyphosate was not detectable in soy milk or tofu (Rubio et al., 2014). Rodrigues and de Souza (2018) tested 10 brands of soy-based infant formulas in Brazil using a validated HPLC method with a stated LOQ = 0.02 mg/kg. Multiple samples of these infant formulas were tested from 2012 to 2017 and they found that 2 of the 10 brands consistently had no detectable concentrations of glyphosate. The eight brands that tested positive were further assessed according to the degree in which the ingredients from soybeans were processed. Three brands were made from minimally processed soybean extract and had, on average, 1.08 mg/kg of glyphosate. The other five brands were made from the more highly processed soy protein isolate and had, on average, 0.11 mg/kg of glyphosate. It is reasonable to expect that soy protein isolate would have lower amounts of glyphosate compared to soybean extract since significant amounts of glyphosate should be removed during the additional water-based processes to produce soy protein isolate.

4.2 Breakfast cereal

A report of breakfast foods was issued by the Alliance for Natural Health (ANH-USA, 2016) in which a variety of products, including grain-based foods, coffee creamer, and eggs were tested for the presence of glyphosate residue. Despite the variety of matrices, all testing was done with the Abraxis ELISA kit and it appears that most foods were tested assuming the kit-sensitivity of 75 µg/L, but it is unclear whether this value was used as an LOD or LOQ. Likewise, the authors report that glyphosate was detectable in eggs, both conventional and organic, and in “organic coffee creamer.” The eggs and organic coffee creamer had reported concentrations of glyphosate >100 µg/L, results that are not consistent with other reports, including regulatory market basket surveys using validated methods, that were unable to detect glyphosate in milk or eggs (EFSA, 2018, 2019; FDA, 2019). All of the other breakfast products that were reported as positive for glyphosate in the ANH-USA report contained wheat or oatmeal. A Swiss agency survey also tested breakfast cereals and 8 of 10 samples had glyphosate residues that were above the LOQ (Zoller et al., 2018).

4.3 Infant foods

Regulatory agencies tested foods that were categorized as infant food. In a survey conducted in Canada (CFIA, 2017), there were 927 samples classified this way and 290 contained quantifiable glyphosate, none of which was more than the MRL (Table 4).

The Infant Total Diet Study (iTDS) conducted in France (ANSES, 2016) surveyed dietary ingredients specifically intended for infants. Modeling intake of infants is critical because total food consumed per unit of body weight is usually greater for children (<6-year old) than for adults, and because children and infants typically consume a smaller variety of food types. This study tested 500 substances from foods prepared as consumed. Glyphosate was detected in both of the two tested breakfast cereal samples.

The Australian and New Zealand Regulatory Authority purchased replicate samples of food items that were analyzed for many agricultural chemicals. Some chemicals, including glyphosate, were tested in a subsample of the foods that were targeted as suspected of being likely sources of exposure to agricultural chemical residues (FSANZ, 2019). Among the subsample of foods tested for glyphosate residue were rice-based, breakfast cereal and mixed infant cereal. For both foods, the percentages of detects (values > limit of reporting) were 25% and mean concentrations of residues were 0.006 mg/kg. Other regulatory agencies did not detect glyphosate in baby food (Zoller et al., 2018) nor in milk-based infant formula (Kolakowski et al., 2020); however, these are difficult to compare because they could contain different ingredients and source of grains, and processing methods could have an impact on residues.

4.4 Ice cream and sugar

The Organic Consumers Association reported an analysis of glyphosate in various flavors of Ben & Jerry's (Unilever, South Burlington, VT) premium (high-fat) ice cream (OCA, 2017). The LOD and LOQ were reported to be 0.05 and 0.25 µg/L, respectively, but no other information about validation was provided, including the assay method. All flavors except one had detectable amounts of glyphosate, but concentrations of glyphosate were not reported. Cream and skim milk are principle ingredients in ice cream and as discussed above milk has consistently been shown not to contain detectable amounts of glyphosate (EFSA, 2018; Ehling & Reddy, 2015; FDA, 2018b; NZ Ministry for Primary Industries, 2012; Steinborn et al., 2016; von Soosten et al., 2016; Zoller et al., 2018). The Canadian survey (Kolakowski et al., 2020) also tested samples of plain yogurt and plain custard and did not detect glyphosate. Furthermore, the milk fat levels, that are higher in premium ice creams compared to other lower fat brands, would make it even less likely that the milk ingredient would be the source of glyphosate since glyphosate is not lipophilic (Bus, 2015; Shelver et al., 2018). Vanilla ice cream is the simplest of the recipes that tested positive. Besides milk, the only other ingredients currently listed for Ben & Jerry's vanilla ice cream are sugar, egg yolk, water, guar gum, vanilla extract, vanilla bean, and carrageenan. None of these ingredients are expected to contain glyphosate. Even sugar derived from either sugarcane or GT sugar beets would not be predicted to contain glyphosate, due to the water solubility of glyphosate (Williams et al., 2000) that would reduce residues during the water-based processing to crystallize food-grade sugar. A recent study used LC-MS/MS and determined that, although glyphosate residues are found in GT sugar beets, due to the water processing steps for crystalizing sugar, this herbicide is not detectable in the final product (Barker & Dayan, 2019).

Combining the lack of information on the analytical method used to measure glyphosate levels with the lack of any ingredients in this premium ice cream product being a plausible source of glyphosate residue, the scientific validity of this report is questionable.

4.5 Honey and sugary syrup

There are several reports about glyphosate in honey and presumably the glyphosate would be from honey bees foraging for nectar from plants exposed to glyphosate. These reports vary in where samples were collected and assays utilized. The ELISA method was used to test honey for glyphosate without any modification or validation of the assay (Rubio et al., 2014). The authors reported that 41 of 69 honey samples had quantifiable amounts of glyphosate (range of 17–163 µg/kg). The highest reported value of 163 µg/kg from this nonvalidated assay would exceed the EU MRL for honey, which is 50 µg/kg (European Commision, 2020a), while the U.S. EPA does not have a tolerance for honey (CFR, 2019). They also tested pancake and corn syrup and said that glyphosate was not detectable in these highly processed sugar-based items.

FDA scientists developed an LC-MS/MS assay to detect glyphosate in honey (Chamkasem & Vargo, 2017). The validated assay had an LOQ = 16 µg/kg, and 9 of 16 samples bought from a local market had glyphosate >LOQ. Of these, the median concentration of glyphosate was 26 µg/kg with a range of 17–121 µg/kg.

Glyphosate was measured in honey from hives on the island of Kaua`I, Hawaii in five “batches” (Berg et al., 2018). The different collections were as follows: (1) two hive samples in fall, 2013; (2) 36 hive samples during the summer of 2015; (3) 21 hive samples in fall, 2016; (4) 21 retail samples in winter, 2013; and (5) three retail samples in fall, 2016. All analyses were done using the ELISA through three different labs depending on the batch. The authors state that 14 samples were tested with both the ELISA and an LC-MS/MS method; however, no validation information is provided. They reported that glyphosate was quantifiable in 27.1% of samples, which are not independent samples, and the mean value was 118.3 µg/L.

An assay was developed and reported by Pareja et al. (2019) using ion chromatography coupled to Q-Orbitrap MS. Assay validation information is reported with an LOQ of 0.005 µg/g. They reported that the assay had a medium matrix effect. Sixteen samples of honey taken from hives, and 16 commercial samples from South America and Europe, were tested and 81% of the samples had detectable concentrations of glyphosate. Half of the honey samples with detectable glyphosate had concentrations greater than the European MRL of 50 µg/kg. Concentrations of glyphosate in the detectable samples were not reported, and AMPA was not detected in any sample.

Another study using derivatization prior to solid phase extraction coupled to LC-MS/MS was used to analyze 200 randomly selected honey samples from those submitted to the lab (Thompson et al., 2019). There was no indication as to whether all of these were independent samples. Validation information was provided, and glyphosate was detectable above the LOQ of 1 µg/kg in 197 of the samples. The greatest concentration of glyphosate was 49.8 µg/kg and they indicated that there were no samples with both glyphosate and AMPA greater than the LOQ. They provide a scatter plot of both analytes, and the ratio of glyphosate to AMPA appeared to vary anywhere from approximately 10/1 to 1/10. There is no explanation for why these would vary to this extent. The source of all AMPA is not necessarily derived as a metabolite of glyphosate as another source may be from degradation of phosphonates used as detergents (Botta et al., 2009).

Honey was tested in two regulatory agency surveys and had detectable glyphosate in 9% of samples (EFSA, 2018, 2019, 2020), which is much less than detections in other studies. LOQs reported by the different countries that tested these honey samples ranged from 0.01 to 0.14 mg/kg. Whereas, in the Swiss survey, honey had detectable glyphosate in 15 of 16 samples (LOQ = 0.001 mg/kg) and with a mean concentration of 0.0046 mg/kg (Zoller et al., 2018). None of the honey samples had detectable amounts of AMPA and 10 honey samples had glyphosate values greater than the LOQ for AMPA, suggesting that for these samples, if AMPA was present, it would not be greater than the concentration of glyphosate. This is inconsistent with the data of Thompson et al. (2019), in which many of the 200 samples had more AMPA than glyphosate. One noteworthy aspect of the publication of Zoller et al. (2018) is that individual LOQs are provided for each food category and for each of the glyphosate and AMPA assays. Therefore, it is meaningful that this validated study showed that the concentration of AMPA was never greater than 42% of the value for glyphosate when both glyphosate and AMPA were greater than or equal to their respective LOQs. This is in contrast to other studies that did not report validation, and occasionally reported AMPA levels that were greater than glyphosate.

4.6 Beer and wine

The Munich Environmental Institute (Guttenberger & Bear, 2016) tested German beers using the ELISA kit method. The reported LOD is 0.075 µg/L, a multiple of the LOD reported by the kit manufacturer for analysis of glyphosate in water. However, Guttenberger and Bear (2016) do not include assay validation data that compare the assay performance between water and beer samples, and yet beer includes alcohol and other constituents that would reasonably be expected to affect ELISA methods. They indicated that all 14 samples had detectable glyphosate with a value ranging from 0.46 to 29.74 µg/L. In 2017, this same group conducted a new round of beer testing, which was reported in the media, but a published report is currently not available online. The German Federal Institute for Risk Assessment (BfR, 2017) issued a statement about these results in which they indicate that detection of glyphosate in beer is not unexpected. Based on these values, the BfR concluded that, regarding glyphosate, an adult would be able to drink 1000 L of beer per day with a reasonable expectation of safety.

The California Public Research Group (Cook, 2019) reported concentrations of glyphosate in five brands of beer and 15 brands of wine. However, interpretation of the results in this report is difficult for two reasons. First, the report is unclear about the assay method employed, because it cites an LC-MS/MS assay method but then gives detailed steps for using the ELISA method. Additionally, independent of which assay was actually used, there was no validation or LOD/LOQ reported. They reported that 19 of 20 tested beers and wine had detectable glyphosate. For nonorganic wine, values ranged from 36.3 to 51.4 µg/L. Nonorganic beers ranged from 9.1 to 49.7 µg/L. Two brands of beer and two brands of wine labeled as organic were tested and one beer and both wine samples were reported to have glyphosate at 5.7, 5.3, and 4.8 µg/L, respectively. Considering that grapes are almost exclusively the whole-food ingredient used in many wines, it is reasonable to conclude that grapes would be the source of glyphosate in wine but because grapes are sensitive to glyphosate they are not sprayed directly. But glyphosate is used between the rows of vineyards because it is not taken up into the vines through the roots. It would have to enter plants predominantly by droplets that get onto the leaves through overspray. Also, it would have to be in an amount that is low and does not kill the plant. Therefore, for glyphosate residues to be present in grapes used to make wine, they would be expected to be very low. Even though grapes are not sprayed directly, there is an MRL for wine grapes in the EU, which is 50 µg/kg. The most common grains used for making beer are barley, wheat, maize (corn), and rice, all of which are crops that can be sprayed with glyphosate preplanting and corn is the only one of these available as GT.

In the Swiss food survey that reported values from a validated assay, beer had detectable amounts of glyphosate in only 2 of 15 samples (median value was <0.5 µg/kg) but glyphosate was detectable in 100% of wine samples (3.1 µg/kg). AMPA was not detectable in any beer sample and was detectable in 4 of the 21 wine samples.

Notwithstanding the questions about the validity of the analytical methods and the agronomic source of glyphosate residues, it is reasonable to conclude that glyphosate has been detected in beer and wine samples, and residues on grains used for beer or grapes for wine are the probable sources.

4.7 Enteral formula

A website (Seneff, 2015) reported testing of 20 samples from a single batch of a commercial enteral formula using the ELISA kit at a commercial lab with a stated LOD = 75 µg/L. As stated previously, this value is provided by the kit for water analysis and there was apparently no validation reported for this lab and sample matrix. Although the samples came from the same batch, 6 of the 20 samples had detectable glyphosate (ranged from 80 to 111 µg/L).

4.8 Summary about reports on processed food

The presence of pesticide residues in foods is regulated by global authorities to ensure food safety by establishing realistic levels of human exposure. Therefore, while reports of the detection of glyphosate residues in certain processed foods can be predicted, as long as the levels are below the established legal limits, food safety is ensured. However, interpreting reported values can be difficult when assays are not validated or there is an unstated application of an appropriate LOD. However, if assuming these reported values for glyphosate are accurate, most values do not approach the MRL. Regulatory Agency surveys have tested many of these same, or similar, processed foods and provide more reliable information. One food item that seems worthy of additional consideration is honey, for which a disproportionate number of samples seem to approach, or exceed, the MRL. More detailed studies with validated assays would help address the current high variability in glyphosate residue levels from the many reports with honey summarized above, since some of the variability might be due to the rapid decline in glyphosate in nectar over time (Thompson et al., 2014).


One key goal for regulatory authorities is ensuring food safety by modeling realistic levels of human exposure to a given pesticide that provide a “reasonable certainty of no harm” for the consumer (Winter et al., 2019). Given that different approaches to exposure modeling exist (see below for examples), after experimental data have established an upper boundary for the level of a given chemical (e.g., pesticides like glyphosate) that might result in an adverse effect, regulatory authorities integrate into their health risk assessment the concept of an “uncertainty factor” (Herrman & Younes, 1999). The desire to use uncertainty factors is founded on the expectation that reducing the potential for exposure by an appropriate “factor” (for the ADI for glyphosate, the uncertainty factor is 100), will yield the required “reasonable certainty of no harm” from exposure to the given chemical for the majority of the population, including vulnerable subgroups. However, consumers have typically looked to health care professionals, nutritionists, and dietitians for their views on whether these regulatory processes ensure the safety of foods from agricultural use of pesticides (both conventional and organic). As mentioned above, the current training curricula for the professions that consumers look to for guidance have limited, if any, coursework on relevant topics, so it is critical to have these professionals become well-informed on how residues relate to overall food consumption and safety (Sanborn et al., 2019; Temte & McCall, 2001). One key concept critical to an understanding of how regulatory processes ensure food safety is the empirical data for pesticide residues that are used by regulatory authorities to establish MRLs and tolerances related to the legal application of the pesticide in agriculture. Importantly, a single residue measurement is insufficient to conclude safety (or imply lack of safety) since a single residue level is only a part of this larger assessment. Notably, since 1995, the Environmental Working Group has annually released reports on pesticide residues in fruits and vegetables (EWG, 2020). These reports rank fruits and vegetables for the presence of residues but they do not provide consumption estimates in comparison to the amounts that are determined to be safe by regulatory agencies. Winter and Katz (2011) responded to this report by calculating from these same data that all exposures were less than 1% of the RfD and most were <0.01% of the RfD. This illustrates the importance of an assay needed to quantify residue accurately. To be meaningful to discussions of food safety, it is critical that reports of “detections” of a residue are put into the context of consumption and toxicity. This section will review published risk assessments that compare dietary exposure of glyphosate to the ADI and these studies are summarized in Table 5.

TABLE 5. Modeling for chronic dietary consumption of glyphosate
References Residue Processing < LOD Food consumption source ADI or RfD(mg/kg/day) Exposure (% ADI or RfD)
EPA (2017) 100% of crop treated and all food at tolerance amounts Default processing factors N/A Daily food intakes from DEEM 1 9%
EFSA (2019) From residue testing Assumes residue = 0 0.5 0.15%
From residue testing Assumes residue = LOD 0.5 0.24%
FSANZ (2019) Mean concentrations from testing Prepared ready-to-eat Assumes residue = 0 2011–2012 Australian National Nutrition and Physical Activity Survey 0.3 <1%
Zoller et al. (2018) Median residues from testing 0.5 0.5%
Max residues from testing 0.5 <10%
FAO/WHO (2019) Median residues from supervised trials Median residues from supervised trials for processed foods GEMS 1 2.7%a
Stephenson and Harris (2016) All foods at EU MRL Uses commodity residues (i.e., sugar beet for table sugar) UK toddler 0.5 80.1%
EU MRL or median residues from supervised trials Uses commodity residues (i.e., sugar beet for table sugar) UK toddler 0.5 16.8%
EU MRL or median residues from supervised trials or monitoring residues Processing factors Irish adult 0.5 2.1%
Actual residues (barley residues matched to commodity intakes) Processing factors Irish adult 0.5 1.2%
From testing, unmeasured residues = 0 Processing factors Assumes residue = 0 Dutch adult 0.5 0.03%
From testing, unmeasured residues = MRL Processing factors If <LOR residue = LOR Dutch adult 0.5 0.58%
  • a Value represents the median of the 17 GEMS Cluster Diets (G01–G17); range across all 17 cluster diets is 0.7–4.2% of ADI.
  • Abbreviations: ADI, acceptable daily intake; LOD, limit of detection; LOR, limit of reporting; RfD, reference dose.

5.1 EFSA assessment

As part of EFSA's most recent report on pesticides in food, dietary assessments of glyphosate were conducted (EFSA, 2020). EFSA conducted a short-term dietary assessment for glyphosate that examined the impact of eating large portions of unprocessed foods that contain the highest reported residues. They concluded that this type of high-intake, short-term dietary exposure to glyphosate would not be expected to be of concern for consumer health. The chronic risk assessment is meant to predict lifetime exposure. EFSA used two approaches for how they dealt with concentrations of glyphosate in a food that were <LOQ: (1) a more conservative approach in which the concentration of glyphosate was set to the LOQ; and (2) a less conservative approach in which concentrations of glyphosate were set to zero. Using these approaches, they calculated that chronic exposure to glyphosate based on assumptions of these models ranged between 0.06% and 0.26% of the ADI. Even though these values excluded residues in drinking water, this exposure is over two orders of magnitude below the ADI and it provides a reasonable certainty that there would be no-harm.

5.2 EPA assessment

EPA (2017) conducted a highly conservative dietary risk assessment that assumed residues of all commodities are at the tolerance, 100% of a commodity is treated, and the default processing factors from the Dietary Exposure Evaluation Model (DEEM Version 7.81) were used. Residue amounts at the tolerance level for each food are multiplied by the daily intake of that food, and these are summed to get the estimated residue intake. For the U.S. general population, the conservative estimated intake was 0.09 mg/kg/day, which is 9% of the U.S. RfD of 1.0 mg/kg/day. The same conservative level of intake would be 18% of the EU ADI.

5.3 Swiss study

In the Swiss study (Zoller et al., 2018), dietary exposure was calculated and the authors considered that this was done conservatively by choosing food consumption values that overestimate actual daily average consumption. They compared individual food categories using median or maximum residue values. If median residues were used, they concluded that individual food categories were less than 0.5% of the ADI/ARfD. The ARfD (acute reference dose), established by the EU, is the amount of a substance that can be consumed in a single meal or over a 24-h period without appreciable health risk. If maximum values were used, all food categories were <10%. They concluded that these residues for any food category were not a cause for a health concern.

5.4 Australia Total Diet Study

The Australia Total Diet Study (FSANZ, 2019) estimates the exposure of the general Australian population to agricultural chemicals, including glyphosate. This modeling used the following conditions: (1) food samples were prepared ready-to-eat; (2) mean concentrations of glyphosate in samples were used; (3) concentrations of glyphosate in food items that were <LOR (limit of resolution) were assumed to be zero; and (4) food consumption data were used from the 2011 to 2012 Australian National Nutrition and Physical Activity Survey. For dietary exposures to glyphosate, they determined that the estimated mean and 90th percentile exposures were 0.022–0.083 µg/kg BW/day and 0.048–0.17 µg/kg BW/day for all age groups 2 years and above, respectively. These values for all age groups were <1% of the Australian ADI, which is 0.3 mg/kg.

5.5 Joint FAO/WHO Meeting on Pesticide Residues

The Joint FAO/WHO Meeting on Pesticide Residues (JMPR) (FAO/WHO, 2019) considered acute and long-term dietary exposures to glyphosate. The 2011 JMPR concluded that an ARfD for glyphosate was not necessary and this meeting likewise concluded that the acute dietary exposure to residues of glyphosate is unlikely to present a public health concern.

For the long-term assessment, the median residues from standardized trials (STMR), or the STMR for processed foods were used, and intakes were calculated for 17 regional cluster diets defined by the Global Environment Monitoring System (GEMS) (WHO, 2020). These calculations assume an ADI of 1 mg/kg of body weight and the daily intake of glyphosate was estimated to range between 0.7% and 4.2% of the ADI (median = 2.7%).

5.6 Deterministic and probabilistic modeling

Modeling of acute exposure to a chemical, like glyphosate, typically follows one of two approaches: deterministic or probabilistic. Deterministic approaches establish an exposure for a single consumer from a single high-residue food with a high-level of consumption (97.5th percentile), while probabilistic approaches (i.e., Monte Carlo Risk Assessment model) establish a distribution of acute exposures for different populations from all sources of the chemical. Both deterministic and probabilistic approaches were used to model exposure to glyphosate in the EU (Stephenson et al., 2018). The deterministic method used either the highest residue value or median residue for a commodity, or when not available, the MRL. Using this conservative method, the majority of foods had glyphosate levels <5% of the ARfD. Moreover, foods with glyphosate levels predicted to be above 5% typically were raw agricultural commodities that are subsequently cooked or processed in a way that often results in a lower residue (Barker & Dayan, 2019; Kolakowski et al., 2020; Stephenson et al., 2018; Williams et al., 2000). By comparison, probabilistic modeling of exposure to glyphosate was done for the Dutch adult and child based on empirical monitoring data collected between 2011 and 2014 in the UK (Stephenson et al., 2018). Probabilistic exposure models account for variations in consumption patterns and variations in residue concentrations in order to estimate optimistic and pessimistic exposure scenarios according to EFSA guidance (EFSA Panel on PPR, 2012). Using these assumptions and scenarios, all individuals had exposures to glyphosate of <10% of the ARfD. They concluded that acute dietary exposure to glyphosate is unlikely to represent a concern.

Probabilistic exposure models account for variations in consumption patterns and variations in residue concentrations in order to estimate optimistic and pessimistic exposure scenarios according to EFSA guidance (EFSA Panel on PPR, 2012). Using these assumptions and scenarios, all individuals had exposures to glyphosate of <10% of the ARfD. They concluded that acute dietary exposure to glyphosate is unlikely to represent a concern.

Similar deterministic and probabilistic approaches were taken to model chronic exposures to glyphosate (Stephenson & Harris, 2016). The worst-case deterministic model was the theoretical maximum daily intake (TMDI). The TMDI approach combines the average daily regional consumption levels with the MRL for each commodity and sums these intakes for all commodities. It indicated that the maximum exposure was approximately 80.1% of the EU ADI (0.5 mg/kg BW/day). As in other models, foods that accounted for much of the glyphosate intake are known to overestimate realistic exposures, such as glyphosate residue levels in whole sugar beets instead of refined sugar; and, milk and cream that have not been shown in market surveys to contain detectable glyphosate. These commodities accounted for 70% of the glyphosate consumption derived by using the TMDI approach. In refined chronic dietary models that used processing factors and STMR values or actual residues, the exposure is progressively reduced to 1.2% of the ADI (Table 5).

Probabilistic modeling was also performed using EFSA guidance on chronic dietary exposure. When there were no monitoring data or unmeasured residues in animal commodities, the pessimistic scenario used MRL values, while the optimistic scenario used a zero. Treatment of residues less than the LOR was set at the LOR for the pessimistic scenario and set to a zero for the optimistic scenario. Using this approach, the 99.9th percentile exposures for adults were determined by the pessimistic and optimistic scenarios to be 0.58% and 0.03% of the ADI, respectively. For the children aged 2–6 years old, exposures were 0.90% and 0.13% of the ADI, respectively.

One report analyzed estimated maternal exposure to glyphosate using two different approaches (McQueen, Callan, & Hinwood, 2012). One approach measured glyphosate in composited aliquots of food consumed for a 24-h period and the weight of this food that each subject consumed. This was done for each of 20 pregnant women in Australia and the assay used was an LC-MS/MS method with an LOD and LOQ of 0.01 and 0.005 mg/kg, respectively. They calculated that the average dietary consumption of glyphosate + AMPA was 0.4% of the Australian ADI (0.3 mg/kg/day). For the second approach, 43 pregnant women were used and exposure to glyphosate + AMPA was determined by recording for 24 h weight and type of food consumed. These amounts were multiplied by the MRL for each of the food types. Using this approach, mean dietary exposure was 4% of the ADI. Similar to modeling studies cited above, regardless of the calculation used (actual residues vs. MRL), both approaches resulted in estimates that were at the low end of the ADI and the method using actual residues was 10 times less than the MRL method (McQueen et al., 2012).

5.7 ANSES Infant Total Diet Study and risk assessment

A study evaluated the impact of chemicals, including glyphosate, in the diets of infants less than 3 years old (Hulin et al., 2014). The conclusion of the study was that the ADI for glyphosate was not exceeded for this population and, using the most conservative assumptions, the 90th percentile consumption was 0.6% of the ADI. This conclusion was based on an older ADI for glyphosate of 0.3 mg/kg BW for the EU that was later increased to the current 0.5 mg/kg BW, suggesting that the difference between modeled intake and the current ADI is now greater.

5.8 Summary about risk assessments

Unlike the warnings from the Environmental Working Group released annually to consumers about produce that cites detection of residues, risk assessments conducted by regulatory authorities put these residue measurements in context by using dietary models that estimate a daily intake that can then be compared to previously established daily exposure thresholds that have a reasonable certainty of no harm. Although risk assessments for glyphosate follow this general approach, each uses various assumptions to test the robustness of their conclusions. The dietary modeling for risk assessments for chronic consumption of glyphosate and some of their assumptions are summarized in Table 5. The six published assessments, five being conducted by regulatory authorities, all reach a conclusion that, regardless of the assumptions being used, exposure of people to glyphosate is less than the ADI, which includes infants, children, and adults. These results are important since it has been suggested that the use of glyphosate has intensified as a result of increased use in recent years due to the adoption of Roundup Ready crops (Benbrook, 2016), creating a perception that exposure is unexpectedly too high. The results from these assessments by regulatory agencies would indicate that exposure to glyphosate is within amounts that are considered to cause no harm.


6.1 Metabolism of glyphosate and presence in urine

The prior sections summarize studies that directly measured glyphosate residue in a variety of sources, including a number of raw and processed foods, which are then used to model dietary exposure and compare these results to regulatory thresholds that ensure public safety. However, exposure to glyphosate can also be assessed by measuring the levels of metabolic excretion of ingested glyphosate. Approximately 20% of ingested glyphosate in mammals is absorbed from the gastrointestinal tract into the circulatory system (Niemann et al., 2015). The half-lives for an oral dose were approximately 6 h for the α phase (distribution) and β phase (elimination), which indicates rapid clearance and poor absorption (FAO/WHO, 2017b; Williams et al., 2000). Virtually, no absorbed glyphosate is metabolized, and it is cleared from the body by the kidneys (Niemann et al., 2015). Consequently, the presence of glyphosate in urine is not unusual, and detection of glyphosate in urine that is within the ranges observed with typical exposures is not indicative of a health risk. Instead, glyphosate in urine reflects absorption and excretion levels that regulatory agencies have reviewed extensively and concluded is safe (BfR, 2016). Elevated glyphosate concentrations can occur when urine volume is reduced for various physiological reasons independent of exposure to glyphosate. Elevated levels of urinary glyphosate can also result from rare instances of high exposure to glyphosate, such as is documented in the Farm Family Exposure Study (Acquavella et al., 2004). Moreover, AMPA is often detected in urine, but since glyphosate is virtually not metabolized within the body itself, AMPA in urine is most likely the result of absorption of ingested AMPA, not metabolism of absorbed glyphosate (Niemann et al., 2015). Possible sources of AMPA are from consumption of foods from crops that have metabolized glyphosate (Reddy et al., 2004) and/or from detergents from a variety of sources that are known to contain AMPA (Botta et al., 2009).

As a result of the known characteristics of absorption, metabolism, lack of bioaccumulation, and excretion of glyphosate described above, total glyphosate ingestion can be estimated reliably by knowing the glyphosate concentration in urine. Comparing the estimated level of exposure to glyphosate from studies measuring urinary glyphosate levels, with the ADI for glyphosate is informative about safety.

For an adult human with a urine output of 2 L/day, the formula to estimate glyphosate ingestion in µg/kg/day is:
where Curine is concentration of glyphosate in urine (µg/ml) and Vurine is daily output of urine (ml/day). This section assumes 60 kg for an average body weight for this calculation based on Niemann et al. (2015).

This formula uses the reasonable assumption that ingestion is the primary route of exposure to glyphosate when tested individuals have not recently sprayed with glyphosate. If so, this formula provides an estimate of total glyphosate and does not rely on subjects estimating, often with low precision (IOM, 2000), their consumptions of different types of food or water and measuring, or estimating, glyphosate in these potential sources. Exposures of glyphosate as a percentage of the ADI or RfD estimated by this method can be compared to the exposures discussed in Section 5 of this paper as a way to determine if values from these two distinct methods corroborate one another (Tables 5 and 6).

TABLE 6. Glyphosate concentrations in urine samples (maximum values) and calculated estimates of ingestion, compared to acceptable daily intake (ADI) or reference dose (RfD)a
Author, year Analytical method (reported LOD and LOQ) Maximum reported urine value (µg/L) Calculated glyphosate ingestiona(µg/kg) % of U.S. reference dose (RfD, 1.0 mg/kg/day) % of EU acceptable daily intake (ADI, 0.5 mg/kg/day)
Brändli and Reinacher (2012) NR 2.0 0.33 0.03 0.07
Connolly et al. (2018) LC/MS-MS (0.5 µg/L, NR) 1.35 0. 23 0.02 0.05
Conrad et al. (2017) GC/MS-MS (NR, 0.1 µg/L) 2.80 0.47 0.05 0.09

Curwin et al. (2007a) Child (nonfarm)

Child (farm)

FCMIA (0.9 µg/L, NR)









Curwin et al. (2007b)


Child (nonfarm)

Child (farm)

FCMIA (0.9 µg/L, NR)













Fagan et al. (2020)



LCMS (0.02 µg/L, 0.1 µg/L)









Honeycutt et al. (2014) ELISA 7.5 µg/L, NR) 18.8 3.13 0.31 0.63
Hoppe (2013) GC/MS-MS (NR, 0.15 µg/L) 1.82 0.3 0.03 0.06
Jayasumana et al. (2015) ELISA (0.6 µg/L, NR) 80 13.33 1.33 2.67
John and Liu (2018) ELISA (0.075 µg/L, NR) 1.03 0.17 0.02 0.03

Knudsen et al. (2017)



ELISA (0.075 µg/L, NR)









Krüger et al. (2014) ELISA (NR, NR), GC/MS (NR, NR) 5 0.83 0.08 0.17
Krüger et al. (2016a) ELISA (0.07 µg/L, NR) 3.57 0.6 0.06 0.12
Krüger et al. (2016b) ELISA (0.07 µg/L, NR) 4.2 0.7 0.07 0.14
Markard (2014) GC/MS (NR, 0.15 µg/L) 0.65 0.11 0.01 0.02
McGuire et al. (2016)

LC/MS (0.02 µg/L, 0.1 µg/L)

1.93 0.32 0.03 0.06

Mesnage et al. (2012)


HPLC/MS (1 µg/L, 2 µg/L) 2.0 0.8c 0.2 0.4
Mills et al. (2017) HPLC/MS (0.03 µg/L, NR) 0.55 0.09 0.01 0.03
Parvez et al. (2018) LC/MS-MS (0.1 µg/L, 0.5 µg/L) 7.2 1.2 0.12 0.24
Rendon-von Osten and Dzul-Caamal (2017) ELISA (0.05 µg/L, NR) 0.48 0.08 0.01 0.02
Sauvage (2020) ELISA (NR, NR) 2.6 0.43 0.04 0.09

Sierra-Diaz et al. (2019)


HPLC/MS (NR, NR) 0.61d,e, d,e 0.20f 0.02 0.04
Soukup et al. (2020) LC/MS-MS (0.05 µg/L, 0.2 µg/L) 1.36 ND ND 0.13f

Trasande et al. (2020)


LC/MS (0.1 µg/L, 0.33 µg/L) 2.13c 0.85c 0.09 0.17
  • a Data are for adults unless noted otherwise. Glyphosate ingestion (µg/kg/day) = Curine * Vurine * 0.20–1 * BW–1. Where, Curine is concentration of glyphosate in urine (µg/ml), Vurine is daily output of urine (ml/day). Calculation is based on a 60-kg adult.
  • Exposure, calculated as a % of RfD and % of ADI, is all calculated using this formula and the max (when available) concentration of glyphosate in urine to enable direct comparisons between studies.
  • b Glyphosate ingestion calculation is based on a median value of 27.2-kg child (farm) and median value of 22.7-kg child (nonfarm) as reported (Curwin et al., 2007a).
  • c Glyphosate ingestion calculation assumes a 25-kg child.
  • d Mean value.
  • e Value for child calculated based on reported average child weight of 30 kg.
  • f This calculation uses actual body weights and 24-h urine volumes for each subject.
  • Abbreviations: AOEL, acceptable operator exposure level; ELISA, enzyme-linked immunosorbent assay; FCMIA, fluorescence covalent microbead immunoassay; G, glyphosate; GC, gas chromatography; HPLC, high-performance liquid chromatography; LC, liquid chromatography; LOD, limit of detection; MS, mass spectrometry; MS/MS, tandem mass spectrometry; ND, not determined; NR, not reported.

6.2 Exposure estimated from urine

6.2.1 Publications in journals

Niemann et al. (2015) provide a comprehensive discussion of using urine to determine exposure and put it into context by comparing it to health-based guidance. They reviewed seven reports of studies in which samples of urine from human subjects were collected and analyzed (Acquavella et al., 2004; Curwin et al., 2007b; Honeycutt et al., 2014; Hoppe, 2013; Krüger et al., 2014; Markard, 2014; Mesnage et al., 2012). However, it should be noted that Niemann et al. (2015) cite an unpublished report by Markard (2014), and apparently a subset of the data of Markard overlaps with the data reported by Conrad et al. (2017). Further complicating an effective review of these seven papers, two (Krüger et al., 2014; Mesnage et al., 2012) are in predatory journals, one is a nonpeer-reviewed laboratory report (Hoppe, 2013) and one is from an internet site that is also not peer-reviewed (Honeycutt et al., 2014). The two publications in predatory journals might not have gone through a rigorous scientific peer review (Beall, 2017). As noted by Acquavella et al. (2004), studies that do not complete peer-review (such as in the popular press or posted to the internet) are more likely to lack basic information about demographics and selection of subjects, sample handling procedures, and basic assay validation information critical to scientific interpretation of the results. Fortunately, there are several studies with reliable residue testing that provide concentrations of glyphosate in urine that enable comparison of the ranges of dietary exposure estimates, thereby avoiding the potential for error from those that lack validation. Studies that report urinary glyphosate are described below and are summarized in Table 6.

In the study by Curwin et al. (2007b), urine samples from farm and nonfarm households in Iowa were measured by immunoassay that used the Abraxis anti-glyphosate antibody with a reported LOD of 0.03 µg/L (Biagini et al., 2004). Urinary concentrations greater than the LOD were detectable in 65–88% of six groups (farm/nonfarm X father/mother/children). Mean glyphosate concentrations for the different groups ranged from 1.2 to 2.7 µg/L and urinary levels of glyphosate were unaffected by farm versus nonfarm individuals. In a follow-up publication (Curwin et al., 2007a), daily exposure to glyphosate for the samples from these same children was normalized by urinary creatinine, instead of estimated daily urine volume. The glyphosate exposures for both farm and nonfarm groups ranged from 0.013 to 0.34 µg/kg/day, and the authors concluded that the overall dose estimates did not exceed either the acute or chronic RfD for glyphosate. Hoppe (2013) examined 182 urine samples from Europeans. The data were not peer-reviewed but were supplied to the BfR. The analytical procedure was conducted by a GC-MS/MS with an LOQ = 0.15 µg/L. They indicated that 44% and 36% of samples had concentrations of glyphosate and AMPA in urine that were >LOD, respectively. Average and maximum concentrations were 0.21 and 1.86 µg/L for glyphosate and 0.18 and 2.63 µg/L for AMPA. The AMPA/glyphosate ratios were highly variable, and the authors suggested this might have been due to variable dietary exposures to other sources of AMPA, which is known to occur (Botta et al., 2009).

Details of one report (Honeycutt et al., 2014) were only provided on an internet site with insufficient information validating the methodology to qualify the data to be scientifically interpretable. Individuals provided samples for analysis using the commercial ELISA kit that was not validated for urine. Minimal information describing sample collection was included in this online report (e.g., age, gender, weight, and zip code). Instead of determining an LOD for this matrix, samples were diluted with water 100-fold and an LOD of 7.5 µg/L was assumed, which was 100 times the LOD for water provided by the ELISA kit's instructions. Thirteen of 35 samples of urine had glyphosate values that were greater than this nonverified LOD, and of the 13 detectable samples, the range was 8.1 µg/L (6-year-old male) to 18.8 µg/L (26-year-old female). Comparing this highest urinary glyphosate value to a health endpoint represents 0.63% of the EU ADI. This calculation is described in Table 6.

McGuire et al. (2016) tested urine of lactating mothers in two states, Idaho and Washington. Samples were tested using an LC-MS/MS assay (Jensen et al., 2016) validated for measuring glyphosate in urine samples with an LOD and LOQ of 0.02 and 0.1 µg/L, respectively. Samples were collected from 40 subjects and glyphosate was detectable in 37 and quantifiable in 29 of them. The highest concentration of glyphosate was 1.93 µg/L and the mean urinary concentration was 0.28 µg/L, considerably less than concentrations reported by Honeycutt et al. (2014). Based on their highest reported glyphosate concentration in urine using a validated assay, these experimental subjects were exposed to 0.06% of the ADI for glyphosate.

Krüger et al. (2014) reported results for glyphosate in urine using an ELISA that did not have validation information. Although the authors reported that the ELISA results correlated with another method, this is insufficient, by itself, to validate an assay. The authors did not provide actual values for glyphosate in human urine, but by extrapolation from a figure in the publication, it appears that the greatest glyphosate concentration was approximately 5 µg/L. The authors claimed that healthy humans had significantly less glyphosate in their urine than chronically diseased humans. They did not provide demographic information or criteria to define the healthy and unhealthy groups. Furthermore, the error bars around the mean glyphosate concentrations in urine from the health and chronically diseased individuals overlapped considerably, albeit the authors report a difference at p < 0.03 between these groups.

Sri Lankan agricultural nephropathy is a kidney disease that is most prevalent among agricultural workers in Sri Lanka and there is no clear etiology. Jayasumana et al. (2015) compared urine samples of patients diagnosed with this disease to two control groups that were nonpatients from endemic and nonendemic areas. Concentrations of glyphosate in urine of the clinically ill group ranged from 28.2 to >80 µg/L and adjusting for creatinine resulted in greater concentrations of glyphosate in urine samples of these patients than either of the control groups. Renal disease is known to reduce urine volume. Accordingly, the urinary concentration of a biomarker depends on the excretion rate and the urinary flow rate. Oliguria has been shown to increase concentrations of biomarkers and absolute concentrations may be misleading because of the difference in flow rates. Likewise, creatinine normalization in the setting of renal disease may be misleading because of variability in the urinary creatinine excretion rate (Waikar et al., 2010). Therefore, elevated concentrations of glyphosate could be an artifact, unrelated to dietary exposure since 24-h exposure requires a reasonable estimate or actual measurement of urine volume. Nevertheless, dietary exposures to glyphosate would be 2.67% of the ADI, when using the same calculations described in Table 6.

The German Federal Environmental Agency conducted an unpublished study of glyphosate levels in human urine that was reported by Niemann et al. (2015). A 24-h urine sample was collected in 1996 and 2012 from 10 male and 10 female students from 20 to 29 years. The subjects lived in Greifswald, Germany and samples were analyzed by gas chromatography with an LOQ of 0.15 µg/L. Glyphosate was detectable in 22 of the 40 samples. Data were not shown but the authors indicated that in 2012 compared to 1996 the values for glyphosate were slightly higher, but values for AMPA declined. The maximum concentration for glyphosate was of 0.65 µg/L, which corresponds to 0.02% of the ADI. A subsequent manuscript (Conrad et al., 2017) apparently shares data with the Markard study. This retrospective study used 24-h urine samples from human subjects who were 20–29 years old, and samples were collected yearly from 2001 through 2015 from Greifswald. Samples were analyzed randomly from 20 males and 20 females in each year. Glyphosate was measured by a validated GC-MS/MS method with a reported LOQ of 0.1 µg/L and 31.8% of the samples had concentrations of glyphosate greater than the LOQ. This report suggests that concentrations of glyphosate in urine increased from 2001 to 2011 and declined after 2012. The maximum detected urinary glyphosate was 2.80 µg/L and was from 2013. This value represents 0.09% of the ADI.

Urine of 13 Danish mothers and 14 of their children were tested for glyphosate using the Abraxis ELISA kit (Knudsen et al., 2017), albeit this assay is not validated for urine samples. They reported that glyphosate was detectable in all samples and that values ranged from 0.49 to 3.22 µg/L and 0.85 to 3.31 µg/L for mothers and children, respectively. Some values at the lower end of these ranges are below the Abraxis LOD for glyphosate in water samples provided in the instructions. However, even if the highest reported values are assumed to be accurate, these would represent a total exposure to glyphosate that are 0.11% and 0.26% of the ADI, for the mothers and children, respectively.

Eighty-four subsistence farmers and fishermen in Mexico had urine tested for glyphosate using the ELISA kit. Based on their supplemental data, 69% of urine samples had a mean concentration of glyphosate of 0.36 µg/L, and the maximum value was from a farmer with 0.48 µg/L (Rendon-von Osten & Dzul-Caamal, 2017). From this highest concentration of glyphosate in the urine, an exposure of 0.02% of the ADI can be calculated.

One report used an HPLC-MS assay, but without reporting validation details, to test urine samples for glyphosate from a total of 100 men and women (Mills et al., 2017). These subjects had samples collected in 1993–1996 and again in 2014–2016. The average age of subjects at the time of the later sampling period was 78 years. The number of detects and the mean values were elevated between 1993 and 1996 and 2014 and 2016, which the authors attributed to increased use of glyphosate. The upper 95% confidence intervals for urinary glyphosate from the earlier and later periods were 0.255 and 0.547 µg/L, respectively. This apparent doubling of concentrations using the high-ends of the confidence intervals could seem like a significant change in exposure; however, exposures as a percent of the ADI would have changed from 0.01% to 0.02% of the ADI, resulting in both values having a >6000-fold reasonable expectation of safety.

Urine samples from pregnant women, 18–39 years old, in Indiana were tested for glyphosate (Parvez et al., 2018). Two samples were collected during gestation, but results were reported only for the samples collected during gestation weeks 11–38. The assay used a validated LC-MS/MS method. They reported that 93% of these women had glyphosate in their urine above the LOD of 0.1 µg/L and the average concentration was 3.40 µg/L. There was no correlation between urinary glyphosate and growth measurements of the baby. They reported that gestation length was shorter for mothers with higher concentrations of glyphosate in urine, but the change was approximately 2 days. Given the small sample size, wide age range, complications arising from two premature deliveries, errors in establishing time of conception, and mean values within normal ranges, the significance of this association is equivocal. Based on the urine values, the woman with the highest exposure would have been at 0.24% of the ADI.

Urine samples were collected from 36 college students from one class in Erie, PA and analyzed for glyphosate using the ELISA kit (John & Liu, 2018), which was not validated. The LOD of the assay was given as 0.075 µg/L, although most reports employing this assay used a dilution that resulted in an unverified LOD of 7.5 µg/L. They reported that one subject had a urinary glyphosate value that was less than the LOD, and that concentrations of glyphosate in urine ranged from 0.096 to 1.033 µg/L. Using the highest value, this subject would have had an exposure of 0.03% of the ADI.

A small pilot study of glyphosate in the urine of 50 Irish adults was conducted in June 2017 (Connolly et al., 2018). Most of these subjects reported using glyphosate-based products at home but none used them as part of their occupation. A validated LC-MS/MS method was used with a reported LOD = 0.5 µg/L. Samples were normalized for creatinine. Glyphosate was not detectable in 80% of the samples. Median and maximum concentrations in the 10 samples that had detectable amounts of glyphosate were 0.87 and 1.35 µg/L, respectively. Based on this highest value, this person would have been exposed to 0.05% of the EU ADI.

A study was conducted in Mexico with 281 children who ranged in age from 5 to 15 years (Sierra-Diaz et al., 2019). They used an HPLC-MS/MS assay, but the manuscript is missing information about validation, LOD, and LOQ. They claim that glyphosate was detected in 73% and 100% of test subjects from two different communities and the mean concentrations of glyphosate in urine were 0.36 and 0.61 µg/L, respectively. The authors indicated that in one of these communities, children are known to apply pesticides, but the authors did not indicate if any of these subjects had sprayed glyphosate prior to sampling urine. Assuming that glyphosate in urine was due to ingestion and that the children weighed 25 kg, this highest urinary concentration of glyphosate would result in an exposure that was 0.17% of the ADI.

Trasande et al. (2020) conducted a study in which urine samples from three groups of children were collected and tested for glyphosate using an LC-MS/MS assay that lacks necessary validation information. Glyphosate was detectable in 30%, 12.5%, and 7.6% of the infant/children subjects in the following age categories: <30 days; 10–19 months old; and 3–8 years old, respectively. The mean detectable urinary glyphosate concentration was 0.278 µg/L, and concentrations ranged from 0.105 to 2.125 µg/L. Their highest exposure would be 0.17% of the glyphosate ADI. Their analysis did not find an association between log-transformed glyphosate concentrations and any of three biomarkers of kidney injury (albuminuria, neutrophil gelatinase-associated lipocalin, and Kidney Injury Marker-1).

In one study published recently, 24-h urine samples were collected from 301 adults for analysis of glyphosate (Soukup et al., 2020). Samples were analyzed using an LC-MS/MS method that was modified from the procedure reported by Jensen et al. (2016) with their own validation. Study subjects were recruited to be healthy, not pregnant or lactating, and not taking medications. For these subjects, 66.5% did not have detectable glyphosate or AMPA. Glyphosate exposures as a percent of the ADI were calculated for each study subject using their own body weight and 24-h urine excretion. Therefore, unlike other studies, there was no need to use assumptions for body weight or volume of urine. Otherwise, they had the same assumptions, as used throughout the calculations in this paper: (1) that 20% of ingested glyphosate is absorbed; (2) that glyphosate is poorly metabolized; and (3) that glyphosate is rapidly eliminated via urine, showing no potential for bioaccumulation. The maximum glyphosate exposure was 1.36 µg/kg BW and, using the actual body weights, they calculated that this person's exposure represented 0.13% of the ADI. This study also used 24-h dietary recalls and did rank-order correlations to estimate food sources of glyphosate and AMPA. This was done based solely on the amount of food consumed and not measured glyphosate content of the food. Nevertheless, they found that consumption of pulses and mushrooms was correlated with glyphosate and AMPA in urine, respectively. Absorbed glyphosate is not metabolized in the body, suggesting that ingestion of AMPA per se, not glyphosate, was responsible for urinary AMPA. The authors conjectured that mushrooms can get exposure to glyphosate from cereal straw and manure, which by microbial degradation is converted to AMPA, but amount of AMPA in the mushrooms was not known.

A study tested glyphosate in urine samples, using a validated MS method, from 16 people prior-to and after switching between conventional and strictly organic diets (Fagan et al., 2020). The greatest concentrations for the children and adults prior to switching diets were 6.22 and 0.82 µg/L, which correspond to 0.50% and 0.03% of the EU ADI. The study reported that the switch to organic diets resulted in a rapid decrease in urinary glyphosate, which is not unexpected as it confirms the short half-life of absorbed glyphosate, which was discussed previously.

The studies reported above use different assays, different diets, and different demographics of test subjects. In spite of these various conditions, they all result in exposures that are low, and in a narrow range, for glyphosate as a percentage of its ADI (0.2–2.67%).

6.2.2 Media reports

Although media reports of data are not typically included in scientific reviews, several media reports about urinary glyphosate have generated noteworthy public attention necessitating a scholarly discussion of their scientific elements. Many of these media reports lack details that handicap accurate interpretation of the results, such as some or all of the following: (1) no information about sample collection, storage, or processing; (2) no information about subjects; (3) no information about assay or its validation; (4) no information about an LOD/LOQ or how it was used in the presentation of the data; and (5) results often indicated that all, or nearly all samples, were positive for residues, which is inconsistent with reports using validated assays and use of an LOD. Some reports did indicate that they used the ELISA commercial kit, but no reports include data showing validation of this method for urine. In spite of these deficiencies, results in these reports can be scrutinized to determine if they provide information that would be pertinent to understanding the safety of glyphosate exposure within a specific region or population.

An article was published in a communication medium for a nonprofit foundation about testing glyphosate in the urine of nonagricultural workers in Berlin (Brändli & Reinacher, 2012). This article indicated that all samples had detectable concentrations of glyphosate, and that they ranged from 0.5 to 2 µg/L. However, information about their analytical methods used for these data was intentionally withheld from inclusion in this publication with a claim that the data about their method would be published within the year, which did not appear to have occurred. Their highest urinary glyphosate represents 0.07% of the ADI.

A report that was the subject of numerous media reports was by the Heinrich Böll Foundation (Krüger et al., 2016b) that analyzed 2011 urine samples. This is the largest number of samples analyzed for this purpose but the Abraxis ELISA kit was utilized, a method lacking validation in urine. Nonetheless, the authors claimed that 99.6% of Germans were “contaminated” and 75% had “urine levels more than five times the legal limit of drinking water.” However, the legal limit for glyphosate in water is not a point of reference directly relevant to a health standard. Their highest value was 0.14% of the ADI, and due to the extensive media coverage of this report, the German Federal Institute for Risk Assessment put out a statement stating that for children, the residues in urine are within the expected range and without any expected adverse health effects (BfR, 2016).

Urine of EU Ministers was sampled and tested for glyphosate, using the Abraxis ELISA for which the report did not provide any information about assay validation (Krüger et al., 2016a). Creatinine was used to correct for differences in water consumption; however, creatinine is related to muscle mass, which necessitates body weight measurements to be useful. The authors reported that all of their 48 subjects had detectable concentrations of urinary glyphosate, with concentrations ranging from 0.17 to 3.57 µg/L. The report did not include LOD or LOQ values for the assay, complicating the interpretability of these reported values. If their maximum value is used, this would result in the exposure for this individual with the highest concentration of glyphosate equal to 0.12% of the ADI, which is well within acceptable exposure levels.

A report was published in French media about 64 farmers providing urine samples for testing of glyphosate (Sauvage, 2020). Apparently, the ELISA was run by two labs and results were compared to the results of an LC-MS/MS assay, run by a third lab, but it is not clear if this was done for all samples. There is no information about LODs for the assays, or if they were even used it in summarizing the results. They did conclude that the ELISA always detected higher concentrations than the LC method. The highest detectable urinary glyphosate was 2.6 µg/L from a sample tested by ELISA. It is noteworthy that this value was described as alarming, but it represents 0.09% of the ADI.

6.3 Summary about concentrations of glyphosate in urine

The absorption and excretion of glyphosate allows for a reasonably accurate way to calculate the transient ingestion of glyphosate without having to make assumptions about what a person consumed, the concentrations of glyphosate in each consumed type of food, the daily consumption of individual foods, and the concentration of glyphosate in water and the amount of water consumed. Therefore, urinary glyphosate measurements allow for an estimate of glyphosate ingestion that can be compared to the ADI. These studies, and the comparison to the ADI, are presented in Table 6 and Figure 2. Although the Abraxis ELISA for glyphosate is not validated for urine samples, these data suggest that urinary glyphosate concentrations appear to be slightly greater when assayed using the ELISA as compared to the more precise and often well-validated LC-MS/MS assays. This could simply be due to the LOD of the ELISA being greater than for the LC-MS/MS, but it is not always clear how the LOD was applied when summarizing results generated by the ELISA. Regardless of the assay used, exposures of individuals are less than 2.67% of the ADI with the greatest dietary exposure to glyphosate determined using the questionable ELISA assay. The studies by Jayasumana et al. (2015), Rendon-von Osten and Dzul-Caamal (2017), and Curwin et al. (2007b) do not clearly distinguish between occupational/farm exposure and dietary exposure. All remaining studies have glyphosate exposures that are <1% of the ADI. Calculating glyphosate exposure from concentrations in urine indicates that the concentrations in reports summarized in this review paper result in exposures that are substantially below the ADI (Figure 2). As mentioned previously, an ADI is an estimate of the amount of a substance in food and drinking water, expressed on a body weight basis, that can be ingested over a lifetime without appreciable health risk (Chemicals Regulation Directorate, 2013). Exposure of humans to glyphosate was reviewed recently by Gillezeau et al. (2019) and they concluded that mean concentrations of glyphosate in urine are typically <4 µg/L, but they did not compare this concentration to a safety standard. It is noteworthy that they also had a concern about the sensitivity of the ELISA and the difficulty of combining data or comparing across studies due to differences in the LOD/LOQ. Using their value (4 µg/L) would result in an estimated ingestion of 0.13% of the ADI, an amount that is over two orders of magnitude below the ADI, and therefore presumed to be safe.

Details are in the caption following the image
Estimate of glyphosate exposure for adults as a percentage of the EU acceptable daily intake (%ADI) derived from the maximum concentrations obtained from each cited study. Urine was tested using an ELISA method, a mass spec (MS) method, or a method that was not reported (NR). None of the studies that used the ELISA provide detailed validation, whereas more detailed validation information are available for the MS assays.


Calculating dietary consumption of glyphosate can be done using two disparate methods. In the first method, residues are measured in individual food items and these are summed based on consumption data for the foods that people eat. This approach requires answering several questions to make reasonable assumptions used in the modeling such as: (1) what residue value for a food is used; (2) is the food processed or cooked; (3) what residue value is used when an analyte's concentration is < LOD; (4) are mean, median, or 90th percentile values used for consumption; and (5) what and how much do people eat? In the second approach, for a pesticide like glyphosate with the knowledge of absorption from the gut, lack of metabolism, and elimination from the body, sampling of urine is an accurate way to calculate ingestion and exposure within the body (Acquavella et al., 2004; Niemann et al., 2015). Regardless of which of the two methods is used, these values need to be compared to a safety standard, such as the ADI or RfD, which are regulatory-derived safety standards. Results of determining exposure to glyphosate by dietary modeling or urinary glyphosate are presented in Table 5, Table 6, and Figure 2. Modeling allows different scenarios using estimates of consumption and data derived from market surveys, whereas urine is a surrogate for estimating actual dietary exposure to glyphosate. The results from these two methods are in relatively close agreement. Dietary estimates range from 0.03% to 18% depending on assumptions. An especially critical assumption used is the residue levels that are used. Urinary glyphosate estimates of exposure are 0.02%–2.67% of the ADI, which do not require an assumption about residue on individual foods.

In spite of that, testing and publishing about glyphosate residues, whether in peer-reviewed journals, by internet postings or in the news media, has become somewhat common in the last decade. Unfortunately, many of the popular press reports are accompanied by value-judgment words like “high” or “contaminate,” or they make scientifically inappropriate comparisons to other standards (i.e., concentrations in urine vs. regulatory defined residue levels for drinking water). Furthermore, some of these reports imply that glyphosate residues were not known to exist previously in a given food or in urine, and, therefore the findings are regarded as novel. Recent publications, such as Winter and Jara (2015), Winter et al. (2019), and Reeves et al. (2019), have attempted to provide more information about the process for risk assessment of pesticides conducted by regulatory agencies. Moreover, timely communications from regulatory agencies, such as BfR responding to reports of residues in food, German beer, or urine (BfR, 2016, 2017), provide helpful information from what should be a trustworthy source in the face of widespread social media communications about food and agriculture (Ryan et al., 2020).

One statistic that is often encountered in publications that also might generate concern by consumers is reports of increased trends over time of usage of a pesticide, either by expanded adoption or as the result of new technology, such as herbicide-tolerant crops (Benbrook, 2016). These statistics intimate that pesticide use has exceeded safe levels established by the original regulatory assessments. In one residue study, the authors suggested that it appeared that MRL values were adjusted due to actual observed increases and not based on toxicity (Bøhn et al., 2014). This is precisely how MRLs are derived. It is important to highlight that alterations in the use of a previously approved pesticide, such as usage of glyphosate on newly approved GT crops, require new residue data to be submitted from the pesticide registrant(s) prior to regulatory approval. These new residue data are reviewed by regulators in order to ensure that the previous ADI or RfD is not exceeded. Additionally, MRLs or tolerances are derived from empirical data of real-world conditions and, once established, MRLs represent for any crop the agricultural practice that results in the highest residue. EPA (1996) stated in their guidance that pesticide use patterns, such as changes in the preharvest interval and/or postharvest treatment, are likely to require residue studies, and potentially another petition for a new tolerance. Expanded usage of a pesticide might change, but by conservatively assuming that 100% of a crop will use the agricultural practice with the highest residue, exposure remaining below the ADI is not subject to changes in commercial adoption. If a new exposure resulted in the sum of all exposures exceeding the ADI, there would need to be a restriction in some use. Moreover, since regulatory authorities use data collected prior to authorization of cultivation or import of the crop, combined with periodic testing to ensure that tolerances are not being exceeded, these media reports of residues do not necessarily provide unexpected data. When properly conducted, independent, peer-reviewed studies are published, they can be a corroboration of the accuracy of previously reported regulatory residue studies.

Putting pesticide residues into context by converting these values to percentages of the EFSA- or EPA-derived ADI or RfD helps one understand the margin of safety, but many consumers want food that is free of synthetic pesticides (Krystallis & Chryssohoidis, 2005). According to Currie (1999), many believe that with improved assays, a concentration of zero might be detected, but that is scientifically not feasible.

More than ever, as in other areas of science, transparency on residues of pesticides and their assessment by global regulatory authorities entrusted by the public to ensure food safety is needed to address complex scientific information (OECD, 2020). The scientific publication process that requires peer-review of the data and conclusions has largely provided the basis for science-based regulatory assessments for the past century (Codex, 2004) . Although peer-review is not a foolproof process, it is a process with the intent of ensuring that results and conclusions from published studies are based on well-conducted and documented scientific experiments. This is in sharp contrast to the essentially unreviewed environment of media and online publications. Adequacy of peer review is increasingly more confusing with predatory journals and electronic publishing (Kelly et al., 2014). Since the public lacks training to help them distinguish information from peer-reviewed journals and science-based regulatory authorities from information they see in media reports and predatory journals, this review has included results on glyphosate residues from both sources to provide them with a single-point reference for an informed discussion of this subject.


It has been observed that the perception of risk is greater with consumers when: (1) it cannot be detected; (2) it is not well understood; and (3) the belief is that the science is not known (Ropiek, 2014). Global regulatory authorities have spent the past decades establishing review practices designed to define, as much as scientifically possible, the data required to establish the parameters needed to define the known science needed for a safe food supply, and what needs to be detected and understood to ensure that such safety requirements are met. However, it is clear that more is needed to demystify these regulatory processes established to ensure their safety. In this review, glyphosate residue data from both regulatory authorities and reports from many groups, both peer-reviewed and in the media, have been summarized. To generalize this large amount of information, glyphosate residue data show that dietary residue exposure is well below established ADIs.


The authors wish to thank Kevin Glenn, Christophe Gustin, and Kristian Kather for their guidance and constructive review. The authors are all employees of Bayer Crop Science, a major manufacturer of glyphosate.


    J. Vicini coordinated the overall process and drafted the manuscript, except for the assay portion that was drafted by P. Jensen. J. Swarthout was responsible for calculations and interpretation of urine values. B. Young was responsible for collecting and interpreting data from residue reports from regulatory agencies.

    • 1 Current residue definition for glyphosate in soybeans is glyphosate only in the EU, and the sum of glyphosate and N-acetlyglyphosate, expressed as glyphosate, in the United States.