Thyroid Cancer and Pesticide Use in a Central California Agricultural Area: A Case Control Study

Abstract

Objective

To examine environmental factors that influence risk of thyroid cancer.

Methods

We performed a case-control study utilizing thyroid cancer cases from the California Cancer Registry (1999-2012) and controls sampled in a population-based manner. Study participants were included if they were diagnosed with thyroid cancer, lived in the study area at their time of diagnosis, and were ≥35 years of age. Controls were recruited from the same area and eligible to participate if they were ≥35 years of age and had been living in California for at least 5 years prior to the interview. We examined residential exposure to 29 agricultural use pesticides, known to cause DNA damage in vitro or are known endocrine disruptors. We employed a validated geographic information system–based system to generate exposure estimates for each participant.

Results

Our sample included 2067 cases and 1003 controls. In single pollutant models and within a 20-year exposure period, 10 out of 29 selected pesticides were associated with thyroid cancer, including several of the most applied pesticides in the United States such as paraquat dichloride [odds ratio (OR): 1.46 (95% CI: 1.23, 1.73)], glyphosate [OR: 1.33 (95% CI: 1.12, 1.58)], and oxyfluorfen [OR: 1.21 (95% CI: 1.02, 1.43)]. Risk of thyroid cancer increased proportionately to the total number of pesticides subjects were exposed to 20 years before diagnosis or interview. In all models, paraquat dichloride was associated with thyroid cancer.

Conclusions

Our study provides first evidence in support of the hypothesis that residential pesticide exposure from agricultural applications is associated with an increased risk of thyroid cancer.

The incidence of thyroid cancer has risen substantially in the United States in the past 30 years, with an average increase of 3% annually (1, 2). This has been partly attributed to incidentally discovered thyroid nodules (1, 3). However, recent reports have questioned this explanation, suggesting environmental, dietary, and genetic risk factors may also explain the increasing incidence (4). Rising incidence of obesity (5) and changing racial/ethnic population distributions (6) have also been implicated in the increasing trend (7). Except for radiation, few studies have addressed environmental exposures’ effect on thyroid cancer occurrence. Some studies found higher risks in the leather, wood, and paper industries and with exposure to certain environmental solvents, flame retardants, and pesticides (8-10).

Certain pesticides are established mutagens or induce tumor growth and chromosomal aberrations in vitro (11, 12). These include glyphosate, the active ingredient in widely used herbicide products, which interferes with estrogen signaling pathways, and 19 pesticides that induce DNA double-strand breakage in thyroid cells in vitro (11, 13). Pesticides can also alter thyroid hormone production, which has been related to thyroid cancer risk (14). Human studies of pesticides and thyroid cancer have been inconsistent (8, 15-18) and had methodological limitations including self-reported exposures, no information on specific pesticides, or small sample sizes.

California ranks first among US states in agricultural production, with 26.2 million acres devoted to farming and ranching. In 2008, agricultural pesticide uses in California totaled 162 million pounds, accounting for one-fourth of all US use (19). Additionally, increased rates of advanced thyroid cancer in California have been documented (20). The purpose of this study is to examine the association between exposure to select pesticides, including 19 that were found to cause DNA cell damage in vitro, and risk of thyroid cancer. We therefore hypothesized that pesticide exposure may be a missing link in need of investigation.

Methods

All procedures described were approved by the University of California, Los Angeles, institutional review board and the State of California Committee for the Protection of Human Subjects.

Subject Selection

Primary thyroid cancer cases of all subtypes, diagnosed from 1999 to 2012, who lived in the central California counties of Fresno, Kern, or Tulare at their time of diagnosis were identified from California’s comprehensive mandated statewide population-based cancer surveillance system, the California Cancer Registry (n = 2578). We excluded anyone <35 years of age (n = 511) as we did not have population controls available for this age group, leaving 2067 for analysis.

Our control population was originally recruited for the Parkinson’s Disease Environment and Gene (PEG) case control study which enrolled controls between 2001 and 2011 in the same counties. Details of our recruitment methods have been described previously (21-23). Briefly, controls were eligible to participate if they (1) did not have Parkinson’s disease, (2) were ≥35 years of age, (3) were currently residing in the designated counties, and (4) had been living in California for at least 5 years prior to the interview. PEG controls with a self-reported history of thyroid cancer were excluded.

Originally, controls aged ≥65 years were to be identified from Medicare lists while younger controls were randomly selected from tax assessor residential units (parcels). However, due to the Health Insurance Portability and Accountability Act prohibiting Medicare from sharing this data after 2001, the entire control population was recruited using tax assessor parcels. We randomly selected residential living units and mailed letters of invitation. Marketing companies and internet searches were utilized to identify head-of-household names and telephone numbers. A total of 2085 eligible controls were contacted, and 1022 participants (49%) enrolled. Out of the total 1022 controls, 812 provided complete residential address histories, 191 completed a short interview providing us with their last 3 residential addresses only. Nineteen had missing/incomplete residential address information and were excluded from analyses. Therefore, 1003 controls are included in the present analyses.

Exposure Assessment

As described elsewhere (22, 24, 25), residential pesticide exposures were estimated with a validated geographic information system (GIS)-based system that combines California state-mandated Pesticide Use Reporting (PUR) data from commercial agricultural applications, land-use surveys providing locations of specific crops, and geocoded residential addresses for each participant. The California Department of Pesticide Regulation has recorded PUR data for any commercial application of restricted-use pesticides (agents considered harmful for humans or the environment) (26) since 1974 and, in 1990, of all pesticide use. PUR records provide information on pesticide active ingredients, the poundage applied, the crop and acreage, the application method and date, and the location according to a Public Land Survey System grid. We used land-use maps to more precisely locate applications (24, 25), and we estimated pounds applied per year within a 500-m radius buffer around subjects’ homes for each pesticide. A 500-m radius has been used previously (27-29), and dispersion studies suggested that 500-m buffers may allow exposures to be measured most accurately (30).

We included pesticides in analyses that cause DNA damage/breakage in a human thyroid cell line using gamma-H2AX immunofluorescence (13) and those that have been linked to endocrine disruption in previous studies (14, 31-36). We generated cumulative residential pesticide exposure estimates for the entire 5-, 10-, and 20-year period before diagnosis/interview based on address histories in controls and the diagnosis address only in cases. As cases were ascertained from the California Cancer Registry, only the address at the time of diagnosis was available for thyroid cancer cases. Older adults move less often (37), and colleagues have found that residential moving behaviors among adults aged ≥65 in our control population were stable in the 26-year period before interview (unpublished data). Although exposure estimates from earlier (5- or 10-year) time periods may best reflect recent residential addresses, most cancers have a longer induction period, and we therefore report estimates from our 20-year model in the main tables, with 5- and 10-year results in Supplementary Table 2 (38). Individuals were considered exposed if they had any ambient exposure to our pesticide of interest within a 500-m radius of their home, irrespective of other pesticide co-exposures.

Statistical Analyses

In multinomial logistic regression models, we controlled for sex, age (35-40, 41-50, 51-60, 61-70 and >70 years), race/ethnicity, and socioeconomic status (SES) at the census block group level, described in the following discussion. Selection of variables for adjustment was based upon previous literature (6, 31) as well as our exploration of associations in our data. We attempted adjustment for smoking status, education, and income; however, effect estimates did not change by >10%, and thus these variables were excluded from the final model. Racial/ethnic group identification was available from the California Cancer Registry for cases, while race/ethnicity was self-identified by controls (non-Hispanic, White; non-Hispanic, Black; Hispanic; Asian/Pacific Islander; non-Hispanic American Indian; other/unknown). SES was coded based on the Yost composite scoring system and presented as quintiles of block group SES from principal components analysis that combined seven indicator variables (education, proportion in blue collar jobs, proportion unemployed among those >16, median household income, proportion below 200% of the federal poverty level, median rent, and median house value) (39).

We calculated cumulative yearly pesticide exposures during the 20-year exposure period to explore dose-response; in other words, we examined whether exposure to a greater number of unique pesticides in the 20-year window resulted in a higher risk of thyroid cancer among study participants. To investigate exposure misclassification due to unknown residential mobility in cases when using a long (20-year) exposure period, we also conducted sensitivity analyses restricting to those ≥65 years of age. We furthermore stratified the data by sex (thyroid cancer is more common in women and more aggressive in men) and race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Asian/Pacific Islander, non-Hispanic American Indian, other/unknown) to assess effects on advanced (regional/metastatic) disease (6). To understand better how exposure combinations influence risk, we examined highly correlated pesticides in separate and combined analyses.

Data analyses were conducted using SAS 9.3 (SAS Institute Inc., Cary, NC, USA). The level of significance used was α = 0.05, and all tests were 2-sided.

Results

A greater proportion of cases compared with controls were female (76% vs 54%) and ≤50 years of age (44% vs 10%) (Table 1). Race/ethnicity and SES were similar for cases and controls.

Table 1.

Demographic distribution of thyroid cancer cases and Parkinson’s Disease Environment and Gene/Center for Gene Environment Studies in Parkinson’s Disease controls

	Controls (n = 1003)	Cases (n = 2067)
Sex
Male	462 (46)	504 (24)
Female	541 (54)	1563 (76)
Age, years
35-40	32 (3)	309 (15)
41-50	73 (7)	603 (29)
51-60	227 (23)	527 (26)
61-70	304 (30)	348 (17)
>70	365 (37)	280 (14)
Missing	2 (0)	0
Race/ethnicity
Non-Hispanic, White	654 (65)	1238 (60)
Non-Hispanic, Black	36 (4)	53 (3)
Hispanic	231 (23)	608 (29)
Asian/Pacific Islander	34 (3)	136 (7)
Non-Hispanic American Indian	45 (5)	17 (1)
Other/Unknown	3 (0)	15 (1)
Neighborhood SES
Lowest SES	279 (28)	664 (32)
Lower-middle SES	262 (26)	468 (23)
Middle SES	204 (20)	418 (20)
Higher-middle SES	224 (22)	412 (20)
Highest SES	34 (3)	105 (5)

Data are given as n (%).

Abbreviation: SES, socioeconomic status.

Odds ratios (OR) for thyroid cancer associated with a 20-year exposure window for each pesticide in single pollutant models are presented in Table 2. Nearly all point estimates were elevated, and we found 10 pesticides to be positively associated with thyroid cancer [including captan OR: 1.24 (95% CI: 1.00, 1.52); propargite OR: 1.27 (95% CI: 1.07, 1.51); glyphosate OR: 1.33 (95% CI: 1.12, 1.58); mancozeb OR: 1.26 (95% CI: 1.01, 1.56); and permethrin OR: 1.26 (95% CI: 1.01, 1.57)]. However, many pesticide exposures were correlated [Supplementary Table 1 (38)]. Paraquat dichloride, which we herein refer to as paraquat, was the only chemical consistently associated with risk of thyroid cancer in both single and multipollutant models [OR in single pollutant model: 1.46 (95% CI: 1.23, 1.73)]. None of the other associations from single pollutant models remained once we restricted to individuals not also exposed to paraquat (Table 3). Here, paraquat-only associations were stronger or the same as paraquat-combination exposures. Additional sensitivity analyses adjusting for paraquat and glyphosate co-exposures that occurred during the same time period also dropped all exposure estimates to the null except for paraquat [OR: 1.42 (95% CI: 1.13, 1.78); data not shown].

Table 2.

Adjusted risk of thyroid cancer with pesticide exposures in the 20 years prior to diagnosis (cases) and interview (controls)

Chemical Name	Exposed cases (n = 2067)	Exposed controls (n = 1003)	Adjusted ORa	Adjusted 95% CI
Thyroid cell DNA damage/breakage
Abamectin	749	310	1.27	0.94, 1.35
Amitraz	164	68	0.96	0.69, 1.34
Captan compoundsb	508	202	1.24	1.00, 1.52
Difenoconazole	30	7	2.37	0.96, 5.81
Diquat dibromide	143	67	1.03	0.74, 1.45
Maneb	392	152	1.23	0.97, 1.55
Milbemectin	7	1	3.68	0.40, 33.94
Naled	589	261	1.10	0.91, 1.34
Propargite	1077	440	1.27	1.07, 1.51
Rotenone compoundsc	93	41	1.19	0.77, 1.82
Tribufos	231	92	1.09	0.82, 1.45
Known endocrine disruptors
2,4-D compoundsd	835	364	1.17	0.98, 1.39
Benomyl	633	283	1.07	0.89, 1.30
Carbofuran	226	86	1.31	0.98, 1.76
Chlorpyrifos	1106	470	1.20	1.01, 1.42
Copper compoundse	1073	467	1.18	1.00, 1.40
Cyhalothrin, lambda karate	244	123	0.94	0.73, 1.22
Diazinon	981	434	1.11	0.94, 1.32
Dicamba compoundsf	546	256	0.98	0.80, 1.18
Glyphosate compoundsg	1307	547	1.33	1.12, 1.58
Linuron	282	99	1.25	0.95, 1.64
Malathion	677	280	1.22	1.01, 1.46
Mancozeb	449	177	1.26	1.01, 1.56
Metolachlor compoundsh	250	123	0.86	0.66, 1.11
Oxyfluorfen	1097	460	1.21	1.02, 1.43
Paraquat dichloride	1209	482	1.46	1.23, 1.73
Permethrin	450	175	1.26	1.01, 1.57
Petroleum oil	902	399	1.14	0.96, 1.35
Talstar	553	216	1.16	0.95, 1.42

^aAdjusted for race/ethnicity, sex, age, and Yost socioeconomic status (39).

^bChemicals included are captan and related compounds.

^cChemicals included are rotenone and related compounds.

^dChemicals included are 2,4-D, diethanolamine salt; 2,4-D, dimethylamine salt; 2,4-D, isopropyl ester; 2,4-D triethylamine salt; and 2,4-dichlorophenoxyacetic acid.

^eChemicals included are copper, copper ammonium complex, and copper hydroxide.

^fChemicals included are dicamba, diglycolamine salt; dicamba, dimethylamine salt and dicamba, dimethylamine salt, and other related chemicals.

^gChemicals included are glyphosate, glyphosate-salt, glyphosate-diammonium salt, glyphosate-isopropylamine salt, glyphosate-potassium salt, and glyphosate-trimesium.

^hChemicals included are metolachlor and S-metolachlor.

Table 3.

Adjusted risk of thyroid cancer for pesticides in combinations with paraquat in a 20-year exposure period

Chemical name	Exposed cases (n = 2067)	Exposed controls (n = 1003)	Adjusted ORa	Adjusted 95% CI
Captanb
Neither captan nor paraquat	840	510	Reference	Reference
Captan without paraquat	18	11	0.64	0.26, 1.59
Paraquat without captan	719	291	1.43	1.17, 1.74
Paraquat and captan	490	191	1.43	1.14, 1.80
Maneb
Neither maneb nor paraquat	844	513	Reference	Reference
Maneb without paraquat	14	8	0.92	0.33, 2.53
Paraquat without maneb	831	338	1.46	1.21, 1.76
Paraquat and maneb	378	144	1.47	1.14, 1.91
Propargite
Neither propargite nor paraquat	736	475	Reference	Reference
Propargite without paraquat	122	46	1.35	0.90, 2.04
Paraquat without propargite	254	88	1.82	1.35, 2.47
Paraquat and propargite	955	394	1.45	1.21, 1.75
Chlorpyrifos
Neither chlorpyrifos nor paraquat	724	446	Reference	Reference
Chlorpyrifos without paraquat	134	75	0.97	0.68, 1.38
Paraquat without chlorpyrifos	237	87	1.78	1.31, 2.43
Paraquat and chlorpyrifos	972	395	1.39	1.16, 1.68
Copperc
Neither copper nor paraquat	703	438	Reference	Reference
Copper without paraquat	155	83	1.14	0.81, 1.60
Paraquat without copper	291	98	1.86	1.39, 2.49
Paraquat and copper	918	384	1.42	1.17, 1.73
Diazinon
Neither diazinon nor paraquat	743	446	Reference	Reference
Diazinon without paraquat	115	75	0.90	0.63, 1.30
Paraquat without diazinon	343	123	1.68	1.29, 2.19
Paraquat and diazinon	866	359	1.36	1.13, 1.65
Glyphosated
Neither glyphosate nor paraquat	644	406	Reference	Reference
Glyphosate without paraquat	214	115	1.11	0.82, 1.50
Paraquat without glyphosate	116	50	1.75	1.15, 2.65
Paraquat and glyphosate	1093	432	1.49	1.24, 1.80
Malathion
Neither malathion nor paraquat	782	479	Reference	Reference
Malathion without paraquat	76	42	1.03	0.65, 1.61
Paraquat without malathion	608	244	1.44	1.16, 1.78
Paraquat and malathion	601	238	1.48	1.20, 1.83
Mancozeb
Neither mancozeb nor paraquat	814	501	Reference	Reference
Mancozeb without paraquat	44	20	1.26	0.68, 2.33
Paraquat without mancozeb	804	325	1.47	1.21, 1.78
Paraquat and mancozeb	405	157	1.48	1.15, 1.89
Oxyfluorfen
Neither oxyfluorfen nor paraquat	739	460	ref	ref
Oxyfluorfen without paraquat	119	61	1.08	0.74, 1.57
Paraquat without oxyfluorfen	231	83	1.87	1.37, 2.55
Paraquat and oxyfluorfen	978	399	1.41	1.17, 1.70
Permethrin
Neither permethrin nor paraquat	821	507	Reference	Reference
Permethrin without paraquat	37	14	1.66	0.81, 3.40
Paraquat without permethrin	796	321	1.48	1.22, 1.79
Paraquat and permethrin	413	161	1.50	1.17, 1.91

^aAdjusted for race/ethnicity, sex, age and Yost socioeconomic status (39).

^bChemicals included are captan and related compounds.

^cChemicals included are copper, copper ammonium complex and copper hydroxide.

^dChemicals included are glyphosate; glyphosate, salt; glyphosate, diammonium salt; glyphosate isopropylamine salt; glyphosate, potassium salt; glyphosate, trimesium.

Our 5- and 10-year single pollutant model analyses generally produced very similar point estimates as reported for the 20-year exposures except for amitraz (OR: 3.72 (95% CI: 1.67, 8.32)] and rotenone (OR: 2.41, 95% CI: 1.18, 4.92), for which effect sizes were larger at 10 years than at 20 years of exposure [Supplementary Table 2 (38)]. However, these associations did not remain once we adjusted for paraquat co-exposure.

We found that risk of thyroid cancer increased in an exposure-response fashion with the number of pesticides that subjects were exposed to in the 20-year cumulative exposure window (Table 4); that is, larger numbers of pesticides contributed to increasing risk [OR for 1-9 pesticides: 1.29 (95% CI: 1.03, 1.61) vs OR for ≥10 pesticides: 1.34, (95% CI: 1.08, 1.68)].

Table 4.

Risk of thyroid cancer relative to total number of pesticides exposed to in the 20 years prior to diagnosis (cases) and interview (controls), based on address at time of diagnosis and interview

	Exposed cases (n = 2067)	Exposed controls (n = 1003)	Adjusted ORa	95% CI
All participants
0	411	260	Reference	Reference
1-9 pesticides	788	376	1.29	1.03, 1.61
≥10 or more pesticides	868	367	1.34	1.08, 1.68

	(n = 470)	(n = 545)
Participants ≥65 years of age
0	92	141	Reference	Reference
1-9 pesticides	193	220	1.41	1.00, 1.98
≥10 pesticides	185	184	1.60	1.12, 2.27

^aAdjusted for race/ethnicity, sex, age and Yost socioeconomic status (39).

Sensitivity analyses that restricted to controls with a lifetime address history (n = 812) produced minimal differences in point estimates (1-12%). When we limited participants to those ≥65 years of age, most associations strengthened [Supplementary Table 3 (38)]; for example, captan compounds, propargite, carbofuran, and glyphosate compounds were each associated with a ≥60% increased risk of thyroid cancer.

While analyzing effect of each race/ethnicity individually, restricting to White non-Hispanic cases and controls increased all effect estimates [Supplementary Table 4 (38)]. We estimated effects similar to total population results for all Hispanics for most pesticides; however, smaller sample sizes resulted in wider CIs (data not shown). After adjusting for paraquat co-exposure, only paraquat and glyphosate exposures [among non-Hispanic Whites only, OR: 1.33 (95% CI: 1.01, 1.75)] were associated with risk of thyroid cancer.

Stratifying by disease stage, we found estimates to be consistently larger for distant/regional disease compared with localized disease; particularly for glyphosate compounds [OR: 1.37 (95% CI: 95% 1.08, 1.73) vs OR: 1.26 (95% CI: 1.04, 1.52), distant/regional vs localized, respectively], paraquat [OR: 1.65 (95% CI: 1.31, 2.08) vs OR: 1.34 (95% CI: 1.11, 1.62)], and propargite [OR: 1.38 (95% CI: 1.10, 1.73) vs OR: 1.21 (95% CI: 1.00, 1.46)]. These associations did not remain after adjustment for paraquat.

Figure 1 shows the use trend in California between 1979 and 2012 for 5 pesticides shown to increase risk of thyroid cancer in our study: glyphosate, oxyfluorfen, paraquat, naled, and benomyl. Glyphosate use rose substantially over the last 3 decades as did use of paraquat and oxyfluorfen, although to a lesser degree. Both naled and benomyl were heavily applied in earlier years; however, after 1998, the use of both chemicals decreased. Supplementary Figure 1 depicts the geographic distribution of each pesticide during the same years and highlights differences in the distribution of various pesticides across the tricounty region (38).

Figure 1.

Trends in the use of glyphosate compounds, oxyfluorfen, paraquat dichloride, naled, and benomyl in California, 1979-2012.

Discussion

We examined associations between 29 preselected hazardous chemicals and risk of thyroid cancer. Paraquat exposure increased risk of thyroid cancer in all pollutant models. Our study found that cumulative long-term exposure to a greater number of unique pesticides increased risk of thyroid cancer (ie, risk of thyroid cancer increased in each of the following exposure categories: ≤1 pesticide, 1-9 pesticides, and ≥10 pesticides).

Paraquat is one of the most widely applied herbicides in the United States and is primarily used for weed and grass control, particularly for fruit orchards and plantation crops as it does not harm mature bark. Paraquat is considered the most highly toxic herbicide to be marketed over the last 60 years and has been banned in 32 countries; however, in the United States, it is considered a restricted-use pesticide and is available to commercially licensed applicators. We detected an association between paraquat exposure and thyroid cancer in all exposure windows and in isolation from other chemical exposures. While there are no studies of paraquat and thyroid cancer in humans, postmortem analyses of persons with paraquat poisoning found measurable levels in the thyroid (40). The Agricultural Health Study examined whether pesticide exposure among female spouses of male applicators was associated with thyroid disease and found ever use of paraquat to be associated with hypothyroidism among the women (31). Further, thyroid adenomas have been reported in Fischer rats exposed to increasing concentrations of paraquat (41). These findings suggest that paraquat may be a critical agent in the development of thyroid cancer.

Although paraquat appears to have the strongest association with thyroid cancer among individual pesticides, we found that exposure to greater numbers of pesticides also increased risk. No other studies have examined the association between combined pesticide exposures during a specified time-period and risk of thyroid cancer; however, one study found that men who were sampled during “high-pesticide use” seasons had higher levels of thyroid-stimulating hormones (TSH) (42). Similar results were also observed among aerial fungicide applicators in Minnesota, who had greater TSH levels during high- vs low-exposure seasons (43). Another in vitro study evaluated the association between 12 currently used pesticides in Denmark and thyroid hormone function in cell culture and found that a 4-component pesticide mixture acted additively to disrupt thyroid hormone function (44).

Thyroid hormones regulate metabolism and influence growth and immune function. Mounting evidence suggests a role for hormones in the development of neoplasms and cancer risk (45-47). Uncontrolled hypothyroidism and the effects of elevated TSH on thyroid growth promotion have been suggested as possible risk factors in thyroid cancer development (48). Hashimoto’s disease has also been postulated to be a risk factor for thyroid cancer due to thyroid autoimmune inflammatory changes, although the association is uncertain given conflicting evidence (49-51). Additionally, it is possible the relationship between Hashimoto’s thyroiditis and thyroid cancer could be attributed to increased use of ultrasounds detecting early-stage thyroid cancers (52, 53). The link between hyperthyroidism and subsequent thyroid cancer development is also uncertain, yet some studies suggest a link, particularly among patients who have the autoimmune disorder Graves’ disease (54-56).

Permethrin exposure showed an elevated effect estimate for thyroid cancer, with wide CIs, in multipollutant models. There are no epidemiologic studies of permethrin exposure and thyroid cancer; however, in vitro studies have found permethrin to alter levels of thyroid hormones (57). As such, to the best of our knowledge, this study provides the first epidemiologic evidence of the effects of permethrin exposure on thyroid cancer risk.

There are several possible mechanisms for which paraquat could plausibly increase the risk of thyroid cancer. Utilizing the 10 key characteristics of carcinogens identified by the International Agency for Research on Cancer (58), the following likely applies to paraquat and possibly other pesticides: (1) it alters DNA repair, as noted in a study by our co-author (13); (2) it has been found to generate reactive oxygen species, causing cellular damage in many organs (59); and (3) cell proliferation is implicated in cancer in general and some pesticides induce cell proliferation (60, 61), including in thyroid cells (62). However, few studies examine the biological mechanisms for pesticide exposure and risk of thyroid cancer; as such, strong conclusions cannot be made.

There are some limitations to the present study. Although our GIS model can determine the number of pounds of each active ingredient applied per acre within a 500-m radius of addresses, we cannot consider this a quantitative measure of exposure since that would depend on many other factors including drift, presence during application times, etc. Also, the quantities are not comparable across pesticides since pounds of active ingredient applied do not accurately translate into a measure of human toxicity, and we currently have no way to standardize these measures. Thus, participants were considered exposed if pesticides were applied within a 500-m buffer of their residence.

An additional limitation to our analyses is that we were only able to utilize residential addresses rather than a combination of residential and occupational addresses. Most agricultural pesticides are applied during working hours; therefore, employing occupational addresses in our study would have allowed us to estimate exposures experienced at the workplace and may have more accurately reflected total pesticide exposure. We also had access to just one address for cases (at time of diagnosis); however, residential moving behaviors in our control population were stable in the 20-year period before enrollment interview (unpublished data), and age and neighborhood SES were similar between cases and controls in our study, leading us to presume cases would have similar moving behaviors. On average, PEG study controls lived at their enrollment address for 18.4 years prior to interview. Nevertheless, the absence of residential address history for our thyroid cancer cases remains a limitation of the study.

Nondifferential exposure misclassification due to address reporting error is a possibility in our study and may have attenuated our effect estimates. The accuracy of our GIS-based pesticide exposure measures relies on the quality of reported addresses, which we expect to be high. To estimate the effects of residential mobility rates in our case population (for whom we only had 1 address for exposure measurements), we performed a subanalysis restricting to older individuals, who we anticipated were more likely to have been living in the same location for a longer time, and found point estimates to be stronger [Supplementary Table 3 (38)]. Finally, we did not have access to risk factor data for our case population. Specifically, we do not know the patients’ history of radiation exposure, family history, or thyroid function. Other known risk factors for which we did have data (ie, sex, age, race) were included in our adjusted models.

The strengths of this study include the large sample size and detailed information on pesticide exposures based on residential addresses combined with our land use and PUR data based pesticide exposure assessment. This allowed us to determine exposure for subjects living near agricultural pesticide applications even though they may be unaware of their exposure due to pesticide drift. This approach is superior to relying on recall alone since it eliminates the possibility of differential recall bias. Additionally, a previous validation study of our GIS-PUR pesticide exposure model found that it has good specificity (87%) to identify individuals with high serum levels of dichlorodiphenyldichloroethylene (63).

In conclusion, we found paraquat exposure to be positively associated with thyroid cancer. The association between paraquat and thyroid cancer is elevated in multipollutant models, suggesting an increased risk of thyroid cancer with exposure to other pesticides in combination with paraquat. Further, independent of paraquat exposure, exposure to a greater number of unique pesticides within a 20-year time frame subsequently increased the risk of thyroid cancer. More epidemiologic research is needed to better understand the relationship between certain pesticides and risk of thyroid cancer in the population, particularly research that uses prospectively collected pesticide exposure and disease data. The present study reports several novel associations, and further research is needed to confirm these findings and to evaluate the mechanisms of action.

Financial Support

A.H. (principal investigator): University of California Cancer Research Coordinating Committee—“Relation of pesticide exposure to thyroid cancer incidence and stage distribution.” Grant no. CRN-15-380517.

Disclosures

The authors have nothing to disclose.

Data Availability

Some or all data sets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.