You are here

Article Text

Article menu
Workplace
Original article
Occupational exposure to pesticides and endotoxin and Parkinson disease in the Netherlands
  1. Marianne van der Mark1,
  2. Roel Vermeulen1,2,
  3. Peter C G Nijssen3,4,
  4. Wim M Mulleners5,
  5. Antonetta M G Sas6,
  6. Teus van Laar7,
  7. Maartje Brouwer1,
  8. Anke Huss1,
  9. Hans Kromhout1
  1. 1Division of Environmental Epidemiology, Institute for Risk Assessment Sciences, Utrecht University, Utrecht, The Netherlands
  2. 2Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, The Netherlands
  3. 3St Elisabeth Hospital Tilburg, Tilburg, The Netherlands
  4. 4TweeSteden Hospital Tilburg, Tilburg, The Netherlands
  5. 5Canisius-Wilhelmina Hospital Nijmegen, Nijmegen, The Netherlands
  6. 6Vlietland Hospital Schiedam, Schiedam, The Netherlands
  7. 7University Medical Center Groningen, Groningen, The Netherlands
  1. Correspondence to Marianne van der Mark, Division of Environmental Epidemiology, Institute for Risk Assessment Sciences, Utrecht University, PO Box 80.178, Utrecht 3508 TD, The Netherlands; m.vandermark{at}uu.nl

Abstract

Objectives Previous research has indicated that occupational exposure to pesticides and possibly airborne endotoxin may increase the risk of developing Parkinson disease (PD). We studied the associations of PD with occupational exposure to pesticides, specifically to the functional subclasses insecticides, herbicides and fungicides, and to airborne endotoxin. In addition we evaluated specific pesticides (active ingredients) previously associated with PD.

Methods We used data from a hospital-based case–control study, including 444 patients with PD and 876 age and sex matched controls. Exposures to pesticides from application and re-entry work were estimated with the ALOHA+job-exposure matrix and with an exposure algorithm based on self-reported information on pesticide use. To assess exposure to specific active ingredients a crop-exposure matrix was developed. Endotoxin exposure was estimated with the DOM job-exposure matrix.

Results The results showed almost no significant associations. However, ORs were elevated in the higher exposure categories for pesticides in general, insecticides, herbicides and fungicides, and below unity for endotoxin exposure. The analyses on specific active ingredients showed a significant association of PD risk with the fungicide benomyl.

Conclusions This study did not provide evidence for a relation between pesticide exposure and PD. However, the consistently elevated ORs in the higher exposure categories suggest that a positive association may exist. The possible association with the active ingredient benomyl requires follow-up in other studies. This study did not provide support for a possible association between endotoxin exposure and PD.

Keywords
  • Parkinson disease
  • pesticides
  • endotoxins

https://doi.org/10.1136/oemed-2014-102170

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Keywords

    What this paper adds

    • Occupational pesticide exposure has been associated with an increased risk of developing Parkinson disease (PD), but it is unclear which specific active ingredients are responsible.

    • It has been postulated that occupational exposure to endotoxin may increase PD risk as well.

    • In a multicentre case–control study on PD in the Netherlands, no significant associations with insecticides, herbicides or fungicides were observed, but active ingredient-specific analyses revealed a possible association with benomyl, a benzimidazole fungicide.

    • No association between endotoxin exposure and PD was apparent.

    Introduction

    It has been frequently reported in epidemiological and toxicological studies that exposure to pesticides may increase the risk of developing Parkinson disease (PD).1–3 Pesticides are widely used in agriculture and therefore farm work is an important source of exposure. Exposure occurs during the actual application of pesticides, but also during contact with crops treated with pesticides when carrying out so-called re-entry activities such as weeding or thinning. Exposure from re-entry work could be substantial depending on the crop.4 ,5

    Since a wide range of pesticides (active ingredients) have been used in the past, it is unclear which active ingredients are responsible for the reported increase in PD risk in epidemiological studies. The accuracy of self-reported information on performed applications in the past is limited for specific active ingredients.6 Moreover, farm workers who have performed re-entry work but were not involved in the application or purchase of pesticides might not be able to accurately report on the use of pesticides. The use of occupational histories to assign exposure is less affected by recall problems.7 However, only a few of the previous epidemiological studies on PD risk used job titles8–11 or a combination of job titles with self-reported information12 ,13 to assess exposure to pesticides or functional subclasses of pesticides, for example, insecticides, herbicides and fungicides.

    Recently, it has been postulated that not only pesticides may increase PD risk, but also endotoxin exposure, by inducing inflammation-mediated neurodegeneration.11 ,14 Endotoxins are the lipopolysaccharide components of Gram-negative bacterial cell walls, and exposure is common during agricultural work.

    We present the results of analyses on the possible associations between occupational exposure to pesticides and endotoxin and PD risk within a recently conducted hospital-based case–control study in the Netherlands. We used the existing general population ALOHA+job-exposure matrix (JEM)15 and self-reported exposure information to estimate exposure to insecticides, herbicides and fungicides through mixing and application work and through re-entry work in treated crops. Furthermore, we constructed a time-dependent crop-exposure matrix to estimate exposure to specific active ingredients based on self-reported cultivation of crops. For exposure to endotoxin we used the recently developed DOM-JEM.16

    Methods

    Cases and controls

    Details about the study methodology were described previously.17 Briefly, cases and controls were recruited between April 2010 and June 2012 from five hospitals in four different areas in the Netherlands. We set out to include all patients who had an initial PD diagnosis in one of the participating hospitals between January 2006 and December 2011. In the Netherlands, with a universal healthcare system, all patients with PD are seen in a hospital. Patients with PD included can therefore be regarded to be representative for all patients with PD in the service areas of the participating hospitals. In each hospital, one neurologist reviewed the medical files of all potential participants. For each confirmed patient with PD, two matched controls were selected from persons who were seen at the department of neurology between January 2006 and December 2011 for non-neurodegenerative symptoms (median nerve neuropathy; International Classification of Diseases, 10th revision (ICD-10) G56.0 and G56.1, ulnar nerve neuropathy; ICD-10 G56.2, thoracic and lumbar disc disease; ICD-10 G55.1, G54.3 and G54.4, and sciatica; ICD-10 M54.3 and M54.4). The controls were matched to the cases on hospital, visiting date (within 3 years of the case diagnosis year), sex and age. Cases and controls were initially contacted via an invitation letter from the hospitals’ neurology departments together with a reply form for giving informed consent or to decline study participation. The study information explained that the study objective was to assess risk factors for neurological disorders, without further specification. Non-responders were sent a reminder after 1 month, and one phone call attempt was performed after another month.

    At recruitment, 1001 (93% of total) eligible patients with PD were still alive and of 993 of those we had a valid current address. There were 448 persons who participated (45%), 406 who declined participation and 139 who did not reply. The participation rate for controls was 35%. About 50% of the non-participants provided a reason for their decline. Health-related reasons were reported most frequently, but compared with cases, more controls reported to be not interested. For 12 cases only 1 suitable control was found and for 4 cases no controls were found, leaving 444 cases and 876 controls who were included in the analyses.

    Data collection

    Participants were interviewed in a standardised computer-assisted telephone interview by one of three trained interviewers. The questionnaire contained an occupational history in which all jobs performed for at least 6 months were included. Study participants reported on years and hours per week worked, job title, type of industry, company name and main tasks. Supplemental questions about occupational application of insecticides, herbicides and fungicides were asked to those who reported to have worked on a farm or as a gardener. The annual number of days on which applications were performed at the job (<1/year, 1–5/year, 6–20/year, 21–50/year or >50/year) was asked. Questions about application method and use of protective equipment were asked to participants who had personally applied pesticides. Furthermore, participants who worked on a farm were asked to name the main crop types cultivated at the farm with a maximum of three. Participants reporting having worked at their parents’ farm during childhood were not always asked additional questions on working hours and pesticide applications (n=40). Based on what was reported by other participants with similar jobs, we assumed that those participants helped 8 h per week at the farm from age 12 to 18 and in case no farm type was provided we assumed it was a mixed farm.

    All jobs were coded according to the International Standard Classification of Occupations 1968 and 1988 (ISCO68 and ISCO88).

    Assessment of pesticide exposure

    Exposures were estimated from 1955 until the calendar year before diagnosis, as after this year synthetic pesticides became commonly used in the Netherlands. Occupational exposure to pesticides was estimated using three different methods.

    The first method estimated pesticide exposure by linking all reported jobs to the ALOHA+JEM.15 This JEM assigns exposure to pesticides, and to functional subclasses (ie, insecticides, herbicides and fungicides) using arbitrary weights of 0, 1 and 4 for no, low and high exposure. For farm and gardener jobs the JEM score was set to 0 if the participant reported that no insecticides (n=40), herbicides (n=41) or fungicides (n=77) had been applied. For jobs coded ISCO88 code 9333 (freight handlers), exposure to insecticides was only assigned to those jobs coded ISCO68 code 97120 (dockers). Cumulative exposures were estimated by multiplying the JEM scores with years worked in a job, summed across all jobs of a participant's occupational history.

    The second method estimated more specifically the exposure for participants who held a farm or gardener job. For these jobs, we developed an exposure model (algorithm) for insecticides, herbicides and fungicides, based on the algorithm developed for the US Agricultural Health Study (AHS) for estimating applicator exposure.18 We extended the algorithm by including estimates for exposure due to re-entry work.

    Applicator exposure was calculated for participants who reported to have applied insecticides, herbicides or fungicides personally. Exposure intensity was estimated based on the application method and use of personal protective equipment (PPE) in accordance to the AHS algorithm. However, as we had no information on performing maintenance or repair of application equipment, and because essentially all applicators in our study mixed the pesticides before use, these factors included in the AHS algorithm were left out in our study.

    For the application method, the AHS model uses relative exposure values of one for distribution of tablets/granules, three for a boom sprayer on a tractor, eight for a backpack sprayer and nine for a hand sprayer. Since the European Predictive Operator Exposure Model (EUROPOEM) shows roughly a factor two difference in exposure between manual and tractor spraying, we adjusted the AHS values slightly to one for distribution of tablets/granules, four for tractor application and eight for manual application by backpack or hand spray.19 ,20

    The values for the use of PPE were based on the AHS algorithm and were 1 for individuals not using PPE, 0.8 for individuals using gloves, rubber boots or goggles, and 0.5 for individuals who also used impermeable clothing or facemasks. If a participant reported that the use of PPE had changed over the years of working in a particular job, average values of categories were assigned.

    Yearly applicator exposure was calculated by multiplying the intensity level with the number of applications per year using the midpoint of the answer categories and the percentage of applications performed by the participant at the farm (mostly/always: 0.9, sometimes: 0.5 and rarely: 0.1).

    Yearly applicator exposure=application method×PPE× applications/year×percentage self-application by participant.

    Yearly exposure from re-entry work was estimated by multiplying the intensity level for a day of re-entry work with the number of days of re-entry work. Based on the EUROPOEM applicator and re-entry worker models, we estimated that a day of re-entry work would result in 10% of the exposure of a typical day of application work.20 ,21

    Since most applications were conducted by tractor or manually by backpack or hand spray, which in the applicator part of the model have on average an intensity level of 6, we used an intensity level of 0.6 for a day of re-entry work. The yearly number of days with re-entry work was calculated from the reported number of applications per year at the farm, the number of days a pesticide was assumed to be present on the crop after application and an estimated number of days that a participant performed re-entry work after each application. The number of applications per year at a farm was imputed for participants who did not know if pesticides were applied at a job or at what frequency, based on the most frequently reported answer by other participants for similar jobs. Although the number of days a pesticide is present on the crops after application depends on pesticide type, crop type and weather conditions, we assumed a period of 14 days based on data in the literature.5 ,22 Furthermore, we estimated that workers who worked 40 h/week at farms with horticultural or fruit crops performed 5 days/week re-entry work (ISCO68 job codes: 61230, 62320, 61270, 62720, 62730) and workers at farms with field crops 1 day/week (ISCO68 job codes: 61110, 62105, 61220, 62210). These estimations were adjusted for the number of hours per week a participant had worked.

    Cumulative total exposure was calculated by multiplying the yearly exposure from pesticide application plus re-entry work with the number of years in a job, summed across all jobs of a participant's occupational history.

    The third method assigned exposure to specific active ingredients by linking reported crops cultivated at the participant's farm to a crop-exposure matrix. In this crop-exposure matrix, per-decade estimations are given for the percentage of farms that applied a specific active ingredient on a type of crop and the yearly frequency of application. Active ingredients included in the crop-exposure matrix were based on previous studies linking specific pesticides to PD risk and for which we had sufficient data to estimate historical application. The included active ingredients were the insecticides: dichlorvos, lindane, parathion and permethrin; the herbicides: 2,4-D, atrazine, dinoseb and paraquat; and the fungicides: benomyl and maneb.

    Expert judgment on the probabilities and frequencies of application were provided by former extension workers, two per crop type. These experts estimated probability and frequency of use of active ingredients allowed for use on potatoes, cereals, beets, maize, tulip bulbs and fruit, back to the year 1960. More details about the expert estimations can be found elsewhere (Brouwer et al, submitted). Estimates for other field crops, and vegetables and flowers in green houses, which were not covered by the experts, were derived from data of Statistics Netherlands that gathered statistics on use of specific active ingredients since 1995. For earlier decades, probability and frequency of application for those crops were extrapolated from time trends for crops for which expert estimations were available. Statistics Netherlands also gathered data since 1976 on active ingredient use in public places, which we used to estimate exposure for gardeners.23

    Cumulative exposures to these active ingredients were calculated by summing the yearly probability of application of a specific active ingredient at the farm multiplied with the yearly frequency across all years worked at farms. For farms where exposure to an active ingredient was assigned to more than one crop, the probability and frequency of use for the crop with the highest probability of use were taken for calculating cumulative exposure. As in the approach with the JEM analyses described earlier, no exposure was assigned to a farm job when on that farm, according to the participant, no insecticides, herbicides or fungicides had been applied.

    Assessment of endotoxin exposure

    High, low or no exposure (weights of 4, 1, 0, respectively, were assigned to reflect the multiplicative nature of occupational exposure distributions) to endotoxin was assigned by linking the DOM-JEM with the reported jobs.16 Cumulative exposure was estimated by multiplying the JEM scores with years worked in a job, summed across all jobs of a participant's occupational history.

    Statistical analysis

    ORs and 95% CIs were calculated using conditional logistic regression. The exposed participants were categorised in two or three groups based on either median or tertiles of the distribution of the different exposures among controls. In Model 1, all analyses were adjusted for pack-years of smoking (5 levels), total coffee consumption (4 levels) and categories for occupational skill and status (high-skilled white-collar worker, low-skilled white-collar worker, high-skilled blue-collar worker and low-skilled blue-collar worker). In Model 2, the pesticide analyses were additionally adjusted for cumulative endotoxin exposure (4 levels).

    Results

    Of the patients with PD, 63.3% were men with a median age at diagnosis of 67 years (see table 1). Cases more often had high-skilled white-collar jobs than controls, smoked less and consumed less coffee. Prevalence of pesticide exposure was 19.3% for cases and 19.1% for controls as assessed by the JEM approach (see table 2). The prevalence of exposure to the functional subclasses insecticides, herbicides and fungicides was slightly higher for cases compared with controls for the JEM approach and the exposure algorithm. More controls (41.7%) than cases (38.1%) ever had a job with low or high endotoxin exposure. Most participants who held a job with high endotoxin exposure were individuals who worked at a farm with livestock. The most reported jobs with low endotoxin exposure were other farm jobs and cleaning jobs. Correlations between exposure to subclasses of pesticides as shown in table 2 were high (Spearman correlation coefficients: 0.66–0.87). More moderate correlations (Spearman correlation coefficients: 047–0.60) were observed between endotoxin exposure and exposure to (subclasses of) pesticides.

    View this table:
    Table 1

    General characteristics of cases and controls

    View this table:
    Table 2

    Prevalence of pesticide, herbicide, insecticide, fungicide, and endotoxin exposure and their correlations

    Pesticide exposure as assessed by the JEM approach

    In table 3, the analyses of cumulative pesticide exposure based on the ALOHA+JEM augmented with self-reported information on actual use of pesticides are presented. The analyses revealed no statistically significant results, but overall, relatively more cases than controls were in the higher exposure tertiles for all groups of pesticides. Increased ORs were most pronounced for exposure to insecticides. Analyses without using augmentation on self-reported actual use of insecticides, herbicides or fungicides within farm or gardener jobs, resulted in ORs closer to 1 (see online supplementary material table S1).

    View this table:
    Table 3

    Cumulative exposure to pesticides, specific subclasses and endotoxin and Parkinson disease risk: job-exposure matrix (JEM) approach

    Pesticide exposure as assessed by the exposure algorithm

    A total of 94 cases (21%) and 183 controls (21%) stated to have ever worked on a farm or as a gardener and consequently were asked supplemental questions on pesticide applications. The results of the analyses using the adjusted AHS exposure algorithm are shown in table 4. Relatively more cases than controls were in the third tertile of cumulative exposure for insecticides, herbicides and fungicides, although the elevated ORs did not reach statistical significance. Sixty-five per cent of the insecticide-exposed, 63% of the herbicide-exposed and 55% of the fungicide-exposed had not personally applied pesticides, thus for those participants only re-entry work contributed to the exposure estimates. This was especially the case for women: of the 60 women exposed to insecticides, herbicides and/or fungicides, there were only two women who had actually applied pesticides. We also performed analyses on application work only (see online supplementary material table S2). Ever having performed applications showed higher non-significant elevated ORs for insecticides and herbicides than for fungicides.

    View this table:
    Table 4

    Cumulative exposure to specific subclasses of pesticides and Parkinson disease risk: exposure algorithm

    Exposure to specific active ingredients as assessed by the crop-exposure matrix

    Table 5 shows the results for specific active ingredients as assessed based on self-reported crops, self-reported actual use of pesticides and applying the active ingredient-specific crop-exposure matrix. For the active ingredient benomyl (a benzimidazole fungicide), a positive association with PD was observed for the highest exposed individuals (OR=2.46; 95% CI 1.16 to 5.22), which remained statistically significant after adjustment for potential confounders. Analyses without reclassifying to non-exposed if persons reported that insecticides, herbicides or fungicides had not been applied generally resulted in lower ORs (see online supplementary material table S3).

    View this table:
    Table 5

    Exposure to specific active ingredients and Parkinson disease risk: crop-exposure matrix

    Exposure to endotoxin

    In table 3, the results for cumulative exposure to endotoxin and PD risk based on the DOM-JEM are reported. ORs below unity were observed for the highest tertiles, but no trend with cumulative exposure was observed. Previous pesticide and endotoxin exposure in one model resulted in lower ORs for endotoxin and higher ORs for pesticides (see adjusted model two in tables 3⇑–5).

    Discussion

    We performed a case–control study on PD and used complementary methods to assign pesticide exposure. We used (1) a JEM that accounts for pesticide exposures in all jobs and industries, (2) an exposure algorithm that accounts for exposure during application and re-entry work at farms and (3) a crop-exposure matrix enabling estimation of exposure to specific active ingredients. The comprehensive evaluation revealed no evidence for an association with pesticides and the functional subclasses: insecticides, herbicides and fungicides. However, elevated ORs, which were observed in most analyses for the highest exposure categories, suggest that an overall effect may exist but that the overall limited number of high-exposed cases precluded any statistical significance. Our analyses on specific active ingredients suggest an association with benomyl, a fungicide. In addition, we found no indication for an increased PD risk following occupational exposure to endotoxin.

    A strength of our study was the use of a JEM and a crop-exposure matrix to estimate exposure. The results of our recent meta-analysis showed that studies estimating pesticide exposure based on job titles found higher ORs than studies using self-reported data only.1 An explanation for this might be that study participants are not able to accurately remember past exposures leading to non-differential exposure misclassification and bias towards the null when solely relying on self-reported data.24 We only downgraded the estimated exposures from the JEM and crop-exposure matrix for farm workers and gardeners who informed us that insecticides, herbicides or fungicides were not used on the farm where they had worked. We believe that this approach increased the specificity of the exposure assessment, and this is indirectly supported by the analyses that showed regression to the null if this information was not used. However, we cannot exclude that some recall bias was introduced if cases were less likely than controls to report that pesticides had not been applied at the job.

    In addition, we analysed exposure to pesticides for individuals with farm or gardener jobs using an exposure algorithm to estimate exposure in more detail by relying more on self-reported data on exposure determinants. The exposure algorithm was adapted from an existing algorithm for applicator exposure.18 We added re-entry work to the model, because exposures during re-entry work can be substantial,4 ,5 and the majority (57%) of the participants who had worked at farms where pesticides were applied did not personally perform applications but did potentially perform tasks that included re-entry work. We found ORs in the same range as with the JEM approach, showing the robustness of our results. In addition, we also analysed mixing and application work only so as to keep the analyses comparable to previous work. These analyses showed non-significant increases in ORs for participants who applied pesticides, which seemed to be stronger for exposure to insecticides and herbicides than fungicides. This is in line with our recent meta-analysis that showed increased summary estimates for ever applying insecticides and herbicides but not for fungicides.1 Interestingly, our analyses using a crop-exposure matrix to estimate exposure to specific pesticides found an increased OR for benomyl, which is a fungicide. Exposure to benomyl was only assigned to 1/3 or 1/2 of the participants who were assigned an exposure to fungicides based on the JEM and exposure algorithm, respectively. This points to the necessity of assigning exposure to specific active ingredients as noteworthy observations might be missed when grouped together.

    A limitation of the study was the relatively low participation rate, especially among the oldest participants. The participation rate among cases and controls age 70 years or younger was 66% and 39%, respectively. Sensitivity analyses limited to younger participants resulted in higher ORs than in the overall analyses (data not shown). The finding of associations in a subgroup with a higher participation rate strengthens the evidence for a relation between pesticides and PD. However, this finding might also reflect a better recall of exposures by younger participants compared with older participants, or it could relate to differences in active ingredients used between earlier and later decades.

    A limitation of using hospital controls is that the conditions included in the control group might relate to pesticide or endotoxin exposure or may suffer from referral bias and therefore may have influenced results. However, repeating the analyses leaving out one of the four categories of neurological conditions from the control group at a time resulted in almost identical results, suggesting that the results are not unduly driven by one specific control group.

    No formal validation of the JEMs has been carried out, but the JEMs were based on expertise from experts with a long experience with agricultural-related exposures such as pesticides and endotoxin. A potential source of exposure misclassification in the JEM analyses is that the same exposure is assigned to all participants with the same job code. For the analyses on pesticides, this was partly solved by adjusting for self-reported non-application of pesticides at the farm, but some non-differential misclassification may still exist because differences in performed tasks within similar jobs were not taken into account. Also, the exposure algorithm used to assess pesticide exposure has some limitations. Owing to no specific questions on job tasks being asked, days with re-entry work had to be estimated from farm type only. In addition, weighting factors in the algorithm were based on the AHS and EUROPOEMs, and it is uncertain how well they correspond to the actual exposures in our study. These uncertainties most likely resulted in some non-differential exposure misclassification and likely attenuation of results.

    The crop-exposure matrix analyses were limited in that exposures to active ingredients were assigned based on estimated probability and frequency of use. Especially when probability of use on a crop was low this could have resulted in incorrect assignment of exposure. For this reason, we categorised those exposed into two exposure groups, and based our conclusions on the highest exposure category for which exposure was most certain. Another limitation is that crops in greenhouses and field crops other than potatoes, beets, cereals or maize were not covered by the experts and probabilities of exposure were based on information from Statistics Netherlands. Exposure for those crops had to be extrapolated for time periods before 1995, resulting in higher chance of incorrect estimations and non-differential exposure misclassification. Also, for some potentially interesting active ingredients that were withdrawn from the market before 1995 and that were mainly used on crops not covered by the experts (eg, some organochlorines such as dieldrin and DDT), we could not assign an exposure based on the crop-matrix. Therefore our study is not informative for these active ingredients.

    Our crop-specific analyses add evidence for a possible relation between benomyl and PD. Benomyl is a fungicide from the benzimidazole family that has been used for three decades on a wide range of crops. In 2001, it was banned in large parts of the world including the Netherlands. Benomyl has been investigated in two previous epidemiological studies on PD. A non-significant elevated OR of 1.9 for self-reported use of benomyl was found within a nested case–control study in the AHS.25 A case–control study using registration data of pesticide applications showed a trend for increased risk for ambient exposure to benomyl at occupational addresses but, however, not at residential addresses.26 Besides these epidemiological studies there is also toxicological evidence supporting a possible association between benomyl and PD risk, through a mechanism where benomyl inhibits microtubule assembly thereby stimulating aggregation of α-synuclein, or by inhibition of aldehyde dehydrogenase activity resulting in accumulation of a toxic dopamine metabolite.26 ,27

    Although only active ingredients previously linked to PD were analysed, the observed association for benomyl might be caused by other factors such as other pesticides related to the crops associated with benomyl exposure. Therefore, we conducted separate analyses on those crop groups. In our study, most individuals in the highest exposure category for benomyl had worked for a large part of their career on farms with field crops, mainly potatoes, cereals and/or beets. Analyses on working 10 or more years after 1955 on a farm with those field crops resulted in non-significant elevated ORs for beets (OR=1.84 (95% CI 0.82 to 4.12)) and cereals (OR=1.71 (95% CI 0.80 to 3.66)) but not for potatoes (OR=1.09 (95% CI 0.47 to 2.56)). In addition, among those participants with the highest benomyl exposure a number of participants had worked on farms cultivating strawberries. Ever working on a farm with strawberries showed a non-significant association with PD (OR=2.87 (95% CI 0.87 to 9.44)). Given that we observe increased risks for most of these crops suggests that the observed effect might be benomyl specific or attributable to a pesticide used in combination with benomyl.

    No increase in PD risk was observed after exposure to endotoxin. The fact that more controls than cases had endotoxin exposure and adjusting the pesticide results for endotoxin exposure led to higher ORs also points to a possible protective effect of exposure to endotoxin. As no exposure response relation for endotoxin exposure was observed, this result should be interpreted with caution and the association between endotoxin exposure and PD should be investigated in more detail, for example, using quantitative data on endotoxin exposure.

    Conclusions

    In summary, we studied the relation between exposure to insecticides, herbicides, fungicides and endotoxin and PD in a multicentre case–control study in the Netherlands. The results did not provide evidence for the postulated increase in risk after endotoxin exposure. Also, no evidence for an association with exposure to pesticides was found. However, statistically non-significant elevated ORs observed in the higher exposure categories for pesticides, insecticides, herbicides and fungicides are in line with earlier evidence that exposure to pesticides might increase PD risk. Active ingredient-specific analyses revealed a possible association with benomyl, a benzimidazole fungicide that has previously been associated with PD risk.

    References

    1. ↵
      1. van der Mark M,
      2. Brouwer M,
      3. Kromhout H,
      4. et al
      . Is pesticide use related to Parkinson disease? Some clues to heterogeneity in study results. Environ Health Perspect 2012;120:340–7.
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Moretto A,
      2. Colosio C
      . Biochemical and toxicological evidence of neurological effects of pesticides: the example of Parkinson's disease. Neurotoxicology 2011;32:383–91.
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      1. Freire C,
      2. Koifman S
      . Pesticide exposure and Parkinson's disease: epidemiological evidence of association. Neurotoxicology 2012;33:947–71.
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. de Cock J,
      2. Heederik D,
      3. Kromhout H,
      4. et al
      . Exposure to captan in fruit growing. Am Ind Hyg Assoc J 1998;59:158–65.
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Tielemans E,
      2. Louwerse E,
      3. de Cock J,
      4. et al
      . Exposure to fungicides in fruit growing: re-entry time as a predictor for dermal exposure. Am Ind Hyg Assoc J 1999;60:789–93.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Engel LS,
      2. Seixas NS,
      3. Keifer MC,
      4. et al
      . Validity study of self-reported pesticide exposure among orchardists. J Expo Anal Environ Epidemiol 2001;11:359–68.
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      1. Teschke K,
      2. Olshan AF,
      3. Daniels JL,
      4. et al
      . Occupational exposure assessment in case-control studies: opportunities for improvement. Occup Environ Med 2002;59:575–93.
      OpenUrlAbstract/FREE Full Text
    8. ↵
      1. Baldi I,
      2. Cantagrel A,
      3. Lebailly P,
      4. et al
      . Association between Parkinson's disease and exposure to pesticides in southwestern France. Neuroepidemiology 2003;22:305–10.
      OpenUrlCrossRefPubMedWeb of Science
    9. ↵
      1. Baldi I,
      2. Lebailly P,
      3. Mohammed-Brahim B,
      4. et al
      . Neurodegenerative diseases and exposure to pesticides in the elderly. Am J Epidemiol 2003;157:409–14.
      OpenUrlAbstract/FREE Full Text
    10. ↵
      1. De Palma G,
      2. Mozzoni P,
      3. Mutti A,
      4. et al
      . Case-control study of interactions between genetic and environmental factors in Parkinson's disease. Lancet 1998;352:1986–7.
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Rugbjerg K,
      2. Harris MA,
      3. Shen H,
      4. et al
      . Pesticide exposure and risk of Parkinson's disease—a population-based case-control study evaluating the potential for recall bias. Scand J Work Environ Health 2011;37:427–36.
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Dick FD,
      2. De Palma G,
      3. Ahmadi A,
      4. et al
      . Environmental risk factors for Parkinson's disease and parkinsonism: the Geoparkinson study. Occup Environ Med 2007;64:666–72.
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Manthripragada AD,
      2. Costello S,
      3. Cockburn MG,
      4. et al
      . Paraoxonase 1, agricultural organophosphate exposure, and Parkinson disease. Epidemiology 2010;21:87–94.
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Niehaus I,
      2. Lange JH
      . Endotoxin: is it an environmental factor in the cause of Parkinson's disease? Occup Environ Med 2003;60:378.
      OpenUrlFREE Full Text
    15. ↵
      1. Matheson MC,
      2. Benke G,
      3. Raven J,
      4. et al
      . Biological dust exposure in the workplace is a risk factor for chronic obstructive pulmonary disease. Thorax 2005;60:645–51.
      OpenUrlAbstract/FREE Full Text
    16. ↵
      1. Peters S,
      2. Kromhout H,
      3. Olsson AC,
      4. et al
      . Occupational exposure to organic dust increases lung cancer risk in the general population. Thorax 2012;67:111–16.
      OpenUrlAbstract/FREE Full Text
    17. ↵
      1. van der Mark M,
      2. Nijssen PC,
      3. Vlaanderen J,
      4. et al
      . A case-control study of the protective effect of alcohol, coffee, and cigarette consumption on Parkinson disease risk: time-since-cessation modifies the effect of tobacco smoking. PLoS ONE 2014;9:e95297.
      OpenUrlCrossRefPubMed
    18. ↵
      1. Dosemeci M,
      2. Alavanja MC,
      3. Rowland AS,
      4. et al
      . A quantitative approach for estimating exposure to pesticides in the Agricultural Health Study. Ann Occup Hyg 2002;46:245–60.
      OpenUrlAbstract/FREE Full Text
    19. ↵
      1. van Hemmen JJ
      . EUROPOEM, a predictive occupational exposure database for registration purposes of pesticides. Appl Occup Environ Hyg 2001;16:246–50.
      OpenUrlCrossRefPubMed
    20. ↵
      1. EUROPOEM I
      . Operator exposure. 1996. http://www.ctgb.nl/gewasbescherming/toetsingskader/drr (accessed 17 Dec 2012).
    21. ↵
      1. EUROPOEM II
      . Worker exposure. 2003. http://www.ctgb.nl/gewasbescherming/toetsingskader/drr (accessed 17 Dec 2012).
    22. ↵
      1. Boleij JSM,
      2. Kromhout H,
      3. Fleuren M,
      4. et al
      . Reentry after methomyl application in greenhouses. Appl Occup Environ Hyg 1991;6:672–6.
      OpenUrlCrossRef
    23. ↵
      Statistics Netherlands 1979–1995. Use of pesticides by Dutch government institutions. The Netherlands.
    24. ↵
      1. Daniels JL,
      2. Olshan AF,
      3. Teschke K,
      4. et al
      . Comparison of assessment methods for pesticide exposure in a case-control interview study. Am J Epidemiol 2001;153:1227–32.
      OpenUrlAbstract/FREE Full Text
    25. ↵
      1. Tanner CM,
      2. Kamel F,
      3. Ross GW,
      4. et al
      . Rotenone, paraquat, and Parkinson's disease. Environ Health Perspect 2011;119:866–72.
      OpenUrlCrossRefPubMedWeb of Science
    26. ↵
      1. Fitzmaurice AG,
      2. Rhodes SL,
      3. Lulla A,
      4. et al
      . Aldehyde dehydrogenase inhibition as a pathogenic mechanism in Parkinson disease. Proc Natl Acad Sci USA 2013;110:636–41.
      OpenUrlAbstract/FREE Full Text
    27. ↵
      1. Kim M,
      2. Jung W,
      3. Lee IH,
      4. et al
      . Impairment of microtubule system increases alpha-synuclein aggregation and toxicity. Biochem Biophys Res Commun 2008;365:628–35.
      OpenUrlCrossRefPubMed

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      Files in this Data Supplement:

    Footnotes

    • Contributors PCGN and HK initiated the study. All authors were involved in the design of the study. MvdM, RV, AH and HK were involved in the statistical analyses and interpretation of the data. The manuscript was drafted by MvdM and was revised with contributions from all authors.

    • Funding This work was supported by Stichting Internationaal Parkinson Fonds (The Netherlands); and The Netherlands Organization for Health Research (ZonMW) within the programme Electromagnetic Fields and Health Research under grant number 85800001.

    • Competing interests None.

    • Patient consent Obtained.

    • Ethics approval Medical Ethics Committee of St Elisabeth Hospital, Tilburg, The Netherlands.

    • Provenance and peer review Not commissioned; externally peer reviewed.