2.4.1 Evidence from studies in laboratory animals
The EOS article claims that glyphosate is carcinogenic, based on an increase in testicular tumours in rats treated via their diet for two years at 3, 10 and 32 mg/kg bw/d. However, pesticide regulatory agencies have not classified glyphosate as a carcinogen because the effect did not occur at higher doses in another two-year rat study. EOS argues that endocrine effects are more potent at low doses than higher doses, and so regulators should re-classify glyphosate as a carcinogen. EOS also claims that George et al (2010) have demonstrated that glyphosate induces cancer in mouse skin.
126.96.36.199 Carcinogenicity via the oral route
The study in which testicular tumours occurred (Lankas, 1981) has been reviewed by the Australian DoHA (1985), WHO (1994) and US EPA (1993). The German BVL did not evaluate this study for the JMPR (2004b), but the EU review includes a summary of the WHO assessment. Rats were treated with glyphosate for 26 months at dietary doses of ca 3, 10 and 31 mg/kg bw/d in males and 3, 11 or 34 mg/kg bw/d in females. The incidence of testicular interstitial (Leydig) cell tumours at termination was 0/15 among controls and 2/26, 1/16 and 4/26 at the three respective doses. The total incidence for all males was 0/50, 3/50, 1/50 and 6/50. The BVL did not attribute the finding to treatment, noting that Leydig cell tumours are common in ageing rats, that the incidence at 31 mg/kg “only slightly exceeded the historical control range,” and that no such effect had been observed in several more recent rat studies at much higher doses. In the absence of treatment-related effects, the NOEL was set at 31 mg/kg bw/d.
The WHO (1994), US EPA (1993) and DoHA (1985) all agreed that the tumours were not treatment-related because their incidence lay within the HC range. This interpretation was supported by data shown in the Australian assessment, showing that the incidences of Leydig cell tumours in glyphosate-treated rats were not different to those in male controls from concurrent studies at the same laboratory (4/65, 3/11, 3/26, 3/24 and 3/40).
Furthermore, glyphosate has not caused cancer in the testis – or at other sites – in any of the other dietary carcinogenicity studies assessed the Australian DoHA (1985, 1991 and 1992), US EPA (1993), EU (1998) and JMPR (2004b). The database comprises:
A 20-month study in mice at ca 11.3 – 45 mg/kg bw/d (Indian Institute of Toxicology, undated);
A 22-month study with glyphosate trimesium in male and female mice treated at 11.7 – 991 and 16.0 – 1341 mg/kg bw/d respectively (Stauffer Chemical Co, 1987a);
Two-year studies in mice at 100 – 1000 mg/kg bw/d (Atkinson et al, 1993a) and 157 – 4841 and 190 – 5874 mg/kg bw/d in males and females, respectively (Knezevich and Hogan, 1983); and
Two-year studies in rats at 89 – 940 and 113 – 1183 mg/kg bw/d in males and females respectively (Stout and Ruecker, 1990); 10 – 1000 mg/kg bw/d (Atkinson et al, 1993b); 121 – 1214 and 145 – 1498 mg/kg bw/d in males and females (Brammer, 2001); 6.3 – 595 and 8.6 – 886 mg/kg bw/d in males and females (Suresh, 1996); and at 4.2 – 41.8 and 5.4 – 55.7 mg/kg bw/d in males and females (glyphosate trimesium salt; Stauffer Chemical Co, 1984).
Despite EOS’s argument that endocrine-mediated effects are specifically low dose phenomena, dietary doses of between 4 and 16 mg/kg bw/d (which lie within the range given by Lankas, 1981) have failed to cause any testicular effects in two mouse and three rat carcinogenicity studies. Therefore, the weight of evidence does not support the EOS assertion that glyphosate is a testicular carcinogen.
188.8.131.52 Dermal carcinogenicity
George et al (2010) tested Roundup Original (a product containing 360 g/L glyphosate and 15% POEA) in a mouse two-stage initiation / promotion model of skin cancer. Following a single dermal dose of the tumour initiator DMBA (7,12-dimethyl benz[a]anthracene) mice were treated dermally, three times per week for 32 weeks, with Roundup (25 mg/kg bw) or a positive control chemical (the tumour promoter TPA (12-O-tetradecanoyl-phorbol 13-acetate) at 5 µg/mouse). Skin cancers (squamous cell papillomas) were present on eight/20 Roundup-treated mice and 20/20 positive controls at termination. By contrast, tumours did not develop on untreated (negative control) animals or further mice that received a single dose of DMBA without a promoter; or 32 weeks’ treatment with Roundup or TPA without prior initiation; or one dose of Roundup followed by TPA for 32 weeks.
Before discussing the significance of George et al’s findings, we must briefly consider the biological basis for the two-stage initiation / promotion model they utilised. This experimental model has been developed in light of the multistage model of carcinogenesis8, the current scientific explanation of how cancers are formed from normal cells. In their experimental design, George et al used a single dose of DMBA to initiate skin tumours and repeated doses of TPA to promote them. Tumours did not develop on animals that received the initiator without subsequent promotion, or on mice treated with the promoter without prior initiation. When substituted for DMBA, Roundup did not behave as a tumour initiator, as tumours did not form on mice treated subsequently with TPA. Furthermore, Roundup was not a complete carcinogen, since tumours did not develop on animals that received it without prior initiation. However, Roundup did behave as a tumour promoter on mice that had already received DMBA.
Because George et al did not apply pure glyphosate or POEA to the test animals, their study could not identify which component(s) of Roundup Original was responsible for the promoting activity. Therefore, EOS’s assertion that “glyphosate induces cancer in mouse skin” is not strictly correct. Furthermore, while single doses of Roundup and TPA induced similar changes in dermal protein expression, it remains unclear whether the formulation and positive control shared a common mode of action (see assessment in Appendix 3).
However, the most important issue raised by this study is whether Roundup Original or other GBHFs are likely to pose a dermal carcinogenicity hazard to persons preparing them for application. In this regard, several factors require consideration:
The weight of evidence suggests that neither glyphosate nor POEA are genotoxins, either alone or in combination. Furthermore, glyphosate has been shown not to be carcinogenic via the oral route in ten studies in two laboratory species.
Roundup Original was not a complete carcinogen in the mouse initiation / promotion model. Tumour initiation was a prerequisite for the eventual development of dermal cancers. Therefore, this and similar products would not be expected to promote tumour formation on human skin in the absence of prior initiation.
Roundup Original was a markedly less potent promoter than the positive control, TPA. George et al applied the formulation at a 150-fold higher dose than TPA (25 mg/kg bw compared with 5 µg/mouse, equivalent to ca 0.17 mg/kg assuming a 30 g bodyweight). Despite this, Roundup promoted tumour formation more slowly than did the positive control. Tumours first appeared after 130 days on Roundup-treated mice, compared with 52 days on those receiving TPA. Fewer, smaller tumours developed on Roundup-treated mice than on those receiving TPA. Moreover, tumour formation occurred on all positive control mice, compared with 40% of those receiving Roundup.
Tumour promotion is reversible, requires prolonged and repeated exposure to the promoter, and the promoted cell population depends on the continued presence of the promoter (Derelanko, 2002). On mice, tumours did not appear until 130 days of treatment with Roundup Original. Assuming a lifespan of 80 years, humans would have to be exposed to Roundup for three days per week for ca 14 years to achieve the equivalent of 130 days of the ca 730-day mouse lifespan. Few herbicide mixer / loaders, if any, would experience such prolonged uninterrupted exposure, especially in situations where GBHFs have a seasonal pattern of use.
Mice received Roundup Original at 25 mg/kg bw/d, which is equivalent to 1500 mg/d for a 60 kg human. The mass of Roundup formulation that must be handled per day to attain a dermal dose of 1500 mg can be estimated using the US EPA (2012) Exposure Surrogate Reference Table. Based on monitoring studies of operators mixing and loading liquid pesticide concentrates under field conditions, this nominates a mean unit dermal exposure of 0.083 mg/kg handled for persons wearing a single clothing layer and gloves9. Therefore, to attain a dermal exposure of 1500 mg, 1500 ÷ 0.083 = 18 072 kg of the product would have to be handled, which is at least ten times higher than could be achieved in a working day.
2.4.2 Evidence from human populations
Citing human epidemiology studies by De Roos et al (2005), Hardell and Eriksson (1999), Hardell et al (2002) and Eriksson et al (2008), EOS claims that there is an association between exposure to glyphosate / GBHFs and the blood system cancers multiple myeloma (MM) and non-Hodgkin’s lymphoma (NHL).
In 2005, the Australian DoHA evaluated epidemiological evidence of associations between use of glyphosate and cancer.
According to the DoHA, McDuffie et al (2001) found no significant association between previous use of Roundup and the occurrence of Non-Hodgkin’s Lymphoma (NHL) among Canadian men (119 test and 301 control), although the study did suggest an association between increased risk of NHL and the use of multiple pesticides.
The Agricultural Health Survey, a prospective cohort study of 57 311 licensed pesticide applicators in Iowa and North Carolina (De Roos et al, 2005a) found no association between glyphosate exposure and NHL. Based on 22 of 32 cases10, mixing or using glyphosate products was claimed to be associated with an elevated risk of multiple myeloma (MM), with an odds ratio of 2.6 (95% Confidence Interval = 0.7–9.4), although the lower CI of 0.7 limited the strength of the finding. There was also a possible relationship between the risk of MM and the cumulative exposure days (years of glyphosate use X days per year) but not intensity-weighted exposure (years of glyphosate use X days X intensity level). However, when Sorahan (2012) re-analysed the complete dataset of 32 cases, the relative risk for ever using glyphosate was only 1.1 (95% CI = 0.5–2.4) when adjusted for age. Additional adjustment for education, smoking, alcohol use, family history of cancer and use of 10 other pesticides had little effect (OR = 1.2; 95% CI = 0.5–2.9). This demonstrates that glyphosate use is not associated with increased risk of MM.
De Roos et al (2005b) found a possible association between NHL and the use of glyphosate in a pooled analysis of 650 males participating in case-control studies performed by the US National Cancer Institute during the 1980s. An OR of 2.1 (95% CI = 1.1–4.0) was detected by logistic regression, but the association was weaker (OR = 1.6; 95% CI = 0.9–2.8) when analysed by hierarchical regression.
In a study of 515 cases and 1141 controls, Hardell et al (2002) obtained elevated risk of NHL or hairy cell leukaemia (HCL) among men who had used glyphosate. However, the DoHA considered the finding as equivocal because of the small sample size (8 cases and 8 controls), inconsistency between the odds ratios obtained by univariate analysis (3.04; 95% Confidence Interval = 1.08–8.52) and multivariate analysis (1.85; 95% CI = 0.55–6.20), and the wide breadth of the 95% confidence intervals.
In a follow-up study (see assessment in Appendix 3), Eriksson et al (2008) examined exposure to pesticides as a risk factor for NHL in 910 cases and 1016 controls. Univariate analysis revealed a significant association between NHL and exposure to glyphosate (29 cases and 18 controls; OR = 2.02; 95% CI = 1.10–3.71), exposure to glyphosate with a latency of >10 years between exposure and diagnosis
(OR = 2.26; 95% CI = 1.16 – 4.40) and exposure to glyphosate for >10 days (17 cases and 9 controls; OR = 2.36; 95% CI = 1.04–5.37). However, NHL was not associated with exposure to glyphosate with a latency of 1–10 years between exposure and diagnosis (OR = 1.11; 95% CI = 0.24 – 5.08) and was, at most, only weakly associated with exposure to glyphosate for <10 days (12 cases and 9 controls; OR = 1.69; 95% CI = 0.70 – 4.07). Multivariate analysis did not demonstrate any association between NHL and glyphosate exposure
(OR = 1.51; 95% CI = 0.77–2.94).
Of the epidemiology studies assessed in Australia, three have suggested an association between glyphosate use or exposure and NHL, but obtained inconsistent results depending on the type of statistical analysis performed. Two other studies have searched for but did not find any such association. Possible associations between glyphosate and HCL and MM were observed in one study each, although the association with MM has subsequently been discounted following a re-analysis of the data.
When weighing up the significance of these results, it is worth taking account of the limitations in the design of the studies, which (with the exception of De Roos, 2005a) collected exposure data in questionnaires relying on the accuracy of the respondent’s memory. This would result in recall bias, misclassification of pesticide exposure, and increased uncertainty regarding the actual level of exposure. Epidemiological studies of this type are also potentially confounded by exposure to multiple pesticides and by established risk factors for haematopoietic system cancers, such as immunosuppression and Epstein-Barr virus (DoHA, 2005).
The JMPR (2004b) review of glyphosate reached similar conclusions from its assessment of epidemiology studies by Hardell and Eriksson (1999), Nordstrom et al (1998) and McDuffie et al (2001), commenting that the claimed associations between glyphosate and lymphopoietic cancers were weak, were not controlled for confounding factors including other pesticides, and did not meet generally accepted criteria for determining causal relationships.