Occupational exposure to pesticides and central nervous system tumors: results from the CERENAT case–control study

The etiology of the central nervous system (CNS) tumors remains largely unknown. The role of pesticide exposure has been suggested by several epidemiological studies, but with no definitive conclusion. To analyze associations between occupational pesticide exposure and primary CNS tumors in adults in the CERENAT study. CERENAT is a multicenter case–control study conducted in France in 2004–2006. Data about occupational pesticide uses—in and outside agriculture—were collected during detailed face-to-face interviews and reviewed by experts for consistency and exposure assignment. Odds ratios (ORs) and 95% confidence intervals (95% CI) were estimated with conditional logistic regression. A total of 596 cases (273 gliomas, 218 meningiomas, 105 others) and 1 192 age- and sex-matched controls selected in the general population were analyzed. Direct and indirect exposures to pesticides in agriculture were respectively assigned to 125 (7.0%) and 629 (35.2%) individuals and exposure outside agriculture to 146 (8.2%) individuals. For overall agricultural exposure, we observed no increase in risk for all brain tumors (OR 1.04, 0.69–1.57) and a slight increase for gliomas (OR 1.37, 0.79–2.39). Risks for gliomas were higher when considering agricultural exposure for more than 10 years (OR 2.22, 0.94–5.24) and significantly trebled in open field agriculture (OR 3.58, 1.20–10.70). Increases in risk were also observed in non-agricultural exposures, especially in green space workers who were directly exposed (OR 1.89, 0.82–4.39), and these were statistically significant for those exposed for over 10 years (OR 2.84, 1.15–6.99). These data support some previous findings regarding the potential role of occupational exposures to pesticides in CNS tumors, both inside and outside agriculture.


Introduction
Epidemiological knowledge regarding central nervous system (CNS) tumors, including incidence data and etiological research, remains limited. Data from several population-based cancer registries demonstrate that their annual incidence reaches as much as 20/100 000 (Ostrom et  study in Northern Italy found a four times higher risk of glioma with exposure to nitrosable fungicides or insecticides (Musicco et al. 1988). More recent studies have used re ned pesticide exposure assessment based on detailed questionnaires and/or expertise and analyzed CNS tumor by subtypes. In the Upper Midwest Study, a large case-control study in the USA including 341 brain tumors in women and 457 in men, no increase in risk was observed for glioma with pesticide exposure globally (Ruder et al. 2006;Yiin et al. 2012), in men (Ruder et al. 2004), and only a trend restricted to carbamate insecticide use in women (Carreón et al. 2005). Another case-control study in Nebraska observed an increase in risk only for workers involved in farming for 55 years and over (Lee et al. 2005). Using a job-exposure matrix for pesticide use de nition, a third case-control study in the USA evidenced a doubling in the risk of meningioma in women exposed to herbicides, while no association was found between gliomas and pesticide exposure (Samanic et al. 2008). In a case-control study conducted in Southwestern France, an increase in risk was shown among the individuals who were the most exposed to pesticides, especially in vineyards and it was more pronounced for gliomas (Provost et al. 2007). More recently, additional results were obtained from two large prospective agricultural cohorts. The Agricultural Health Study in the USA found a signi cant fourfold increase in farmers most exposed to chlorpyrifos (Lee et al. 2004) and a threefold increase in their spouses using domestic or occupational organochlorines (Louis et al. 2017). In the French Agrican cohort, elevated risks were observed in farmers exposed to speci c crops (peas, beets, potatoes) and risk doubled in those who applied pesticides (Piel et al. 2017). Analysis focused on carbamate insecticides, fungicides and herbicides pointed to some speci c active ingredients being associated with a high increase in the risk (Piel et al. 2019a(Piel et al. , 2019b.
We present here new results from a multicenter case-control study in France, exploring the relationship between CNS tumors and pesticide occupational exposures, taking into account the CNS subtypes and accurate de nitions of pesticide exposures.

Population
Brie y, the CERENAT case-control study, has been described in detail elsewhere (Coureau et al. 2014): it was conducted in four French administrative areas (Gironde, Calvados, Manche and Hérault) and included i) patients aged 16 and over with a diagnosis of incident benign or malignant CNS (brain and spinal cord) tumor between June 2004 and May 2006, established either by a neuro-pathological assessment or by clinical and radiological assessment for cases with no histological diagnosis and living in one of the four areas when diagnosed, ii) for each case, two controls with no history of CNS tumor, randomly selected from the local electoral rolls during the period 2005-2008, individually matched on age (+/-2 years), sex and area of residence. Primary brain tumors with the following ICDO-3 topography codes were included: C70.0-C70.9, C71.0-C71.9 and C72.2-C72.9. Patients with recurrent tumors, metastases, pituitary tumors, genetic syndrome or AIDS were excluded. Cases were grouped by morphology codes as gliomas, meningiomas, and other tumors (acoustic neurinomas, lymphomas and other unspeci ed primary brain tumors).

Data collection
Data were collected in standardized questionnaires by trained interviewers during face-to-face structured interviews. When patients were in a severe clinical condition or deceased, a proxy was invited to complete a simpli ed questionnaire. For all participants, socio-demographic characteristics, medical history, lifestyle, environmental data and complete occupational histories with beginning and end dates (including kind of industry, activity and main tasks) were collected.
Pesticide exposure assessment Besides job calendars, in-depth life-long questionnaires on pesticide exposures were completed by all subjects who gave a positive answer to one or more of the following 4 screening questions "Have you ever worked on a farm?", "Have you ever applied pesticides on crops?", "Have you ever worked in contact with crops?", "Have you ever used pesticides in non-agricultural jobs?". The speci c questionnaires included detailed information on various tasks (beginning and end dates, number of days per year, equipment used) including treatment (mixing, spraying, equipment cleaning) and other tasks (pruning, thinning, bending/tying up, harvesting, etc.). Three experts in pesticide exposure (PFP, PL, IB) reviewed all the job calendars and speci c questionnaires independently for consistency and to determine exposure parameters for each individual: direct or indirect exposure in vineyard, open eld, fruit and vegetable growing, gardening, wood industry and pest-control. Duration of exposure (in years and in cumulated life-long days) was determined for each type of exposure and the median of the distribution was considered as the threshold for exploring risks in those most exposed.

Analysis
The index date for each case and his two matched controls was the date of case diagnosis. Pesticide exposure was considered as a binary variable (yes/no) for each exposure parameter, and also according to the median of cumulative exposure de ned in controls. Cumulative exposure was de ned as the sum of lifetime days of treatment for agricultural exposures and as the duration of jobs (in years) for non-agricultural exposures. Conditional logistic regression for matched sets was used to estimate odds ratios and 95% con dence intervals. All statistical tests were two-sided and a global test for each categorical indicator was performed. The following variables were considered as potential confounders: level of education (primary school or less, secondary school, high school and university), smoking (in pack-years), alcohol consumption (classi ed as excessive in men over 21 glasses of wine, cider, beer or spirits per week and over 14 glasses per week in women). None of these variables was retained in the nal analysis as alcohol and smoking were not associated with health outcome and level of education was closely correlated to the jobs and exposures under study. Separate analyses were run for each histological type. We used 2-sided statistical tests and a 5% signi cance level. Analyses were performed with the SAS ® software, version 9.2 (SAS Institute Inc, Cary, NC, United States).

Population characteristics
Among the eligible subjects, 95% of cases and 61% of controls were reached. A total of 596 (73%) cases and 1,192 (45%) controls were nally included in the CERENAT study. The participation rate was 66% for glioma cases and 75% for meningioma cases. The main reasons for non-participation were refusals, severe condition or death without proxy. Non-included cases were older than included cases (mean age: 63 years vs. 58 years).
The cases were 273 patients with gliomas, 218 with meningiomas, and 105 with other brain tumors (68 neurinomas, 12 lymphomas, and 25 unde ned or others). The proportion of proxy interviews was 25% for gliomas, 6% for meningiomas and 18.5% for other brain tumors. The average age was 55.4 years for patients with gliomas, 60.2 years for meningiomas and 57.6 for other tumors (Table 1). Women represented 42.1%, 75.7% and 53.3% of the population for each type of tumor respectively. The level of education was higher in controls than in cases whatever the type of tumor. The proportion of alcohol excessive consumers was slightly higher in controls (8.5%) than in cases (6.2%). Life-long tobacco smoking concerned half of the population, in a comparable proportion in cases and controls, although the proportion of smokers who exceeding 20 pack-years life-long was slightly higher in cases (41.2%) than in controls (34.8%). Description of pesticide exposure Information on occupational pesticide exposure was available for all subjects, and about 2% of data on exposure duration were missing. Figure 1 presents the proportion of individuals exposed to pesticides according to the status (all controls, all cases, gliomas, meningiomas) and the type of exposure (direct and indirect in agriculture and outside agriculture).

Agricultural pesticide exposures
A total of 754 subjects (42.2%) were classi ed as exposed to pesticides in agriculture. One hundred and twenty ve individuals (

Non-agricultural pesticide exposures
One hundred and forty-six subjects (8.2%) were classi ed as exposed to pesticides in non-agricultural occupations: 36 (2.0%) in gardening & landscape (24 of them exposed in treatment tasks and 12 through indirect exposures), 90 (5.0%) in the wood industry (20 directly and 70 indirectly exposed), and 21 (1.2%) in insect control for public health (15 directly and 6 indirectly exposed). The medians of the duration of exposure were 10 years in gardening, 14 years in wood industry and 15 years in pest control.

Risks associated with agricultural exposures
Associations between agricultural exposures and CNS tumors are presented in Table 2. Direct agricultural exposure was not signi cantly associated with an increase in risk for all brain tumors together and all types of exposures (OR=1.04, 0.69-1.57) but a slightly positive trend was observed when restricting the analysis to gliomas (OR=1.37, 0.79-2.39), while a decrease in risk was observed in meningiomas (OR=0.79, 0.36-1.76) and other tumors (OR=0.65, 0.23-1.87) ( Table 3). In the most exposed farmers (life-long number of days of treatment over the median), the risk of brain tumors overall was increased (OR=1.58, 0.83-3.01) and the risk of glioma was doubled (OR=2.22, 0.94-5.24).

Risks associated with non-agricultural pesticide exposures
Associations between non-agricultural exposures and CNS tumors are presented in Table 3. In green spaces, where 16 cases and 20 controls were exposed, overall risk of brain tumors tended to increase in subjects directly exposed to pesticides (OR=1.89, 0.82-4.39), and to a lesser extent in indirectly exposed subjects (OR=1.43, 0.45-4.50). The risk was more than doubled and signi cant for subjects with the longest exposures (over 10 years) (OR=2.84, 1.15-6.99).
Increases in risk were found in analysis restricted to gliomas (direct exposure: OR=2.19, 0.58-8.6) while numbers were too limited for conclusions in meningioma (5 cases and 1 control).  In the wood industry, where 34 cases and 56 controls were exposed, a slight increase was observed for all CNS tumors, in indirectly (1.27, 0.78-2.08) and In pest control workers, only 9 cases and 12 controls were exposed and the risk of brain tumor was increased only in directly exposed (OR=1.75, 0.64-4.83) with too limited numbers for conclusions by subtypes.

Discussion
In this case-control study in France, increases in the risk of central nervous system tumors were observed in relation to various occupational pesticide exposures in agriculture but also outside agriculture. In the most exposed individuals, the risk of glioma was non-signi cantly trebled in open eld and in fruitgrowing and multiplied by 1.68 in vine-growing. Elevated risks of glioma were also observed for non-agricultural use of pesticides, especially in workers in green spaces and in the wood industry, and to a lesser extent in pest control workers. No increase in risk was seen for meningioma, nor in indirectly exposed individuals in agriculture. Although the total number of cases included was quite large, most of the elevated risks we found did not reach the statistical signi cance because of limited numbers when considering tumor subtypes together with speci c types of pesticide exposures. Thus, these results can only be interpreted as trends, but they are important to consider because of the strength of some associations that we observed. Moreover, these results were globally consistent, showing higher risks in gliomas and in most exposed individuals for almost all the types of pesticide use.
The main strengths of this study include the enrolment of incident cases supported by population-based cancer registries, face-to-face interviews, the analysis of sub-types of tumors (gliomas, meningiomas), accurate pesticide exposure assessment (exploring agricultural and non-agricultural jobs, direct and indirect exposures). Considering the 73% participation rate in cases and 45% in controls, we cannot rule out selection bias. However, the lower participation of subjects with gliomas and elderly people, more frequently exposed to pesticides as shown by observations of participants, is likely to have decreased our risk estimates and biased our results toward the null. Apart from this, we do not see any clear reasons why the participation would be related to pesticide exposures especially as the study was presented to participants as dealing with environmental and occupational factors and CNS health in general, without mentioning the hypothesis on pesticides. Recall bias is a concern in our study as in any retrospective study. However, we believe that this bias was limited by the review of exposure data by experts, who considered job titles as well as responses to speci c questionnaires to ensure exposure assessment consistency.
One of the lessons from our study, as already raised by results from a previous study, is the di culty of highlighting associations when histological types of CNS tumors and kind of exposures are not analyzed separately. This could explain why studies using imprecise metrics for pesticide exposure, such as job Results on fruit and vegetable growing are less conclusive, but indicated a trend towards an increased risk of glioma in the most exposed, for which no other evidence in the literature has been found. Few studies have explored the association of CNS tumors with pesticide exposures outside agriculture. We observed a signi cant doubling in risk in green space workers, consistent with a study in golf course workers that found an elevated mortality for CNS tumors compared to the USA general population (Kross et al. 1996). The non-signi cant doubling of risk that we found in workers exposed in the wood industry is in line with a case-control study on gliomas that showed a raised risk among wood workers, attributed to exposure to organochlorine wood preservative and solvents (Cordier et al. 1988). Our conclusions on pest control workers are limited because of small numbers, but the slight increase in risk we found in the most exposed is in line with a study in Roma that found an excess in CNS tumor mortality in a retrospective cohort of pest control workers (Figà-Talamanca et al.

1993).
In this study, we have not explored the role of speci c active ingredients, as people generally cannot remember them life-long. However, the increases in risk we observed across several crops and also outside agriculture, suggest either of the role of a large range of pesticides or of the role of pesticides that have been ful ll the second hypothesis as they have been used as insecticides on crops (including seed treatment) and animals, as well as herbicides and fungicides (mainly dithiocarbamates), but we cannot rule out the possibility that several other molecules, among the more than 1,000 that have been marketed since 1950, could play a role.

Conclusions
In conclusion, our study brings new evidence on the association of pesticides and CNS tumors in agriculture: the associations were statistically signi cant in open eld workers, but trends were also observed in vine-growing and fruit growing, especially for gliomas and for the most exposed workers. Even newer evidence are the associations we found outside agriculture, with almost a trebling in risk among green spaces workers exposed for more than 10 years. These results have implications for enhancing preventive measures in agriculture but also for paying particular attention to pesticide exposures outside agriculture.