Main text

Gillezeau et al.’s [1] review of glyphosate biomonitoring published in this journal earlier this year demonstrated that few glyphosate biomonitoring studies have been conducted since 2007. Because of glyphosate’s widespread use and frequent presence in the environment worldwide, it is important to understand routes and cumulative levels of exposure and potential human-health effects.

Between 1971 and 2014, the US applied nearly 1.6 billion kilograms of glyphosate, 66% of which had been applied in the previous decade [2]. Glyphosate-based herbicides (GBHs) are approved for over 100 different agricultural and non-agricultural uses including weed control along roads, canals, railroads, powerlines; in and around commercial and industrial facilities; and in a myriad of public spaces and around homes. GBHs have been detected in air, water, food, and companion and farm animal feedstuffs [3] and it can remain in food for over a year even after washing, cooking, and freezing practices [4].

In 1996, Roundup Ready, a glyphosate-tolerant seed type, was approved for farm use in the US and has since been integrated into the global genetically modified (GM) crop industry. Today, the US grows a variety of glyphosate-resistant plants including soybeans, maize, cotton, alfalfa, and sugarbeets. In 2012, approximately 94% of all soybeans (30 million hectares) planted in the US were Roundup Ready [5].

There are few glyphosate biomonitoring studies to date largely because the analytic methods for detecting glyphosate and its metabolite aminomethylphosphonic acid (AMPA) have not been well established. This is likely due to previous assumptions that glyphosate was non toxic based on prior genotoxicity testing, could only be applied to crops pre emmergence, and did not pose an exposure risk via groundwater runoff or food residue [6]. More accurate and sensitive methods using mass spectrometry (MS) have advanced more recently, which can improve accurate expsoure assessment and evaluation of health effects. For example, a recently published study involving 71 women in central Indiana found that over 90% of the women studied had detectable glyphosate levels in their urine; a finding that was significantly associated with a shorter pregnancy [7].

To determine if glyphosate exposure could be identified historically, we examined urine samples collected from dairy farmers between 1997 and 98 who self-reported glyphosate application.

Methods

Urine samples were collected following consent from farmers working on family farms in Wisconsin between 1997 and 1998, and stored at − 80 C. Recruitment and screening procedures have been discussed in detail elsewhere [8]. Briefly, urine samples were collected by farmers 8 h following first of the season pesticide application on cropland. Because the main study focused on restricted use pesticides only and because urinalysis measures were not readily available at the time of the original study, glyphosate was not analyzed in urine and no additional information beyond self-reported use was collected for glyphosate.

In 2018, we identified 18 samples from farmers who had reported applying glyphosate, and randomly selected 18 additional samples from the remaining archive of over 200 stored urines from the original study. Samples were analyzed by Neotron Laboratories to determine whether glyphosate or its metabolite, AMPA could be detected. Neotron used the US Food and Drug Administration’s liquid chromatography tandem mass spectrometry (LC-MS/MS) method [9] to detect and quantify glyphosate and AMPA residues. The limit of detection (LOD) for glyphosate was 0.4 μg/kg urine (0.4 ppb) and AMPA was 1 μg/kg urine (1 ppb).

Pearson Chi-square tests were used to compare detections among applicators and non applicators, and to determine if there was an association between glyphosate detection and farmer age, amount of land farmed, owned, or rented, gross annual farm income, hours of application of non GBH pesticides or number of acres to which non GBH pesticides were applied.

Results

One non-applicator sample could not be analyzed, resulting in 18 applicator and 17 non-applicator samples for urinalysis. Seven of the 18 applicator and none of the 17 non applicator samples had glyphosate detections above the limit of detection (Χ2 = 8.3; p-value < 0.01).

Table 1 compares concentration levels found in this study to the 8 occupational studies reviewed by Gillezeau et al. [1]. The mean level of glyphosate detected among Wisconsin dairy farmers was 4.04 μg/kg (4.04 ppb) across the seven positive samples (range 1.3–12.0 μg/kg). One individual, who had the highest glyphosate concentration (12.0 μg/kg), also had AMPA detected above the LOD (4.1 μg/kg). No other AMPA detections were observed.

Table 1 Comparison of occupational biomonitoring studies for glyphosate in urine, 1991–2018
Full size table

There was no association between glyphosate detection and farmer age, amount of land farmed, owned, or rented, gross annual farm income, hours of application of non GBH pesticides or number of acres to which non GBH pesticides were applied.

Discussion

Of the 35 farmer applicators whose urine was tested for glyphosate and AMPA, none of the non applicators had positive detections, whereas 39% of the applicators had positive detections. The sample with the highest concentration of 12.0 μg/kg also tested positive for AMPA. Both findings support the sensitivity of LC MS/MS to detect the actual analyte and its metabolite.

The occupational biomonitoring studies conducted to date and summarized in Table 1 demonstrate a) most have been conducted with small samples (range 1–76 participants; average 22); b) LODs vary widely, from 100 to 0.5 ppb; c) percent detections >LOD also vary widely (range 20–100%); c) there has a been a marked decrease in LODs over time; and d) none of the prior studies measured AMPA. Also of note is that creatinine adjustment has been used inconsistently, the importance of which for accurate glyphosate characterization in human urine is unclear [10].

Newly published reports are demonstrating previously unrecognized health impacts of glyphosate exposure, including a meta analysis of epidemiologic studies showing a link between glyphosate based herbicide expsoure and increased risk of Non-Hodgkin Lymphoma [11] and a long-term low dose toxicologic assessment that showed glyphosate and Roundup had developmental and endocrinological impacts in sprague dawley rats [12].

Precise measurements of glyphosate in human urine and serum is critical for informing human-equivalent doses in experimental models, particularly at low doses. Importantly, recent evidence from in vivo investigations suggests glyphosate bioaccumulates in rats, which requires further exposure assessment and biomarker investigation in humans [13]. Glyphosate is never used alone and is combined with adjuvants to increase plant penetration which can be more toxic than glyphosate alone [14]. To fully characterize health effects, adjuvants need to be identified and biomonitoring assessments need to be able to detect them in conjunction with glyphosate and AMPA.

While the effects of long-term specimen storage on glyphosate molecular integrity has not been well studied (see [15] for a 1998 report using nuclear magnetic resonance spectroscopy for detecting glyphosate proteins using flash freezing), the impacts of different freezing methods with and without cryoprotectants on glyphosate recovery need to be further assessed experimentally, and with other historical urine specimen repositories.

Detecting both glyphosate and AMPA in this small sample of Wisconsin farmers demonstrates LC-MS/MS can be used to detect concentrations in urine samples undergoing long-term cryopreservation (samples remained in -80c without cryoprotectant since first collection), and that glyphosate exposures among US farmers were occurring 20 years ago.