2.1 The association between glyphosate / glyphosate—based herbicides and developmental malformations
2.1.1 Effects in toad and bird embryos
According to EOS, Roundup causes developmental malformations in toad and chicken embryos at doses “much lower than those used in agricultural spraying” and “ten times lower than the MRL”. These claims are based on an article by Paganelli et al (2010; see Appendix 3), who treated African clawed toad (Xenopus laevis)embryos with glyphosate (360 or 500 pg by intracellular injection) or a 480 g/L Roundup formulation (present in the incubation medium at a 5000-fold dilution, or 96 mg glyphosate/L). The test compounds decreased the expression of genes that regulate embryonic development, impaired the formation of neurons (nerve fibres) and the neural crest, and also caused microphthalmia and microcephaly (abnormally small eyes and head).
Incubation with Roundup at 4000- and 3000-fold dilutions caused increases in retinoic acid (RA) signalling activity within toad embryos, whereas co-treatment with a RA-receptor antagonist blocked increases in RA signalling and prevented microcephaly. The study authors also found that injecting Roundup into chicken eggs (20 µL of 3500- or 4500-fold dilutions, equivalent to 2.7 or 2.1 µg glyphosate/egg) caused microphthalmia and microcephaly in the embryos. However, they did not investigate whether the malformations occurred in response to stimulation of RA signalling, as in Xenopus.
Retinoic acid, a metabolite of vitamin A, has a pivotal role in the development of the central nervous system and causes microcephaly, microphthalmia and neural tube defects including spina bifida when administered in excess to pregnant laboratory animals (Maden, 2002). Therefore, in principle, Paganelli’s study suggests that glyphosate and glyphosate-based herbicides may have the potential to cause developmental malformations by a mechanism involving RA.
However, caution should be exercised in extrapolating from findings in amphibians and birds to predicting risks for humans. The absorption, distribution, excretion and toxicokinetics of chemicals in pregnant mammals are fundamentally different to those in organisms whose development occurs in the external environment. Furthermore, the experimental routes of administration used by Paganelli (incubation or injection) do not reflect the likely routes of human exposure (oral, dermal, or inhalational) or the protective effect of the placental barrier (BVL, 2010).
Above all, as discussed later in this Section, glyphosate has been tested in numerous developmental studies over a 20—year period in rats and rabbits without causing malformations of the head and neural tube, even at doses high enough to be toxic to the mother and foetus.
2.1.2 Effects in laboratory animals
The major theme of the EOS article is that glyphosate has shown teratogenic activity in industry-sponsored developmental toxicity studies in rats and rabbits, with effects on foetuses including mortality, reduced ossification (bone formation) and increased incidences of skeletal and visceral abnormalities. Furthermore, EOS claims that these findings were wrongly dismissed in the EU (1998) review of glyphosate performed by the German Bundesamt fur Verbraucherschutz und Lebensmittelsicherheit (BVL). In particular, EOS criticises the use of historical control (HC) data to assist in deciding whether foetal malformations and anomalies were related to treatment, or occurred by chance. EOS was concerned that HC data introduced variability into the analysis and obscured the teratogenic effects of glyphosate.
EOS’s comments on the BVL evaluations of specific studies can be summarised as follows:
The BVL evaluation of Tasker et al (1980a) did not consider that an increased incidence of foetal malformations in rats at the highest dose (3500 mg/kg bw/d) was treatment-related, because the incidence lay within the HC range. EOS regards this as unjustifiable, due to the findings of malformations in other studies with glyphosate. EOS also criticises the BVL’s definition of sternebral unossification as a variation, rather than a malformation.
EOS disagrees with the BVL assessment of Suresh (1993a), a developmental study in rabbits at 20, 100 and 500 mg/kg bw/d in which there was an increase at all doses in major visceral anomalies, including dilated heart. Suresh concluded that the NOEL for maternotoxicity was 20 mg/kg but there was no NOEL for foetal visceral malformations. The BVL dismissed the biological significance of the foetal findings, and set the NOEL at 100 mg/kg bw/d based on comparison with HC data.
The BVL evaluation of Brooker et al (1991b; a gavage study in rabbits at 50, 150 and 450 mg/kg bw/d) was criticised for dismissing an increased incidence of foetal heart malformations at the high dose by reference to HC data.
EOS criticises the BVL’s assessment of Bhide and Patil (1989; a developmental study in rabbits at 125, 250 and 500 mg/kg bw/d), which assigned a NOAEL of 250 mg/kg for developmental toxicity based on embryo- and foetal lethality and visceral and skeletal malformations at the high dose. EOS believes that heart, lung and kidney malformations were increased at all doses, while rudimentary 14 rib was increased at 250 and 500 mg/kg.
EOS does not concur with the BVL evaluation of an anonymous (1981) oral feeding study in rabbits, in which increased foetal mortality at 50.7 and 255 mg/kg bw/d was not attributed to treatment because the doses were “far below those at which foetal effects were found in the gavage studies.”
The mammalian toxicology of glyphosate has been reviewed by several national and international pesticide regulatory agencies and scientific organisations, including the APVMAth, the US EPA, the EU and the Joint FAO/WHO Meeting on Pesticide Residues (JMPR). Between them, these agencies have evaluated eight developmental toxicity studies with glyphosate in rats and seven in rabbits. Kimmel et al (2013) and Williams et al (2012) have also reviewed developmental studies with glyphosate in laboratory animals.
220.127.116.11 Effects in rats
The German BVL assessed six rat developmental studies for the EU and/or JMPR. These are summarised in Appendix 1. There was a wide span of doses, ranging from 22 to 3500 mg/kg bw/d. According to the BVL, maternotoxicity was seen as clinical signs and reduced bodyweight gain at >1000 mg/kg, with maternal deaths at 3500 mg/kg. Effects on foetuses comprised increased incidences of wavy ribs, unossified sternebrae2, and incompletely ossified finger / toe bones, cranial centre and vertebral arches at >1000 mg/kg; with increased mortality and depressed litter and mean foetal bodyweights at 3500 mg/kg. Overall, the lowest NOEL for maternal and foetal effects in rats was 300 mg/kg bw/d, a dose 1000-times higher than the Australian ADI for glyphosate.
After closely examining the German evaluations for the EU and JMPR, the APVMA supports the BVL’s conclusions, including those relying on HC data3. Indeed, it is possible to rebut EOS’s claim that the BVL incorrectly dismissed the treatment-relatedness of dwarfism and bent tail seen at 3500 mg/kg bw/d in Tasker et al (1980a). The US EPA, Australian DoHA and Kimmel et al (2013) have also evaluated this study, and independently reached the same conclusions as the BVL. The DoHA (1985) attributed the malformations to genetic factors because all dwarf foetuses were in one litter, all those with bent tails were confined to another litter, and the control and 3500 mg/kg groups had the same number of litters with malformed foetuses. Furthermore, based on the available evaluation reports, neither dwarfism nor bent tail occurred at any dose in the other rat studies, or in rabbits.
However, there are possibly significant findings in Dallegrave et al (2003), a developmental toxicity study in which pregnant rats were dosed orally from
GD 6–15 with a Roundup formulation containing 360 g/L glyphosate and 18% w/v polyethoxylated tallow amine (POEA)4. The doses were equivalent to 500, 750 or 1000 mg glyphosate/kg bw/d. Based on increased mortality in dams at the highest dose, the apparent NOEL for maternotoxicity was 750 mg/kg bw/d but this is uncertain because Dallegrave et al did not report clinical signs, even in dams which died. The test formulation did not affect foetal survival or growth, but from 500 mg/kg upwards caused skeletal abnormalities including ossification deficits, absent and wavy ribs, absent vertebrae, and divided sternebrae and supraoccipital and interparietal bones.
The fact that the test formulation caused malformations at half the lowest foetal LOEL in rat studies with glyphosate active constituent (1000 mg/kg bw/d; see above) suggests that formulation adjuvants caused or contributed to the effects. When Holson (1990) administered POEA to pregnant rats by gavage on GD 6–15 at 15, 100 and 300 mg/kg bw/d, there was significant maternal toxicity at 300 mg/kg while decreased food consumption and mild clinical signs occurred in dams at 100 mg/kg. The maternal NOEL was 15 mg/kg bw/d. Foetal growth and development were unaffected, so the NOEL for developmental toxicity was 300 mg/kg bw/d. In Dallegrave et al (2003), by comparison, rat dams were exposed to POEA in the test formulation at ca 250, 375 or 500 mg/kg bw/d, exceeding the maternal LOEL of the pure surfactant by 2.5 to 5-fold. Furthermore, Dallegrave’s mid and high dose dams received more POEA than administered in Holson’s study (Williams et al 2000; Williams et al, 2012).
Williams et al (2012) have also noted anomalies in the numbers of foetuses, corpora lutea and implantations reported by Dallegrave et al (2003), and commented that Dallegrave used a non-standard method for fixing and protein-digesting foetuses prior to skeletal examination, which may have created areas that appeared to be incompletely ossified. Given the reporting and methodological issues identified in Dallegrave et al (2003), and because there are no other known developmental toxicity studies with GBHFs that can be compared with Dallegrave’s study, the APVMA can not reach any further conclusions on Dallegrave’s findings.
18.104.22.168 Effects in rabbits
Six of the nine known developmental studies with glyphosate in rabbits have been assessed by the German BVL for the EU and/or JMPR. Two other sponsored regulatory studies have been assessed by Kimmel et al (2013), and a further study (Stauffer Chemical Co, 1983b) was evaluated by the Australian DoHA. The doses spanned from 10 to 500 mg/kg bw/d. Evidence of maternotoxicity was fairly consistent between studies, but the threshold doses for each effect varied widely. Clinical signs and bodyweight depression occurred at >40 mg/kg, with increased maternal mortality and abortion at >100 mg/kg and decreased food consumption and bodyweight gain at >150 mg/kg. Due to the varying LOELs, maternal NOELs lay between 20–250 mg/kg bw/d.
Four gavage studies did not demonstrate any effects on foetuses at the highest doses administered (100 mg/kg bw/d in Stauffer Chemical Co, 1983b; 300 mg/kg in Hojo, 1995; 350 mg/kg in Tasker et al, 1980b and 400 mg/kg in Coles and Doleman, 1996). In four gavage studies there was fetotoxicity, seen as bodyweight depression and reduced skeletal ossification at 300 mg/kg, increased mortality at >450 mg/kg and extra 13 rib or unilateral 14thth rib at 500 mg/kg bw/d.
Visceral abnormalities occurred in six studies. These included heart or ventricular dilation and cardiomegaly, the incidences of which were elevated at 20, 100 and 500 mg/kg bw/d in Suresh (1993a). By reference to HC data, the BVL concluded that the effects were biologically significant only at the high dose, and set the foetal NOEL at 100 mg/kg bw/d. Intra-ventricular septal defect (either alone or combined with other cardiac abnormalities) was reported in Brooker et al (1991b), Bhide and Patil (1989), Hojo (1995) and Moxon (1996). Brooker et al observed incidences of 3.6% and 5.3% at 150 and 450 mg/kg bw/d, compared with 0.6% among study controls. However, given that the incidences lay within the HC range (0.7–5.9%), the BVL did not ascribe the finding to treatment at either dose. Septal defect was increased at 125, 250 and 500 mg/kg bw/d in Bhide and Patil (incidences were 0.9, 0.8 and 2.6% vs zero among controls). The BVL evaluator reasoned that the finding was unlikely to have been caused by glyphosate but could not exclude a relationship to treatment at 500 mg/kg bw/d. The APVMA concurs with this view, especially in the absence of HC data from the study laboratory. Also in Bhide and Patil, but no other study, there were elevated incidences of absent kidney (0.9, 1.8, 1.6 and 7.7% at 0, 125, 250 and 500 mg/kg bw/d) and postcaval lung lobe (0, 0.9, 1.6 and 5.1% in the respective groups). Again, the BVL attributed the findings to treatment at 500 mg/kg but not at lower doses.
Hojo (1995) reported one foetus affected by interventricular septal defect and hypoplasia of the pulmonary artery at 100 mg/kg bw/d, but no cardiac abnormalities at 10 or 300 mg/kg. Coles and Doleman (1996) observed a foetus with a heart and great vessel defect at 200 mg/kg bw/d but no cases at 50 or 400 mg/kg. Moxon (1996) found three foetuses having heart defects involving septation, one each at 0, 100 and 300 mg/kg bw/d. In these latter three studies, it is clear that the cardiovascular abnormalities were unrelated to treatment.
Overall, the range of foetal NOELs in rabbits was 100–400 mg/kg bw/d, overlapping the lowest foetal LOEL of 300 mg/kg bw/d. The margin between the lowest foetal NOEL and the Australian ADI is 333. Examining the dose-effect relationship in the rabbit gavage studies, the most sensitive end-points are foetal bodyweight and skeletal ossification, which were depressed at 300 mg/kg. If cardiac dilation, ventricular septal defect and major visceral malformations (including missing lung lobes and kidney) were indeed caused by glyphosate, by any reasonable interpretation they are confined to the 450 and 500 mg/kg groups. The margin between the doses causing these effects and the Australian ADI is 1500.
The final issue in rabbits involves a seriously-deficient study report of increased foetal deaths occurring at 50.7 and 255 mg/kg bw/d in a developmental study by dietary administration (Anon, 1981). The BVL assigned a NOEL of 10.5 mg/kg bw/d but highlighted the inconsistency between these particular findings and the results in the gavage studies, in which foetal mortality was not enhanced below 300 mg/kg bw/d. Based on the comparative weight and strength of evidence, this comment is entirely reasonable.
2.1.3 Epidemiological evidence
According to EOS, a report commissioned by the state government of Chaco, Argentina (CPICA, 2010), found an increase of nearly four-fold in the rate of malformations over a decade, coinciding with the expansion of agriculture into the region and a corresponding rise in the use of agrochemicals, including glyphosate.
EOS, Paganelli et al (2010) and Carrasco (2011) cite Benitez-Leite et al (2009) as finding that Paraguayan women exposed to herbicides during pregnancy were more likely than unexposed women to deliver offspring with malformations. These included microcephaly or anencephaly (small head or absence of a cranium), facial defects, myelomeningocele (protruding brain), cleft palate, synotia (ears extended below the jaw), polydactyly (too many fingers / toes) and syndactyly (fused digits). The specific risk factors identified were living near treated soy fields, dwellings located <1 km from treated fields, storage of pesticides in the home, and contact with pesticides (Carrasco, 2011).
EOS also claims that Savitz et al (1997) found high levels of premature births and miscarriages in female members of Canadian farming families that used pesticides, including glyphosate.
According to the BVL (2010), Mulet (2011) and Saltmiras et al (2011), the database studied by Benitez-Leite et al was small and confined to children born in one hospital. Benitez-Leite et al suspected a relationship between malformations and pesticide (not specifically herbicide)exposure but did not provide evidence of maternal exposure to glyphosate, or even mention glyphosate in their article. The association between “living near treated fields” and congenital malformations was weak, with an odds ratio (OR) 1/6th of the reported association between malformations and pesticide storage at home.
The “Ontario Farm Family Health Study” (Savitz et al, 1997) has been assessed by the JMPR (2004b), the Australian DoHA (2005), Mink et al (2011) and Williams et al (2012). In a cross-sectional study of 1898 couples and 3984 pregnancies, Savitz et al examined the association between pregnancy outcome and the father’s exposure to pesticides during the three months before conception. The study relied on mail questionnaires, with telephone interviews of non-respondents. Couples were asked to provide information on all pregnancies (of which over 1/3rd had occurred over 10 years previously) and farm activities and pesticide use over the previous five years. Not all reports of adverse pregnancy outcomes were confirmed from medical or other records, and the study was uncontrolled for maternal age, smoking and previous history of spontaneous abortion.
There were no statistically significant associations with the use of glyphosate alone. There were slightly increased odds ratios (OR) but no statistically significant associations between miscarriage and paternal use of herbicides and glyphosate on crops (17 exposed cases, OR = 1.5; 95% Confidence Interval = 0.8–2.7) or in the yard (13 exposed cases, OR = 1.4; 95% CI = 0.7–2.8). Based on five exposed cases, the OR for pre-term delivery and use of herbicides and glyphosate on crops was 2.4 but the risk estimate was of low precision (the 95% CI was 0.8–7.9). There was no association between the use of farm chemicals and small-for-gestational age births or sex ratio.
DoHA questioned the apparent association between miscarriage and herbicide/glyphosate application due to the small number of cases and the imprecision of the risk estimate, noted that the study authors had not directly tested for association between glyphosate use and reproductive effects, and observed that the study was further weakened by the lack of quantitative exposure assessment and data on the time spent using pesticides. The JMPR assessment commented that the claimed associations were weak, were not controlled for confounding factors including other pesticides, and did not meet generally accepted criteria for determining causal relationships.
Sanin et al (2009) undertook a retrospective cohort study of time to pregnancy (TTP) among 2592 fertile women living in five regions of Colombia, between which there was variation in the use of glyphosate-based herbicides. Glyphosate was not used in the region with lowest risk of prolonged time to pregnancy (TTP). The region with greatest risk (fecundability5 OR of 0.15; 95% CI = 0.12–0.18) was a sugar cane-growing district with a prolonged history of use of glyphosate and other chemicals. Glyphosate was applied to illegal crops in two of three other regions with enhanced risk, but not in the third, an organic agriculture area. The study authors concluded that the observed differences in TTP remained unexplained.
Numerous other epidemiological studies have examined datasets for associations between glyphosate and adverse reproductive outcomes, but found little evidence that glyphosate is causing ill health within human populations. Furthermore, many of these studies are weakened by shortcomings including survey methods prone to inaccurate or biased recall of pesticide exposures; lack of quantitative information on the timing, duration and extent of exposures; and the absence of appropriate controls for smoking habit, maternal age and previous reproductive history. The following publications were included in a review by Mink et al (2011) of research published over a twelve year period:
Rull et al (2006) pooled data from two Californian case-control studies evaluating neural tube defects and residential proximity to areas where pesticides were applied; mothers were considered “exposed” if any crop within 1 km had been treated with to glyphosate. Based on 45 exposed cases and 33 exposed controls, ORs of 1.4–1.5 were found depending on the regression model used for analysis. In each instance, the 95% CIs included 1.0.
In a case-control study performed in an agricultural region of Spain, Garcia et al (1998) observed no significant association between congenital malformations and the fathers’ exposure to glyphosate during the three months prior to conception or the first trimester of pregnancy (OR = 0.94; 95% CI = 0.37–2.3).
In a population of 2110 Ontario farmers’ wives from the Ontario Farm Family Health Study, Arbuckle et al (2001) reported a borderline significant association between pre-conception exposure to glyphosate and spontaneous abortion (33 exposed cases; OR = 1.4; 95% CI = 1.0–2.1), but no significant association with post-conception exposure (22 exposed cases; OR = 1.1; 95% CI = 0.7–1.7). Arbuckle and co-workers considered their investigation as “exploratory” and noted many limitations to their study, including the potential for inaccurate classification of pesticides and timing of exposure relative to conception. They also cautioned that the results should be interpreted with care and confirmed in further investigations.
To investigate whether reported pesticide use by men or women was associated with delayed pregnancy, Curtis et al (1999) measured the conditional fecundability6 ratio (CFR)7 in 2012 planned pregnancies among the Ontario Farm Family Health Study farming couples. The CFR for women who had used glyphosate (regardless of men’s use) was depressed (0.61; 95% CI = 0.30–1.3) but there was no statistical significance. Fecundability was slightly elevated (CFR = 1.3; 95% CI = 1.07–1.56) in men who had used glyphosate but whose wives had not. The study authors attributed this finding to uncontrolled factors or chance.
Self-reported glyphosate exposure during pregnancy was inversely associated with gestational diabetes (OR = 0.61; 95% CL = 0.26–1.48) in a cross-sectional analysis of data from the Agricultural Health Study by Suldana et al (2007).
Self-reported use of glyphosate was associated with a small, statistically non-significant increase in birthweight in the most recent offspring of 700 women in the US Agricultural Health Study (Sathyanarayana et al, 2010).
Garry et al (2002) conducted a cross-sectional analysis of pesticide applicators and their families. Parent-reported ADD / ADHD in children was associated positively and significantly with use of glyphosate, with 6/14 affected children having parents who had exposure to glyphosate or Roundup (OR = 3.6; 95% CI = 1.35–9.65). ADD / ADHD diagnosis was not confirmed by a clinician, however.
A further review of the scientific literature (Williams et al, 2012) concurred with the conclusion of Mink, i.e., that no consistent effects of glyphosate exposure have been found on reproductive health or offspring development in either humans or animals.