Pruitt's EPA Defending Monsanto's Glyphosate Herbicides

The Environmental Protection Agency (EPA) pesticide office continues to side with Monsanto’s interests to protect glyphosate-based herbicide products like Roundup.

NRDC let EPA know that it has strayed far from its mission, in detailed technical comments on EPA’s Draft Human Health & Ecological Risk Assessment (relevant documents on EPA’s website). Based on the combined data of adverse impacts, the EPA needs to severely restrict the use of glyphosate and glyphosate-based products immediately.

In addition, over 56,500 NRDC online members and activists also sent an email to EPA calling for the protection of monarch butterflies from deadly over-use of glyphosate-based herbicides (it’s not too late—you can take action too). Thank you!

Monarch butterfly


Since the last ecological risk assessment was completed for glyphosate in 1993, its use has skyrocketed resulting in glyphosate being the most widely used pesticide in the US.  In that time, mounting evidence has pointed to adverse human health effects as well as the precipitous decline of the North American monarch butterfly population. Research in the prestigious medical journal JAMA (Journal of the American Medical Association) reports on the startling evidence that glyphosate—the main ingredient in Monsanto’s weed-killer, Roundup—is not only getting into our bodies, but has been doing so at increasing levels for decades, as glyphosate crop uses rise across the country! (see my previous blog for details)

EPA Ignoring Cancer Risks

EPA is classifying glyphosate as “Not Likely” to cause cancer in humans, despite studies reporting elevated cancer risk in humans and laboratory animals (Dec, 2017 EPA Revised Glyphosate Issue Paper: Evaluation of Carcinogenic Potential). In our comments to EPA, we continue to push EPA to instead classify glyphosate and its formulated products as posing a risk of cancer. This would be consistent with the following scientific reviews:

  • The International Agency for Research on Cancer (IARC 2016) concluded in a consensus report that the cancer evidence from studies of people comprised “limited evidence” for risk of cancer (Non Hodgkin Lymphoma, NHL), which was supported by “sufficient” evidence of carcinogenicity in animal studies;
  • The EPA FIFRA Scientific Advisory Panel (SAP) did not reach a consensus, but a large proportion of Panelists concluded that there is suggestive evidence that glyphosate is a human carcinogen (SAP 2017, p. 45). EPA Cancer Guidelines define suggestive evidence as, “evidence of a positive response in studies whose power, design, or conduct limits the ability to draw a confident conclusion”, which many SAP Panelists felt was appropriate for glyphosate (SAP 2017, p. 48; EPA 2005, p. 83).
  • Even a Monsanto-sponsored meta-analysis by Chang and Delzell (2016) reported a positive and marginally statistically significant association between any versus no use of glyphosate and risk of NHL cancer (meta-RR = 1.3, 95% confidence interval (CI) = 1.0–1.6, based on six studies) and multiple myeloma (meta-RR = 1.4, 95% CI = 1.0–1.9; four studies)

One Panelist, Dr. Lianne Sheppard, emphasized that, “It is clear to me that we can't conclude, as the Agency has done, that it's not a carcinogen. That's just completely inappropriate based on their criteria" (SAP Transcript, p. 1192).

In October 2017 the Lancet Commission on Pollution and Health issued a consensus statement identifying pollution from chemicals and pesticides as the “largest environmental cause of disease and death in the world today, responsible for an estimated 9 million premature deaths” (Landrigan et al 2017). The paper devotes two paragraphs to glyphosate, highlighting cancer concerns: “Epidemiological studies of agricultural workers who were exposed occupationally to glyphosate and other herbicides have found evidence for increased occurrence of non-Hodgkin lymphoma in these people.” The authors of this report represent dozens of recognized medical and scientific experts from around the world, and the report is published in the Lancet, one of the most prestigious medical journals in the world.

Cancer Evidence in People from Field Studies

Even EPA seems to acknowledge that the studies are all positive for an elevated risk of NHL cancer in exposed people. EPA presents the following figure in its most recent cancer assessment (EPA, p. 65 Fig. 3.2).

EPA 2017 Glyphosate Cancer Issue Paper Figure 3.2

Below I address each of EPA’s excuses for discarding field study evidence of cancer risk:

Confused Use of Confounding: EPA suggested that the field studies failed to adjust for occupational exposure to diesel exhaust fumes, solvents, livestock and other farm animals, UV radiation, and ‘unknown factors’ which EPA proposes as “highly likely confounders in NHL studies” (EPA p.65). This is a mistaken understanding of a confounder which must both cause NHL and occur more frequently in heavy glyphosate users. For example, a study that examines the link between heavy alcohol use and premature death would be confounded if it didn’t also account for smoking, which both causes premature death and is more likely to occur among heavy drinkers (heavy drinkers are 3X more likely to also be heavy smokers than the general population). EPA has presented no information that would suggest that study subjects with high glyphosate exposures are also in the sunlight or around livestock any more than the study subjects with little or no glyphosate exposures.

Biased Approach to Addressing BiasEPA also rationalizes its dismissal of the epidemiologic evidence of cancer by noting that recall bias and missing data are frequent limitations in studies (EPA 2017 Cancer Issue Paper, p.66). However, again EPA has presented no evidence to support or suggest that there is recall bias or missing data at sufficient amounts to alter the study results. EPA should not dismiss evidence of harm based on unsubstantiated suppositions. “Declaring a chemical as not hazardous, or reducing a level of health protection, should require validation, not speculation” (Melnick et al 2003).

Overlooking Cancer Red Flags In The Ag Health StudyBoth Monsanto and EPA frequently support their no-cancer-risk position by citing a long-term prospective epidemiologic study of pesticide applicators in the Midwest, called the Agriculture Health Study (AHS), conducted by the NIH National Cancer Institute, that does not report a link between glyphosate exposure and NHL. However, a recent update of the cohort did report a possible association between glyphosate and another type of blood cancer called acute myeloid leukemia (AML) (Andreotti et al 2017). The study authors warn that, “Given the prevalence of use of this herbicide worldwide, expeditious efforts to replicate these findings are warranted” (Andreotti et al 2017). The increase risk of AML in the study was over 2-fold higher in highest exposed applicators compared with the never exposed applicators. The possible link with leukemia should be very concerning to the public and particularly to pesticide applicators, because AML is a very serious fast-growing cancer, with only about one-quarter of the people that have it surviving longer than 5 years. Although EPA acknowledges these new data, it considers them too limited to inform the assessment and simply states it will continue to follow the literature (Section 3.5.2 (1), p. 53). However, at the very least, these data should prevent EPA from classifying glyphosate as posing no cancer risk.

EPA dismisses much of the systematic reviews and meta-analyses by proposing—again without any evidence—that, “publication bias may play a role in this evaluation given there is a tendency to only publish positive results [data showing a cancer risk] and potential concerns regarding glyphosate have only been raised in recent years” (EPA p. 64). EPA provides no reference or support for this statement, which is not only pure conjecture but there is evidence that the opposite is true. That is, there is a bias for published industry-sponsored studies to report no harm associated with toxic chemicals—see for example: Bero et al 2016; Sass 2006; Vom Saal and Hughes 2005; Hayes 2004; Ong and Glantz 2001; Deerfield et al 1993.

Cancer Evidence in Lab Rodent Studies 

EPA concludes that “none of the tumors evaluated in individual rat and mouse carcinogenicity studies are treatment-related”. This conclusion is a rejection of the recommendation of SAP Panelists to: “Discuss the suggestive association of NHL risk and glyphosate exposure in humans, with the supporting evidence in mouse studies” (SAP 2017, p. 48).

Below I address each of EPA’s excuses for discarding experimental evidence of cancer risk:

Statistical test cherry-pickingEPA stated that the results were statistically significant with a trend test but not with other statistical tests. This argument seems to turn the Cancer Guidelines on its head –using it to eliminate cancer evidence, not to identify it. EPA is using the test that shows no cancer risk, in lieu of the tests that show cancer risks.

Undue emphasis on lack of clear monotonic responseEPA rejected cancer evidence if the dose-response was not also monotonic, meaning a steady increase in the number of tumors with an increase in the treatment dose.  The EPA Cancer Guidelines say in reference to the Bradford Hill criteria for epidemiologic information that “the absence of an exposure-response relationship does not exclude a causal relationship” (EPA 2005, p. 41). EPA should not have dismissed evidence of harm for this reason.

Undue emphasis on lack of pre-neoplastic observationsEPA dismissed tumors if pre-neoplastic changes were not also reported (for example, Table 4.20). EPA violated its Cancer Guidelines by turning them upside down regarding the relevance of pre-neoplastic (pre-cancer) tumors. The Guidelines wisely note that the presence of pre-neoplastic tumors may “lend support to the significance of findings for animal carcinogenicity” (EPA 2005 p. 48), whereas EPA uses the lack of reported pre-neoplastic tumors as an excuse to disregard observed tumors. In other words, EPA is interpreting the lack of supporting evidence as evidence of no harm, despite evidence of harm.

Absence of evidence of tumor progression is not absence of cancerEPA dismissed evidence of tumors in mice due to what it describes as a lack of evidence that the tumor would progress from pre-neoplastic to malignancy (EPA, p. 90). Three of the mouse studies lasted only 18 months: Kumar 2001 reported dose-dependent increases in renal tubular adenoma and malignant lymphoma; Wood et al 2009 reported dose-dependent increases in malignant lymphoma; Sugimoto 1997 reported a statistically significant trend test increase in malignant lymphoma, hemangiosarcoma, and renal tubular adenoma (see summary in Clausing et al 2018).  A mouse of 18 months is equivalent to a human of about 30 to 50 years old, hardly long enough to reliably predict a human cancer risk, since about 80% of all human cancers occur in people over the age of 60. For this reason, the NIEHS cancer experts include higher doses and consider various statistical methods when interpreting study results, with the goal of identifying agents that pose a cancer risk to humans. EPA should not disregard evidence of early cancer or pre-cancer lesions on the basis that it occurred in a test that was likely too short to give the lesions time to progress to full cancer. We would falsely believe that many carcinogens were safe if we only looked in people under the age of forty.

EPA discounted tumors in the highest treatment groups, but Dr. Christopher Portier, cancer assessment expert and Retired Director of the National Center for Environmental Health/Agency for Toxic Substances and Disease Registry, showed that after adjusting the study lengths and combining the data for a pooled analysis, the data demonstrated a highly significant trend for excess cancer risk for male mouse kidney cancer, male mouse malignant lymphoma, and hemangiosarcoma in male mice, even when high dose treatment groups are removed from the analysis (Portier 2016, p. 15).

Misuse of historical control dataEPA stated that the tumors were within the range of historical control data. This violates EPA Cancer Guidelines that state, “Generally speaking, statistically significant increases in tumors should not be discounted simply … because incidence rates in the concurrent controls are somewhat lower than average.” (EPA 2005 p. 48). SAP Panelists suggested to EPA that it may be using historical control data “subjectively” only in situations where it would reduce the impact of the cancer evidence, which may potentially introduce biases (SAP, p. 60). It seems that EPA choses the interpretation of data that makes the evidence of cancer risks go away.

In summary, a re-analysis of the rodent tumor data has recently been conducted by Dr. Christopher Portier, retired US government cancer expert. Dr. Portier reports an excessively high number of malignant lymphomas and hemangiosarcomas in the male mice in multiple studies of CD-1 strain of mice (See Portier Rebuttal Report, Oct 2017); these are blood-based tumors, as are the non-Hodgkin’s Lymphoma (NHL) cases seen in the human epidemiologic studies. These are the analyses that EPA – an Agency charged with protecting human health and the environment – should have come up with.

Monsanto Internal Experts Acknowledge Glyphosate Formulated Products Not Safe

Even Monsanto experts agree in internal memos made public through litigation and posted on the US Right to Know website (US RTK) that the glyphosate-based products carry additional health risks, compared with glyphosate alone. A 2002 internal Monsanto email states, “Glyphosate is OK but the formulated product (and thus the surfactant) does the damage” (Monsanto email, April 25, 2002). And, in another corporate email from Monsanto’s Manager of Toxicology Programs: “You cannot say that Roundup is not a carcinogen… we have not done the necessary testing on the formulation to make that statement”. (Monsanto email, Nov 22, 2003) These corporate disclosures represent a disturbing level of understanding of the potential health harms from its products, while making public assurances of the safety of its products.

Although we are pleased that EPA is working with the National Toxicology Program to evaluate the potential hazards of glyphosate-based products that can be purchased from store shelves, the research results are not available yet. Thus, the potential acute and chronic adverse impacts of glyphosate formulations is a major data gap in EPA’s assessment that should be addressed with the use of additional uncertainty factors.

Contaminated Food—Widespread Population Exposure

The SAP Panelists raised concern at the potential for the general population including young children to be exposed to glyphosate at occupational levels. In its report, the SAP pointed out to EPA that the presumption that  pesticide applicators had significantly higher exposures than the general population is incorrect; EPA’s high-end exposure estimate for 1-2 year old children (presuming all relevant foods have glyphosate residues at the maximum allowable limit) is 0.47 mg/kg/day, which is much higher than the exposure estimates for applicators and within the range of exposure estimates for pesticide mixers and loaders (0.03-7 mg/kg/day) (SAP 2017). 

However, there is still no public information on glyphosate residues in food. A 2014 report of the Government Accountability Office (GAO) was critical of FDA and USDA for both a failure to launch an effective pesticide residue testing program (GAO 2014). In response, the FDA finally started testing for residues of glyphosate in common foods only in the past two years, and no results have been officially made public yet.

Collapsing consumer confidence in the regulatory agencies led to consumer groups and academics doing their own testing, reporting glyphosate residues in breast milk, honey, cereal including oatmeal marketed for kids, infant formula, soy sauce, flour, and other food products (Reuters 2015; HuffPost Dec 6, 2017; US RTK Apr 18, 2018; Guardian, Apr 30 2018). Now reports in the Guardian reveal FDA internal documents showing that almost every food FDA has tested has been shown to be contaminated with glyphosate residues. “I have brought wheat crackers, granola cereal and corn meal from home and there’s a fair amount [of glyphosate] in all of them,” wrote FDA chemist Richard Thompson to colleagues in a 2017 mail  reported in the Guardian (Guardian, Apr 30 2018).

Failing our Families

The failure of the agencies to initiate and incorporate comprehensive glyphosate residue testing results into its assessment is a major data gap that should be addressed with the use of additional uncertainty factors. Under the Food Quality Protection Act,  the agency must prohibit any use for which the registrant has failed to demonstrate that there is a reasonable certainty of no harm to vulnerable populations including infants and children from cumulative and aggregate exposure (from the diet and all other sources.) EPA must apply an “additional tenfold margin of safety for the pesticide chemical residue and other sources of exposure…for infants and children to take into account potential pre- and post-natal toxicity and completeness of the data with respect to exposure and toxicity to infants and children.” EPA may use a different additional margin safety for the pesticide “only if, on the basis of reliable data, such margin will be safe for infants and children.” 

EPA must ensure that it is meeting its FQPA responsibilities in evaluating the exposure and risks to vulnerable populations as well as to the general population from both cancer and non-cancer risks, and has failed to do so.   

Tell the EPA to protect monarchs from toxic herbicides

About the Authors

Jennifer Sass

Senior Scientist, Federal Toxics, Health and Food, Healthy People & Thriving Communities Program

Join Us

When you sign up you'll become a member of NRDC's Activist Network. We will keep you informed with the latest alerts and progress reports.