Eriksson et al (2008): This was a population-based case-control study of exposure to pesticides as a risk factor for non-Hodgkin lymphoma (NHL), consisting of 910 cases and 1016 controls. The subjects were men and women aged 18 – 74 years living in Sweden, diagnosed with NHL between December 1999 and April 2002. All cases were diagnosed and classified histopathalogically according to WHO criteria. Controls were selected from the national population registry.
Exposure assessment was performed by a questionnaire which included work history, exposure to pesticides, organic solvents and several other (unidentified) chemicals. For dose-response analysis of pesticides, information was collected on the number of years, days per year and hours per day of exposure. The questionnaire also included smoking habit, medications, leisure activities and residential proximity to industrial installations, but data on these variables were not included in the review. Supplementary phone interviews were conducted if necessary. All exposures of less than a full day, or occurring during the same calendar year as the diagnosis or one year prior, were disregarded.
Data were analysed by unconditional logistic regression (univariate and multivariate) adjusted for age, sex and year of diagnosis or enrolment. In the univariate analysis, different pesticides were analysed separately, and the unexposed category consisted of subjects who were not exposed to any of the included pesticides. All controls were used in the analyses of NHL subgroups. In the dose-response calculations made for agents with at least 20 exposed subjects, the median number of days of exposure among controls was used as a cut-off. Latency period calculations and multivariate analyses (performed because most pesticide exposures involved more than one chemical) included agents with statistically significantly increased ORs, or with an OR >1.50 and at least 10 exposed subjects.
Univariate analysis adjusted for age, sex and year of diagnosis or enrolment 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 before diagnosis (unstated no. of cases and controls; 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 before diagnosis (unstated no. of cases and controls; OR = 1.11; 95% CI = 0.24 – 5.08) or exposure to glyphosate for <10 days (12 cases and 9 controls; OR = 1.69; 95% CI = 0.70 – 4.07). Multivariate analysis adjusting for exposure to other chemicals yielded a low and statistically non-significant risk estimate for glyphosate (OR = 1.51; 95% CI = 0.77 – 2.94).
When the different sub-types of NHL were analysed separately, exposure to glyphosate was associated with a significantly enhanced risk of small lymphocytic lymphoma / chronic lymphocytic leukaemia (195 cases; OR = 3.35; 95% CI = 1.42 – 7.89) and unspecified NHL (38 cases; OR = 5.63; 95% CI = 1.44 – 22.0). Odds ratios for other types of lymphoma were not statistically significant.
The same research group have published a previous (Hardell et al, 2002) epidemiology study on the association between pesticide exposure and NHL, in which univariate analysis found a significant association with glyphosate (OR = 3.04; 95% CI = 1.08 – 8.52) based on 8 cases and 8 controls. Noting the small sample size and the broad CI, the Australian DoHA (2005) concluded that strength of association was questionable, and it was equivocal whether glyphosate was indeed a risk factor for NHL.
The current follow-up study improves on its predecessor in several respects, as it was based on a larger population (910 vs 515 cases), had larger sample sizes, included both men and women, and collected exposure data from living individuals only.21 The follow-up would therefore have increased statistical power and diminished recall bias. Compared with the 2002 study, the risk estimate was lower (OR of 2.02 vs 3.04) but the association between glyphosate exposure and NHL was strengthened, as evidenced by the narrower 95% CI (1.10 – 3.71 vs 1.08 – 8.52). However, the 2008 and 2002 studies failed to demonstrate associations by multivariate analysis, which yielded ORs of only 1.51 and 1.85, with 95% CIs that had lower bounds of less than 1.0 (0.77 – 2.94 and 0.55 – 6.20). Eriksson et al (2008) noted that many glyphosate users had previously been exposed to MCPA, and suggested this as an explanation for why neither chemical showed a significant OR when subjected to multivariate analysis.
At best, the association between glyphosate and NHL in this study is equivocal, remains potentially confounded by established risk factors such as immunosuppression and Epstein-Barr virus (as noted previously by the Australian DoHA, 2005), and could also have been affected by recall, exposure measurement and information bias if NHL cases or their interviewers believed that their disease may be related to pesticides (Mink, unpublished). Mink has also observed that, by excluding 88 potential cases who died before they could be interviewed, the study population did not represent those cases with more aggressive disease. Furthermore, the dose-response analysis may have been confounded by exposure to other herbicides, and was based on unequal cut-off points for glyphosate (<10 days or >10 days) and “other” herbicides (<32 days or >32 days) (Mink, unpublished).
APPENDIX 5: PHARMACOKINETICS OF GLYPHOSATE AND ITS METABOLITE AMPA IN RATS
Anadon et al (2009): Laboratory grade glyphosate (Sigma Chemical Co, St Louis, MO, USA; purity 95%) was administered to male Wistar rats (Charles River Inc, Margate, Kent, UK; bw 200 – 210 g) at 100 mg/kg bw IV (in 0.1 mL glycerol formal) or 400 mg/kg PO (gavage to fasted animals in 0.5 mL corn oil). Groups of 8 rats were killed and exsanguinated at 0.25, 0.5, 1, 2, 4, 6, 8, 10, 12 and 24 h post-dosing, and the concentrations of glyphosate and aminomethyl phosphonic acid (AMPA) were measured in plasma by HPLC with fluorescence detection.
Glyphosate, IV administration: Following an initial peak concentration (Cmax) of 166 µg/mL plasma pharmacokinetics were biphasic, consistent with a two-compartment open model, with rapid distribution and gradual elimination. The volume of distribution at steady state was 2.99 L/kg, suggesting extensive diffusion into the tissues. Clearance was 0.995 L/h/kg. The elimination half-life from plasma was 9.99 h and the area under the concentration vs time curve (AUC) was 100 mg.h/L.
Glyphosate, PO administration: Absorption from the GIT was gradual, with a Cmax of 4.62 µg/mL occurring in plasma at 5.2 h. Oral bioavailability was poor (23.2%). Clearance was the same as following IV administration and the AUC was similar (at 93.3 mg.h/L), but the elimination half-life from plasma was appreciably more prolonged (14.4 h).
AMPA: The metabolite first appeared in plasma within 0.25 h of PO dosing, and had similar pharmacokinetic behaviour to glyphosate. The Cmax (0.42 µg/mL) occurred at 2.4 h. An AUC of 6.1 mg.h/L was attained, ca 6.5% of glyphosate’s AUC in plasma. The elimination half-life of 15.1 h was similar to that of the parent chemical after PO administration.