|Home | About | Journals | Submit | Contact Us | Français|
Organochlorine (OC) insecticides have been regulated as possible human carcinogens primarily on the basis of animal studies. However, the epidemiologic evidence is inconsistent. We investigated the relationship between cancer incidence and OC insecticide use among pesticide applicators enrolled in the Agricultural Health Study, a prospective cohort study of 57,311 licensed applicators in Iowa and North Carolina enrolled between 1993 and 1997. Information on ever use of seven organochlorine insecticides (aldrin, chlordane, DDT, dieldrin, heptachlor, lindane, toxaphene) was collected from a self-administered questionnaire at enrollment. Lifetime exposure-days to OC insecticides were calculated using additional data from a take-home questionnaire completed by 25,291 participants (44% of total). We found no clear evidence of an association between use of OC insecticides and incident cancers (N=1,150) ascertained through December, 2002. When we focused on individual insecticides and structurally similar groups (aldrin and dieldrin; chlordane and heptachlor), significantly increased relative risks of some cancers were observed for use of some chemicals (rectal cancer and chlordane, lung cancer and dieldrin, non-Hodgkin lymphoma (NHL) and lindane, melanoma and toxaphene, leukemia and chlordane / heptachlor). Some significant decreased relative risks were also observed (colon cancer and aldrin; overall cancer and heptachlor). In Conclusion, we did not observe any clear relationship between cancer risk and the use of OC insecticides. Our chemical-specific findings are based on small numbers and multiple comparisons, and should be interpreted with caution; however, some observed associations (lindane and NHL, chlordane/heptachlor and leukemia) are supported by previous evidence.
Organochlorine (OC) insecticides are a class of insecticides characterized by their cyclic structure, number of chlorine atoms and low volatility. These agents were widely used in agriculture and pest control between the 1940s and 1960s. Concern over their environmental persistence and possible health effects led the U.S. EPA to restrict or ban their use during the 1970s and 1980s. Most organochlorine insecticides are rarely used in the United States today; however, they continue to be employed in some developing countries, most notably India (1–3).
Evidence that OC insecticides demonstrate weak estrogenic and anti-estrogenic properties led to speculation that these chemicals may act as tumor promoters through hormonally-mediated effects (4;5). Some epidemiologic studies have linked these chemicals to increased risks of soft-tissue sarcoma, non-Hodgkin lymphoma, leukemia and cancers of the prostate, lung, pancreas and breast (6–12), although these findings have been generally weak and inconsistent. Breast cancer has received the most attention, though the findings now are largely perceived as showing no overall association (7). OC insecticides have different pesticidal properties, and may also differ in their toxicologic effects; if so, the inconsistency in findings for non-breast cancers may be partly due to differences across study populations in the types of OC chemicals used. Findings from experimental studies of animals are more suggestive of carcinogenic effects (6;7;13–15). The International Association for Research on Cancer has evaluated OC insecticides as being either possibly carcinogenic to humans (DDT, chlordane, heptachlor, toxaphene), or not classifiable as to carcinogenicity (aldrin, dieldrin, lindane) (13–15).
In order to better understand whether OC insecticides are carcinogenic to humans, additional epidemiologic evidence is needed from prospective studies with detailed information on individual exposure to specific OC pesticides and other risk factors for the diseases of interest. With this in mind, we investigated the relationship between site-specific cancer incidence and OC insecticide use among pesticide applicators enrolled in the Agricultural Health Study.
The Agricultural Health Study (AHS) cohort has been previously described in detail (16). Briefly, the AHS is a prospective study of 57,311 Iowa and North Carolina residents licensed to apply restricted-use pesticides. Recruited between 1993 and 1997, cohort members include private applicators (farmers and nursery workers) and commercial applicators (employees of pest control companies or businesses that use pesticides; from Iowa only). Cancers among AHS applicators diagnosed through December 31, 2002 were identified through data linkage with files from the Iowa Cancer Registry, North Carolina Central Cancer Registry, state death registries and the National Death Index, and were coded according to the International Classification of Diseases for Oncology, 2nd edition. Cohort members who were alive but no longer residing in Iowa or North Carolina were identified through current address records of the Internal Revenue Service (address information only), Motor Vehicle Registration offices, and pesticide license registries of the state agricultural departments. Person-year accumulation for cancer incidence of individuals who had moved from the state was censored in the year they departed. The mean time of follow-up was 7.3 years. All participants provided verbal informed consent, and the protocol was approved by the institutional review boards of the National Cancer Institute, Battelle, the University of Iowa, and the research corporations Battelle and Westat, which were contracted to coordinate study management.
A self-administered enrollment questionnaire collected comprehensive exposure data on 22 pesticides and information on ever use of 28 pesticides. Study participants also provided information on personal protective equipment, pesticide application methods, pesticide mixing, equipment repair, smoking history, alcohol consumption, cancer history of first-degree relatives, and basic demographic characteristics. Applicators completing the enrollment questionnaire were also given a self-administered take-home questionnaire which sought additional information on pesticide use and occupational exposures. Reminder cards were mailed to non-respondents at two weeks and ten weeks post-enrollment. Forty-seven percent of the enrolled farmers completed and returned the take-home questionnaire. The questionnaires may be accessed at http://www.aghealth.org/questionnaires.html.
Ever use of seven OC insecticides (aldrin, chlordane, DDT, dieldrin, heptachlor, lindane, toxaphene; see Table 1) was ascertained from the enrollment questionnaire, with more detailed exposure information collected from the take-home questionnaire. We calculated two measures of cumulative exposure for (1) each chemical, (2) groups of structurally similar chemicals (aldrin and dieldrin; chlordane and heptachlor), and (3) for overall OC insecticide use. Lifetime exposure-days for a given chemical, the first measure, was calculated as the product of the number of years a participant personally mixed or applied that chemical and the number of days in an average year that chemical was used. An intensity-weighted measure for each chemical was also calculated by multiplying the lifetime exposure-days with an exposure intensity score. Intensity scores were calculated using an algorithm that takes into account various factors influencing actual exposure, including whether the applicator personally mixed or prepared the pesticides for application, the method of application, the repair of pesticide equipment, and the use of various types of personal protective equipment during these activities (17). Overall measures of lifetime exposure-days and intensity-weighted lifetime days exposed to (1) structurally similar chemical groups and (2) total OC insecticides were calculated as the sum of the respective chemical-specific metrics.
Our criteria for defining different analytic samples within this study are summarized in Figure 1. Briefly, we excluded cohort members from analyses involving ever use of specific chemicals if they were diagnosed with cancer prior to enrollment (N=1,075), missing information on age (N=2), missing information on ever use for all OC pesticides (N=5,055), or if they were lost to follow-up or otherwise did not contribute any person-time (N=296). After these exclusions, data from 51,011 applicators remained. For analyses involving measures of cumulative exposure to each OC chemical, we excluded subjects who did not complete a take-home questionnaire (N=26,428) or for whom cumulative exposure metrics could not be calculated for any of the OC chemicals due to missing data (N=506). Following these exclusions, an analytic data set of 24,077 applicators remained. For analyses of cumulative exposure to all OC pesticides combined, we further excluded subjects with missing cumulative exposure data for any of the OC chemicals (N=1,668); an analytic data set of 22,409 applicators remained.
To investigate the relationship between OC insecticide exposure and cancer risk, we performed Poisson regression modeling for individual cancer sites to calculate rate ratios (RR) and 95% confidence intervals (CI) for different categories of exposure (ever exposure; category of unweighted/weighted lifetime exposure-days). Rate ratios were adjusted for age at enrollment (<40, 40–49, 50–59, 60–69, 70+ years), sex, state, education (≤ high school graduate, > high school graduate), smoking (non-smoker, ≤ 12 pack-years, 12+ pack-years), alcohol use (non-drinker, ≤6 drinks/month, >6 drinks/month), family history of cancer and lifetime days of total pesticide application (<51, 51–115, 116–235, 236–507, 508+ days). Analyses with additional adjustment for body mass index (<18.5, 18.5–24.9, 25–29.9, 30+ kg/m2) yielded virtually identical findings (data not shown).
For measures of lifetime days and intensity-weighted lifetime days to all OC insecticides, we conducted our analyses using two different reference groups. The first reference group consisted of applicators unexposed to any OC insecticides, compared to exposed applicators grouped into three categories of exposure (low, medium, high) using the tertiles among exposed cancer cases as cut-points. The second reference group consisted of exposed applicators in the lowest category of exposure. For analyses of specific chemicals and structurally similar chemical groups, exposed cases were grouped into two categories using the median exposure among exposed cases as a cut-point.
In either analysis, if the upper exposure category contained 10 or more cases, that category was further sub-divided by the median exposure level, but only if the resulting categories contained at least five cases. The rationale for this strategy was to enable estimation of relative risk across a wider range of exposure levels while minimizing the potential for creating unstable relative risk estimates based on small numbers.
Tests for trend of cumulative exposure variables were performed by modeling the exposure category medians as a quantitative score and calculating Wald test statistics. We limited analyses to cancer sites for which there were at least 20 cases exposed to any OC pesticide among take-home questionnaire respondents and at least 5 cases per category of exposure level. Given that nearly three quarters of the subjects aged less than 50 at enrollment reported that they had not used OC insecticides (a reflection of the fact that these chemicals were banned or restricted in the 1970s and 1980s), we also repeated our analyses restricted to applicators aged 50 or older.
Of the 51,011 AHS applicators included in analyses, 24,384 (48%) reported having ever used an OC insecticide on the enrollment questionnaire. A summary of selected characteristics of AHS applicators in relation to OC insecticide exposure level is provided in Table 2. The study cohort is composed primarily of male, white, private applicators. Exposure to OC insecticides was strongly associated with age; 36% of applicators reporting high exposure to total OC insecticides were aged 60 or older at enrollment, whereas 25% of low-exposed and 12% of unexposed applicators were of the same age group. AHS applicators reporting OC insecticide use were more likely to be North Carolina residents, smokers, to have high lifetime overall exposure to pesticides and a family history of cancer, and were less likely to consume alcohol.
Results from analyses of ever use of OC insecticides are summarized in Table 3. Use of any OC chemical was associated with a borderline statistically significant increased risk of leukemia (RR 2.0, 95% 1.0–4.1) and a significantly decreased risk of colon cancer (RR 0.6, 95% 0.5–0.9). No associations with overall cancer incidence or other site-specific rates were found. Chemical-specific associations with leukemia were also observed with lindane and heptachlor (statistically significant two-fold relative risks), and non-significant relative risks of 1.5 or greater were observed for chlordane, dieldrin and toxaphene. Other statistically significant associations with specific chemicals included a 50% excess in lung cancer incidence with lindane use, and an increased risk of rectal cancer with use of chlordane (RR 1.7, 95% CI 1.0–2.8) and toxaphene (RR 2.0, 95% CI 1.1–3.5). Non-significant relative risks of 1.5 or greater were also observed for NHL and toxaphene and for melanoma and heptachlor. Our findings did not materially change when we repeated these analyses only among respondents of the take-home questionnaire.
Results from analyses of unweighted- and intensity-weighted lifetime days of OC insecticide use among take-home questionnaire respondents, with never exposed subjects as the referent group, are provided in Table 4. OC exposure level was not associated with the incidence of all cancers combined nor for any specific cancer site. Among solid tumors, relative risks of cancers of the lung, bladder and rectum increased moderately with increasing level of exposure; however, monotonic increases in relative risk with increasing exposure were not observed, and tests for trend were not statistically significant (although the test for trend for rectal cancer by intensity-weighted days was of borderline significance). The associations with cancers of the lung and bladder disappeared when low exposure was defined as the referent group, while the association with rectal cancer became slightly stronger (unweighted exposure: RR 1.8 (95% CI 0.5–7.5) for moderate exposure, 2.7 (95% CI 0.7–10.7) for high exposure, Ptrend 0.19; intensity-weighted exposure: RR 0.8 (95% CI 0.2–3.4) for moderate exposure, 2.8 (95% CI 0.8–9.3) for high exposure, Ptrend 0.03).
Moderately elevated risks of some hematopoietic cancers were observed among OC-exposed applicators. The relative risk for non-Hodgkin lymphoma (NHL) increased with increasing levels of unweighted exposure (RR 1.5 for highest exposure group) and intensity-weighted exposure (RR 1.7), although these risk estimates and tests for trend were not statistically significant. Organochlorine insecticide use was associated with a two-fold increase in leukemia risk; however, risk did not increase with rising level of exposure. When low-exposed subjects were used as the referent group, there was no evidence of an association with NHL or leukemia.
Separate analyses of exposure level to individual OC insecticides and to groups of structurally similar chemicals (aldrin and dieldrin; chlordane and heptachlor) were performed. For most chemical-cancer site comparisons, there was little evidence of a relationship (online supplement). Statistically significant increases in cancer risk were observed for chlordane and rectal cancer (Table 5; Ptrend 0.03, 0.09 for cumulative and intensity-weighted exposure respectively), dieldrin use and lung cancer (Ptrend 0.02, 0.002), lindane and NHL (Ptrend 0.12, 0.04), toxaphene and melanoma (Ptrend 0.03, 0.24) and the chlordane / heptachlor group and leukemia (Ptrend 0.02, 0.10). Significant decreased relative risks were observed for aldrin and colon cancer (Ptrend 0.19, 0.04) and heptachlor and all cancers (Ptrend 0.03, 0.04).
When we restricted our analyses to subjects aged 50 or older, our findings did not materially change.
In this analysis of data from the AHS cohort, we found no clear evidence of an association between overall use of OC insecticides and cancer incidence. We observed an excess of leukemia and a deficit of colon cancer among workers who reported ever using any OC chemicals; however, no clear dose-response relationship with increasing level of exposure was apparent for these or other cancers. These findings do not suggest that OC insecticides, as a whole, are carcinogenic in our study population, although it should be noted that this study did not consider all types of OC insecticides, as some chemicals (e.g., endrin, methoxychlor, mirex) were not assessed in the AHS questionnaires.
While we found generally no evidence of a relationship between overall OC use and cancer risk, it is important to recognize that analyses of a class of pesticides are informative only to the extent that the effects of intra-class chemicals are similar. Organochlorine insecticides possess very different chemical structures, and may exert different biologic effects in humans. If OC chemicals differ in carcinogenic potency and mechanism of action, then analyses of overall OC use may dilute or conceal chemical-specific effects. We observed some statistically significant associations with lifetime days of exposure to individual chemicals. Most associations involved elevated cancer risks (rectal cancer and chlordane, lung cancer and dieldrin, NHL and lindane, melanoma and toxaphene, leukemia and chlordane / heptachlor), although reductions in risk were also observed (colon cancer and aldrin; overall cancer and heptachlor). Only two of these associations with cumulative exposure were also observed in analyses of ever use of specific chemicals in the entire cohort (rectal cancer and chlordane, leukemia and chlordane/heptachlor). The chemical-specific analyses involved a large number of comparisons, and, consequently, at least some of these findings may be due to chance. However, some of the observed associations (lindane and NHL, chlordane / heptachlor and leukemia) are supported by previous evidence.
The risk of NHL rose with increasing cumulative exposure to lindane. Though not produced in the U.S. since 1976, imported lindane is still used in this country for seed treatment for a limited number of vegetables, and for the pharmaceutical treatment of scabies and head lice. Oral administration of lindane has been observed to increase the incidence of liver tumors in mice and, less clearly, thyroid tumors in rats (13). Previous epidemiologic evidence has suggested a relationship between lindane exposure and NHL. A moderate association between lindane use and NHL was observed in a pooled analysis of three population-based case-control studies conducted in the Midwestern U.S., with stronger relative risks observed for greater duration and intensity of use (18). An increased risk of NHL with lindane use was also found in a Canadian case-control study of NHL (19). In support of these epidemiologic findings is the observation that lindane induces chromosomal aberrations in human peripheral lymphocytes in vitro (20).
Chlordane and heptachlor are structurally related chemicals. Chlordane is metabolized into heptachlor, and technical-grade products of each contain approximately 10–20% of the other compound (15). These chemicals were banned from use in agriculture in the U.S. in the late 1970s, and completely banned in 1988. Chlordane and heptachlor are established carcinogens in animal models, with orally administered doses clearly demonstrated to induce liver tumors in mice and rats (15). Previously published evidence suggests that these chemicals may be leukemogenic. Two reports have been published describing 25 and 11 cases of blood dyscrasias following exposure to chlordane and heptachlor, respectively, usually as a result of pest control treatment in the home or garden (21;22). Additionally, heptachlor has been observed to induce tumor promoting mechanisms in human myeloblastic leukemia and lymphoma cells (23;24). Findings from a U.S. population-based case-control study of leukemia generally did not suggest an association with chlordane, although a three-fold elevated risk was observed among farmers reporting 10 or more days of use annually on animals (25). Some case-control studies of NHL have reported associations with chlordane/heptachlor exposure (26;27), although later studies found no such relationship (19;28). We did not observe a relationship with NHL in our study.
Our finding that lung cancer relative risk increased significantly with increasing dieldrin exposure was previously reported in an earlier AHS analysis by Alavanja et al. (9). Dieldrin produces liver tumors in mice, though not in other animal models (13), and has been reported to induce chromosomal damage in a dose-response manner in a human embryonic lung cell line (29). However, no elevated cancer rates were observed among workers employed in the manufacture of dieldrin, aldrin and endrin (30–32).
We found equivocal evidence of an increase in rectal cancer risk with overall OC use, and stronger evidence for use of chlordane. Pesticide use has generally not been linked with rectal cancer in previous epidemiologic studies, and reports of rectal cancer incidence among farmers are inconsistent, with both increased (33–36) and decreased (37–39) risks observed. An excess risk of rectal cancer accompanying pesticide use was observed in a small cohort of Icelandic pesticide applicators (40); however, the association was based on small numbers, and no specific information on OC insecticides was available (33). A significantly elevated number of deaths due to rectal cancer was observed in a cohort of workers involved in the production of dieldrin and aldrin, although a dose-response relationship with exposure intensity was not present (32).
We also observed an increased risk of melanoma accompanying high exposure to toxaphene, although this finding was based on small numbers. Toxaphene, a complex mixture of chlorinated camphenes, which was widely used in North America until 1982 (41), has been reported to cause liver and thyroid tumors in animal studies and to induce sister chromatid exchanges in vitro (15). In one study, an increased frequency of chromosomal aberrations was reported among workers exposed to toxaphene (42). Otherwise, no previously published epidemiologic evidence suggests carcinogenicity to humans. Case-control studies of NHL and leukemia conducted in Iowa and Minnesota both found no evidence of a relationship with toxaphene use (25;26).
It has been hypothesized that OC insecticides, which demonstrate weak estrogenic and anti-estrogenic properties (4;5), may play a role in the pathogenesis of hormone-related cancers such as breast and prostate (43). However, epidemiologic studies of these agents generally do not support a relationship with breast cancer (7), and are unclear with respect to prostate cancer (44–47). While our analysis of pesticide applicators is not informative with respect to breast cancer and OC use, we did find consistent evidence suggesting no increased prostate cancer risk with exposure to OC insecticides. These findings are consistent with an earlier case-control study of prostate cancer conducted within the AHS cohort (8). In that study, a variable defined from factor analysis that included age greater than 50 and OC insecticide use was found to be weakly associated with increased prostate cancer risk; however, no relationship between prostate cancer and cumulative exposure to OC chemicals was observed. Our current findings offer further evidence that use of OC insecticides is not associated with prostate cancer risk.
This study has several strengths. First, information on exposure to OC chemicals was conducted prior to disease onset, precluding the possibility of differential recall bias. Second, the extensive information collected in the AHS regarding exposure to each insecticide enabled the development of detailed measures of OC exposure for use in our analysis. Third, unlike previous cohort studies investigating exposure to OC insecticides, we were able to control for a variety of potentially confounding occupational, demographic and lifestyle factors in our analysis. Fourth, our outcome for this study is cancer incidence, using data collected from population-based cancer registries, which eliminates issues of survival bias when cancer mortality is the endpoint of interest.
There are also limitations to this study. First, only 44% of the enrolled study subjects completed the take-home questionnaire, which was the source of detailed information regarding OC insecticide use. This raises the question as to whether selection bias may have influenced our findings and may have limited the generalizability of our sample. However, an earlier analysis found that individuals completing the take-home questionnaire were older on average than non-respondents, but otherwise comparable (48). Since we adjusted for age in all analyses, it is unlikely that selection bias is a plausible explanation for our findings. Second, study subjects would mostly have been recalling past, not current, use of OC insecticides at the time of data collection since most OC insecticides were banned long before study enrollment. The subjects’ recall of pesticide exposures several years in the past could introduce measurement error. Since information on pesticide use was collected prior to disease diagnosis, the misclassification should be non-differential and would likely lead to an attenuation of observed risk estimates. Although exposure misclassification may have occurred, previous evaluation of this issue has shown that recall of pesticide use in the AHS is comparable to that of other factors commonly obtained by interview in epidemiologic studies, Test-retest percentage agreement ranged from 70% to more than 90% for ever vs. never use of specific pesticides and from 50% to 60% for duration, frequency, or decade of first use of specific pesticides (49). A majority of subjects reported pesticide use duration information that was plausible (i.e., not an overestimate) in relation to the years that that pesticide had been registered for use (50). Third, the timing of the AHS (recruitment from 1993–1997, follow-up through 2002) may have been too late to capture cancer risks associated with OC insecticides, many of which were taken off the market in the 1970s and 1980s. However, the OC insecticide body burden among AHS applicators may still be high, given the tendency of these chemicals to accumulate and persist in fatty tissue. When we restricted our analysis to subjects aged 50 or older at enrollment, who were most likely to have applied OC insecticides in large amounts prior to their restriction, our findings did not change. Fourth, the follow-up of this cohort is relatively short, and for some cancers the number of accrued cases is small, particularly for chemical-specific analyses; as a result, analyses of these cancers had limited statistical power to detect associations of moderate size. Additionally, given the large number of comparisons performed in our analyses, we cannot rule out the possibility that some of our observed chemical-specific findings may have arisen due to chance.
In conclusion, our analysis of data from the AHS cohort suggests that, overall, use of OC insecticides was not related to cancer risk. We did, however, observe associations among specific chemicals, some of which (lindane and NHL, chlordane/heptachlor and leukemia) are supported by previously published studies and warrant further investigation. New studies of OC insecticide use in Western countries are probably no longer feasible, given that most chemicals have been off the market for many years. Instead, the most suitable settings for such projects nowadays are likely to be found in India and other developing countries where these chemicals are still in use. A future re-analysis of AHS applicators following additional follow-up may also be informative for clarifying whether these insecticides influence cancer risk.
This research was supported by the Intramural Research Program of NIH (National Cancer Institute and National Institute of Environmental Health Sciences).
Manuscript Novelty and Impact: This manuscript describes findings from an investigation of the relationship between site-specific cancer incidence and organochlorine insecticide use among pesticide applicators enrolled in the Agricultural Health Study, a large prospective study that, unlike most other occupational cohort studies, recorded detailed information on exposure to specific organochlorine insecticides as well as on other potentially confounding risk factors. While cancer risk was unrelated to overall organochlorine insecticide use in our study population, we did observe some chemical-specific associations (lindane and NHL, chlordane/heptachlor and leukemia) that are supported by previous evidence.