Article Text
PDF
Abstract
Objective Diazinon, a common organophosphate insecticide with genotoxic properties, was previously associated with lung cancer in the Agricultural Health Study (AHS) cohort, but few other epidemiological studies have examined diazinon-associated cancer risk. We used updated diazinon exposure and cancer incidence information to evaluate solid tumour risk in the AHS.
Methods Male pesticide applicators in Iowa and North Carolina reported lifetime diazinon use at enrolment (1993–1997) and follow-up (1998–2005); cancer incidence was assessed through 2010(North Carolina)/2011(Iowa). Among applicators with usage information sufficient to evaluate exposure-response patterns, we used Poisson regression to estimate adjusted rate ratios (RRs) and 95% CI for cancer sites with ≥10 exposed cases for both lifetime (LT) exposure days and intensity-weighted (IW) lifetime exposure days (accounting for factors impacting exposure).
Results We observed elevated lung cancer risks (N=283) among applicators with the greatest number of LT (RR=1.60; 95% CI 1.11 to 2.31; Ptrend=0.02) and IW days of diazinon use (RR=1.41; 95% CI 0.98 to 2.04; Ptrend=0.08). Kidney cancer (N=94) risks were non-significantly elevated (RRLT days=1.77; 95% CI 0.90 to 3.51; Ptrend=0.09; RRIW days 1.37; 95% CI 0.64 to 2.92; Ptrend=0.50), as were risks for aggressive prostate cancer (N=656).
Conclusions Our updated evaluation of diazinon provides additional evidence of an association with lung cancer risk. Newly identified links to kidney cancer and associations with aggressive prostate cancer require further evaluation.
- diazinon
- insecticides
- organophosphate
- neoplasms
Statistics from Altmetric.com
What this paper adds
This comprehensive evaluation of diazinon and solid tumours in a prospective cohort study included exposure information from two points in time, 337 771 person-years, and 2288 incident cases of solid tumours.
This re-evaluation within the AHS cohort demonstrated that lung cancer risks persist with updated exposure information, 199 additional cases, and 8 more years of follow-up. In this first evaluation of histologic subtypes, there is suggestive evidence for an association with adenocarcinoma.
In this first study to examine kidney cancer risk associated with diazinon exposure, there is suggestive evidence of an association that needs to be confirmed in further studies.
Aggressive prostate cancer was also associated with diazinon exposure.
Introduction
Diazinon (O,O-diethyl O-(2-isopropyl-6-methyl-4-pyrimidinyl) phosphorothioate) is a broad-spectrum organophosphate insecticide first registered for agricultural and residential use in the mid-1950s. It has been available in several physical states and formulations, including as liquid, granules, dust, wettable powders and impregnated materials. Residential lawn and garden use in the USA was phased out by 2004,1 but diazinon still ranks among the top 10 most commonly used active organophosphate insecticides2 and is used agriculturally to control soil and foliage insects on crops and non-lactating livestock. In 2007, crop-specific restrictions were added and granular formulations of diazinon were completely banned. Additional limitations in certain parts of the USA aim to protect endangered species from diazinon exposure due to agricultural runoff and drift.1
Diazinon can be present in the ambient air as a vapour or particulate following application and end up in drinking water due to runoff from agricultural fields. Therefore, human environmental or occupational exposure is possible through ingestion, inhalation or dermal routes.3 Transformation of diazinon to diazoxon, a potent cholinesterase inhibitor, can occur in multiple environmental and biological media.3 While it has not been classified for carcinogenicity by the International Agency for Research on Cancer (IARC) or the US Environmental Protection Agency (EPA),4 ,5 renal, liver and pancreatic toxicity have been identified in animal models.3
Epidemiological evidence of an association between diazinon and lung cancer was observed in previous analyses of pesticide applicators in the Agricultural Health Study (AHS), a large prospective cohort of applicators and their spouses,6 ,7 and in a case–control study nested within a cohort of pest control workers.8 There is some evidence for an association with other solid cancers, including soft tissue sarcoma9 and prostate cancer.10 ,11 Diazinon has also been linked to risks of leukaemia and follicular lymphoma in the AHS,7 ,12 and with non-Hodgkin's lymphoma.13–15 However, most of this evidence comes from case–control studies with relatively few exposed cases or limited exposure information. These findings, as well as changes in the EPA registration status of diazinon, underscore the need to assess exposure at more than one time point to adequately capture changes in usage and subsequent impacts on cancer risks. Moreover, risks for other major solid tumour sites such as bladder and kidney have not been evaluated. The current study investigated the putative associations between diazinon and solid tumours in the AHS with extended follow-up and enhanced exposure information from a follow-up survey.
Methods
Study population
Details of the study design and cohort composition of the AHS are described elsewhere.7 ,16 Initiated in 1993, the AHS is an ongoing prospective cohort of 52 394 licensed private pesticide applicators (primarily farmers) residing in Iowa and North Carolina, 32 345 of their spouses, and 4916 licensed commercial pesticide applicators in Iowa. We identified applicators at the time of their application (1993–1997) for a restricted-use pesticide license at licensing facilities, and asked them to complete an enrolment questionnaire eliciting information on use of pesticides, medical history, smoking and demographic information. We asked all applicators to complete a more detailed ‘take-home’ questionnaire at enrolment, where they provided lifetime use of specific pesticides, including diazinon. Approximately 44% (N=25 291) returned the take-home survey; characteristics of these applicators were not systematically different from those completing the enrolment questionnaire only.7 ,17 We invited all applicators to complete a follow-up telephone survey 5 years later (1998–2005), where they indicated their pesticide use since enrolment. Spouses did not report detailed pesticide usage at both time points, and were not included in this evaluation.
Exposure assessment
At enrolment, we obtained lifetime use of diazinon for the 25 291 applicators completing the take-home survey, and updated use information during the telephone follow-up interview. We used multiple imputation to estimate pesticide exposures for applicators who did not participate in follow-up; the methodology and its validation have been reported previously.18 Briefly, we used logistic regression and stratified sampling to impute pesticide exposure from a series of predictors, including but not limited to demographic and farm characteristics, specific medical conditions at enrolment and characteristics of pesticide usage. For the current analysis, we imputed diazinon exposures for the period between enrolment and follow-up for the subset of applicators (28%) with missing follow-up data.
We created exposure metrics reflecting lifetime diazinon use and intensity. Lifetime (LT) exposure days were the number of application days per year multiplied by the number of years of application. Intensity-weighted (IW) lifetime exposure days further accounted for factors impacting exposure, including application method, whether or not the applicator personally mixed the pesticides, repaired pesticide application equipment or used personal protective equipment.19
The decade of first use of diazinon was reported in categories at enrolment (eg, 1960s, 1970s). We estimated this decade for the 1% of applicators missing this information by subtracting the midpoint of the categorically reported years of diazinon use at enrolment (≤1, 2–5, 6–10 11–20, or 20+ years) from the enrolment year. Although detailed information about diazinon applications was not collected at enrolment, at follow-up applicators reported the physical state (eg, dry/granular, liquid) of the most frequently applied pesticides and type of application (eg, crop, non-crop, animals).
Case ascertainment
We assessed cancer incidence by linkage to North Carolina and Iowa state cancer registries from enrolment through 31 December 2010 for North Carolina and 31 December 2011 for Iowa, and determined vital status by matching to the National Death Index. Follow-up time was censored at the date of any cancer diagnosis, date of death, migration out of state or date of last follow-up, whichever was earlier.
This assessment added 199 incident lung cancers and 780 other solid tumours identified since the last analysis, which included cases diagnosed through 2002 and did not include bladder or kidney cancers.7 We classified prostate cancer as aggressive by tumour characteristics as was described previously,11 including distant stage, poorly differentiated, Gleason score ≥7, or prostate cancer as the underlying cause of death.
Statistical analysis
Among 25 291 enrolled applicators who completed the take-home questionnaire, we excluded individuals with prevalent cancer at baseline (N=622) or who were missing follow-up information (N=145). We further restricted the analysis to males (N=23 861) due to the small number of female applicators (N=663; 188 of whom reported using diazinon at follow-up). Our final analysis subset included the 22 830 male applicators with complete information for LT exposure days (reported or imputed). Because the population exposure distribution at follow-up did not vary substantially from enrolment, we retained the exposure categories from the most recent AHS diazinon analysis7 and evaluated cancer risks for applicators within tertiles of LT and IW exposure days of diazinon use compared to those with no use. We evaluated risks of incident solid tumours for which there were at least 10 diazinon-exposed cases. We estimated lung cancer risks by histological subtype (adenocarcinoma, squamous cell carcinoma, small cell carcinoma and other carcinomas). We also examined cumulative diazinon usage by the decade of first use, and computed the proportions of reported diazinon applications at follow-up by the physical state of product applied and application type.
We conducted Poisson regression to estimate rate ratios (RRs) for solid tumour sites in relation to diazinon with the SAS (V.9.2) procedure MIANALYZE, which yielded a single RR for each cancer site and a 95% CI reflecting the average variance across separate models of five imputed exposures. We adjusted models for age at baseline (<40, 40–49, 50–59, ≥60 years), smoking history (never, tertiles of pack-years among former smokers: <3.75, 3.75–15, >15, tertiles of pack-years among current smokers: <11.5, 11.5–28.4, >28.5), education (high school or less, greater than high school), family history of cancer and state (Iowa, North Carolina), including missing covariates as categories (all covariates included in final models had ≤7% missing). There were several exceptions for covariate inclusion; we excluded alcohol or education variables and included only the categorical smoking adjustment without pack-years for kidney cancer and subtype-stratified models due to sparse data, and additionally adjusted prostate cancer models for race (Caucasian, non-Caucasian). We evaluated additional potential confounders collected at enrolment, including those linked with cancer risk in other AHS evaluations (eg, doctor diagnosis of allergy, farm animal exposures, diesel tractor use and solvent use in farming activities such as equipment cleaning and pesticide mixing). There were no covariates parameterised as time-varying variables (ie, cross-product terms with the underlying time variable) in our regression models. However, we did parameterise our exposure variables to reflect changes in diazinon use in 2-year increments between enrolment and follow-up. We assessed interactions with binary coexposures via cross product terms with continuous variables of diazinon LT and IW exposure days and Wald tests, and tested for linear trends using the median LT or IW exposure days within categories parameterised as continuous variables.
We conducted additional analyses to explore the consistency with prior assessments of cancer risk in the AHS, timing of use, and disease latency. We conducted Spearman rank correlation analyses to identify other pesticide usage correlated with diazinon exposure, and estimated all risks further adjusted for lifetime use of pesticides previously associated with lung cancer in the AHS, including chlorpyrifos, dicamba, dieldrin, metolachlor, pendimethalin,20 carbofuran21 and terbufos,22 and for the top five of other pesticides for which cumulative lifetime usage was correlated with diazinon. To address concerns regarding exposure recency, we repeated analyses after restricting to cases diagnosed in the first 10 years after enrolment and lagged exposure in 5-year and 15-year intervals. To address concerns about imputed exposures, we also repeated main analyses after excluding applicators with missing follow-up information. These models could be adjusted only for age, smoking and state. All data used were from AHS data release versions P1REL201209.00 and P2REL201209.00.
Results
Of the 22 830 male applicators, 5120 used diazinon (table 1), 2% of whom used it for the first time after enrolment (data not shown). Non-exposed applicators were more likely to raise livestock and less likely to raise poultry, and applicators with the highest diazinon usage had nearly twice the average LT exposure days of total pesticide exposures. Private applicators in North Carolina were more likely to have used diazinon and had greater cumulative LT exposure than those in Iowa (mean=25 and 4 LT exposure days, respectively; online supplemental table S2). Commercial applicators had greater lifetime diazinon use overall, including more days per year of use (mean=14 days vs 6 among private applicators; data not shown).
Selected baseline characteristics of male pesticide applicators in the Agricultural Health Study, 1993–2010/2011*, by cumulative lifetime diazinon exposure (N=22 830)
Based on information collected in the follow-up interview, diazinon was applied predominantly in dry form, particularly among private applicators (70.3% of applications, table 2). Private applicators applied diazinon for non-crop uses on the farm (59%), whereas commercial applicators applied it to crops (76%). Applications to animals were less frequent (1–3%). While relative applications to crops and animals did not vary greatly between Iowa and North Carolina, the physical state of the product differed. Private applicators in Iowa applied dry or granular diazinon, whereas the North Carolina private applicators reported usage more comparable to the commercial applicators in Iowa (33.5% and 39.7% applying dry diazinon, respectively). No clear patterns in decade of first diazinon use were evident, except that most usage began in the 1970s and 1980s (online supplemental table S1); decade of first use was uncorrelated with age (data not shown).
Diazinon applications* reported by male pesticide applicators at follow-up in the Agricultural Health Study, 1993–2010/2011
A total of 2288 incident solid tumours were diagnosed through 2010 (North Carolina) and 2011 (Iowa), including 526 cases among those who used diazinon. In multivariable analyses, we observed a significant exposure-response trend for lung cancer with increasing LT exposure days (RR>38.8 days vs non-exposed=1.60; 95% CI 1.11 to 2.31; Ptrend=0.02; table 3). The pattern was similar for IW exposure days (RR=1.41; 95% CI 0.98 to 2.04; Ptrend=0.08). The RR increased with cumulative LT use when we split the top tertile of LT exposure days at the median, with RR in 38.90–108.8 LT exposure days=1.41; 95% CI 0.88 to 2.27 and RR>108.8 LT exposure days=1.80, 95% CI 1.09 to 2.97; Ptrend=0.01. Corresponding risks for IW exposure days were RR=1.39; 95% CI 0.84 to 2.31 and RR=1.39; 95% CI 0.86 to 2.24; Ptrend=0.15, respectively. The RR of lung cancer was elevated among never smokers for both < and ≥ median LT or IW exposure days compared with the non-exposed, but no gradients were present, and RRs were based on small numbers of exposed cases (online supplemental table S3).
Rate ratios (RRs) for selected solid malignancies in association with cumulative lifetime diazinon exposure among male pesticide applicators in the Agricultural Health Study, 1993–2010/2011 (N=22 830)
Risks for kidney cancer were also elevated in the top exposure tertiles compared with the non-exposed (RRs 1.77 and 1.37 for LT and IW exposure days, respectively), although not statistically significantly (table 3). When we split the top exposure tertiles at their medians, the RR continued to increase (RR in 38.90–108.8 LT exposure days=1.58; 95% CI 0.63 to 3.97 and RR among those with >108.8 LT exposure days=2.56; 95% CI 1.01 to 6.47; Ptrend=0.02). Corresponding RRs for a split of the top IW tertile were RR=0.73; 95% CI 0.18 to 3.00 and RR=2.40; 95% CI 1.01 to 5.73; Ptrend=0.05); most of the cases in the top tertile of IW exposure (6 of 9) were above the median of the exposure level. When we restricted to only the renal cell carcinomas (∼94% of cases) the association in the top tertile of diazinon exposure remained (RR=1.81; 95% CI 0.91 to 3.6). We saw no association for prostate cancer overall, but observed non-significantly increased risks for aggressive prostate cancer in the top tertile of LT (RR=1.16, 95% CI 0.83 to 1.63 Ptrend=0.44) and IW exposure days (RR=1.29; 95% CI 0.93 to 1.79; Ptrend=0.22). Splitting the top tertile at the median did not identify further increasing risks for LT exposure days (RR in 38.90–108.8 LT exposure days=1.14, 95% CI 0.76 to 1.73 and RR>108.8 days=1.11, 95% CI 0.65 to 1.90; Ptrend=0.48). However, a suggestion of an association remained evident for IW exposure days (RR=1.21, 95% CI 0.78 to 1.87 and RR=1.29, 95% CI 0.82 to 2.02; Ptrend=0.17). The large number of prostate cancer cases allowed us to categorise exposure into quartiles, which yielded no association with LT exposure days (RR in Q4 vs non-exposed=1.14, 95% CI 0.79 to 1.64; Ptrend=0.49) and a non-statistically significant positive association with IW exposure days (RR in Q4 vs non-exposed=1.39, 95% CI 0.97 to 2.01; Ptrend=0.11). Risks for other cancer sites were generally null (table 3). No differences in any of these observed associations emerged from analyses of lagged exposures (data not shown).
We found subtle differences in the association between diazinon exposure and lung cancer by histological subtype (table 4). There was no apparent association with squamous cell carcinoma, the most common subtype. We observed a significant exposure response for adenocarcinoma and LT exposure days (Ptrend=0.02) but not for IW exposure (Ptrend=0.14). We observed no statistical interactions between diazinon and exposure to solvents, animals, diesel tractor use and self-reported diagnosis of allergy for any cancer site (data not shown). Lifetime exposures to pesticides previously associated with lung cancer risk in the AHS were uncorrelated with diazinon use (ranged from ρ=0.21 to 0.50, the strongest correlation with dieldrin). We found no substantive evidence that the five additional most strongly correlated pesticide exposures (from among those with available usage information) were confounders in our key analyses (online supplementary table S5). Although based on 49 lung and 14 exposed kidney cancer cases and imprecise, when we restricted analyses to the 72% of applicators participating in follow-up, we observed similarly elevated risks for lung, kidney, and aggressive prostate cancers in the top exposure categories for both metrics. When we restricted analyses to cases diagnosed in the first 10 years post-enrolment or to commercial applicators only, we also observed consistent patterns of association for these sites but reduced precision (data not shown).
Lung cancer rate ratios (RRs)* in association with cumulative lifetime diazinon exposure days among male pesticide applicators in the Agricultural Health Study, 1993–2010/2011, by histological subtype
Discussion
In this comprehensive analysis and largest study of diazinon and lung cancer to date, we observed a significantly increased risk of lung cancer among male pesticide applicators reporting over 38 LT exposure days of diazinon exposure, with significant exposure response. This analysis included an additional 199 cases since the most recent AHS assessment of lung cancer and diazinon,7 allowing for analyses by histological subtype and a more robust evaluation of potential confounding and effect modification by smoking and other lung cancer risk factors. It also added 8 years of follow-up for North Carolina participants and 9 for Iowa participants, and further supports a lung cancer–diazinon association. These results are consistent with the only other study to evaluate this association, a case–control study of pesticide workers, which found suggestive associations between diazinon exposure and lung cancer mortality.8 It also suggests possible links with kidney and aggressive prostate cancers. There was no compelling evidence of diazinon-associated risk for other solid tumour sites. For a chemical still commonly used in agriculture with potential to result in environmental exposure, these findings corroborate those previously observed in epidemiological studies as well as offer new information about associations with cancer development.
Animal studies have noted DNA methylation by diazinon, which could cause toxicity and carcinogenic action.23 Diazinon has been linked with renal dysfunction in rats,24 liver and kidney tissue damage in mice,25 and DNA damage in the liver and kidneys of rabbits.26 Non-mutagenic histological changes in lung tissue27 and in airway defence mechanisms28 have also been noted in animal models. Dose-related genotoxic effects of diazinon have been observed in human nasal mucosal cells in vitro.29 One proposed mechanism of altered gene expression resulting from diazinon exposure is impaired DNA excision repair activity by downregulation of the RNRM1 gene, which plays an important role in upregulation of the PTEN tumour suppressor.30 Excision repair protein levels have been found to be lower in patients with lung cancer compared with controls,31 ,32 which may indicate increased cellular mutations and transformation as a consequence of deficient excision repair processes. Oxidative stress may be another mechanism of cytotoxicity,33 as has been demonstrated in human lymphocytes following diazinon exposure.34
We explored whether additional information about application as a liquid or solid could provide additional insight into the observed association with lung cancer. Usage during the follow-up period differed between North Carolina and Iowa; diazinon was largely sprayed as a liquid in North Carolina, while dry applications were more common in Iowa. Although both application modes can result in exposure by inhalation or ingestion, liquid applications could potentially lead to more frequent dermal exposures, which are weighted more heavily in our IW metric19 and may be less relevant for lung cancer aetiology. We also compared application patterns between states and saw no major differences as to whether diazinon was applied to crops or animals. Diazinon historically came in a variety of preparations1 and is used on crops as well as animals, but because details for applications were not collected at enrolment, we could not assess whether these exposure patterns reflect usage during earlier time periods. Finally, we note that the AHS cohort is comprised primarily of private applicators, although the limited information we have indicates that the frequency of commercial use differs from private use on farms. We were constrained by sample sizes to further evaluate these interesting descriptive features of diazinon application in relation to cancer risk. Our observed association with kidney cancer is based on a relatively small number of exposed cases (n=21) and is the first such evaluation, to the best of our knowledge, in a human population. However, the animal literature provides some biological plausibility, including renal dysfunction and dose-depdendent oxidative stress and genotoxic effects.24 ,25 ,26 The kidneys are a site of concentration of xenobiotic compounds, and may be sensitive to chemical insults.35 Thus, although the observed kidney cancer excess may be a chance finding, the association has some biological support and should be further evaluated. We were unable to conduct stratified analyses of kidney malignancies by histological subtype to further explore this finding, but associations in the top tertile of diazinon exposure remained when we restricted to the predominant subtype, renal cell carcinoma.
In a previous evaluation of prostate cancer and pesticides in the AHS, Koutros et al11 did not find a consistent relationship between diazinon use and risk of prostate cancer overall, although there was a suggestion of an increased risk of aggressive cancer in the highest category of IW exposure days (RR=1.31, 95% CI 0.87 to 1.96). Our analysis differs in two important ways. First, our evaluation provides additional cases and follow-up time. Second, because our analysis is focused specifically on the diazinon exposure response, we restricted to applicators who completed the take-home enrolment questionnaire. However, our results were similar in that we observed a non-significantly increased risk of aggressive prostate cancer among applicators in the highest categories of diazinon use. The association with IW exposure days remained when we split the top tertile at its median and in quartile analyses, which are more directly comparable to the analyses of Koutros et al. A case–control study among Canadian farmers also found an association between diazinon use and prostate cancer (OR=1.43; 95% CI 0.99 to 2.07) that was stronger and statistically significant in the highest category of exposure,10 although these tumours were not characterised as aggressive. We observed no risks for melanoma or for bladder, colon or rectum cancers. While animal studies indicate potential liver and pancreatic carcinogenicity,3 we had too few diazinon-exposed cases (zero and four, respectively) for meaningful analysis of these sites.
The strengths of this study include the prospective design and detailed pesticide use assessed at two time points, which allowed us to retain our original analytic subset for nearly 15 years of follow-up. We also had detailed information on smoking history. Analyses of lung cancer restricted to never-smokers indicated positive associations with diazinon, and stratified analyses showed a positive and significant trend for adenocarcinoma, a subtype less strongly linked to smoking compared to squamous and small cell carcinomas of the lung.36 For these reasons, residual confounding by smoking is not a likely explanation for the observed associations. We were also able to adjust for other known or potential confounders and to assess effect modification by important coexposures. We found no evidence of confounding by use of other pesticides, including chlorpyrifos, another phosphorothioate insecticide previously associated with lung cancer in the AHS cohort.6 Sensitivity analyses restricting to those with complete exposure data agreed with our main findings.
We noted some interesting patterns in application preferences and the timing of first use of diazinon, but were limited by sample sizes and lacked sufficient latency to fully explore whether trends in usage by state or time period were aetiologically relevant. Notably, the majority of applicators first began using diazinon in the 1970s and 1980s, roughly 10–20 years prior to the assessment of diazinon exposure at enrolment. Therefore, in addition to the average 15 years of prospective follow-up in this analysis, considerable time since first exposure has elapsed for most applicators.
With additional follow-up, cases and exposure information, results from this prospective cohort study continue to provide evidence of an association between diazinon and lung cancer risk. Furthermore, our update allowed us to evaluate several tumour sites previously unassessed in relation to diazinon exposures in the AHS. A potential association with aggressive prostate cancer and one newly identified with kidney cancer require further evaluation.
References
Footnotes
Contributors RRJ participated in the study design, carried out the primary data analysis, and prepared the first and subsequent drafts of the manuscript. FB-A was involved in the original study design and preliminary analyses and participated in the manuscript preparation. SK and CCL helped to coordinate the data collection, provided advice for prostate cancer analyses and contributed to the manuscript. SLH led the data imputation and contributed to the manuscript. AB, JL, JAH and MCRA were involved in the original study design and participated in the manuscript preparation. LEBF participated in the original study design, data collection, coordination of analyses and helped to prepare drafts of the manuscript. All authors read and approved the final manuscript.
Funding This research was supported by the intramural research programme of the National Institutes of Health, National Cancer Institute, Division of Cancer Epidemiology and Genetics (Z01CP010119).
Competing interests None.
Patient consent Obtained.
Ethics approval Institutional Review Board of the National Cancer Institute.
Provenance and peer review Not commissioned; externally peer reviewed.