Role of the herbicide atrazine in the development of non-Hodgkin's lymphoma.

A. Role of the her bicide atrazine in the development of non-Hodgkin's lymphoma. Scand J Work Environ Health 1993;19:108-14. Atrazine is the most commonly used herbicide in the United States and is a wide spread groundwater contaminant in the Midwest. The role of atrazine in the development of human non-Hodgk in's lymphoma (NHL) was investigated in three case-referent studies conducted in four midwestern states in the United States. A total of 993 white men with NHL and 2918 population based referents were interviewed concerning their agricultural practices. When the results of the three studies were combined, atrazine use was associated with an odds ratio of 1.4 [95% confidence inter val (95% CI) 1.l-1.8, 130 cases, 249 referents) for NHL. However, adjustments for the use of 2,4 dichlorophenoxyacetic acid and organophosphate insecticides reduced the apparent association between NHL and atrazine in all but one state and reduced the associations for the long-term and frequent users in Nebraska. Detailed analyses suggested that there was little or no increase in the risk of NHL attributable to the agricultural use of atrazine.

In 1990, over 75 million pound s (34 million kilograms) of the triazine herbicide atrazin e was used in the United States for weed control on com, sorghum, sugarcane, macadamias, and pineappl e (1-3). Over two-third s of the use was associated with com (4). Atraz ine was also used on golf courses, lawns, pastureland , and in Christmas-tree production (3). Atrazine is the most commonly used herbicide in the United States (1) and is a wide spread groundwat er contaminant in the Midwest (3). The use of atrazine in Germany (5) and parts of Italy (P Crosignani, personal communicati on, 1990) has been banned to prevent further groundwater contaminat ion.
Atrazine forms genotox ic N-Nitroso compound s under certain conditions (6)(7) and has been shown to increa se the incidence of malignant mammary tumors in female Sprague Dawley rats (8), beni gn mammary tumors in male Fisher 344/LATI rats (9), and uterine adenocarcinoma s and lymphatic and he-matopoietic malignancies in female Fisher 344/LATI rats (9). Lymphom as have also been found to occur in excess in male and female Swiss mice (10, II ), but the International Agency for Research on Cancer (12) has noted that these results have limitations.
Atrazine may playa role in hormonall y mediated tumors, such as breast, endometrium, ovary, and prostatic cancer, because of its action on the hypothalamic-pituitary-gonadal axis (12). Atrazine and its metabolite impair steriod hormone metabolism (12). Atrazine has been shown to inhibit 5-alpha-steroid reductase in the pituitary of rats, reduce 3 alpha-hydroxysteroid dehydrogenase activity in pups, and decrease the number of androgen-specific bind ing sites in the prostate of pups (12). In vitro studies have also shown atrazine to inhibit androgen metabol ism (12). While atrazine inhibits some hormone s, it appears to induce others. Notably it causes increases in luteinizing hormone and follicular-stimul ating hormone (13).
In humans, atrazine has been linked to ovarian cancer (14), colon cancer ( 15), and non-Hodgkins lymphoma (16), and it is currently classified as a possible human carcinogen by the United States Environmental Protection Agency (class C) (I) and by the International Agency for Research on Cancer (group 2B) (12).
The association of atrazine with non-Hod gkin 's lymphoma was observed in a case-referent study conducted in Kansas ( 16). Subsequentl y, two additional case-referent studies of non-Hodgkin's lymphoma, using similar study designs, were conducted in three other midwestern states (Nebraska, Iowa, and Minnesota) of the United States (17)(18). In this report , we have combined the data from these three studies to evaluate the role of atrazine in the development of non-Hodgkin' s lymphom a.

SUbjects and methods
The three studie s evaluated in this report were population-based case -referent studies conducted in Nebraska, Iowa-Minn esota , and Kansas. Detail ed descriptions of the method s for each study have been publi shed elsewhere (16)(17)(18)(19)(20)(21)(22)(23)(24). Each study included several malignancies of the lymph atic and hematopoietic systems and, in Kansas only, soft-tissue sarcoma. The studies in Iowa-Minne sota and Kansas included white men, while the Nebrask a study included both white men and white women . This report eva luates non-Hodgk in' s lymphoma among white men.

Cases
In Nebraska, all cases of non-Hodgkins lymphom a diagno sed between I July 1983 and 30 June 1986 among white men aged 21 years or older and residing in one of the 66 counties of easte rn Nebraska were identi fied through the Nebraska Lymphom a Study Group and area hospitals (N = 227) (table 1). In the Iowa-Minnesota study, all newly diagno sed cases of non-Hodgkin ' s lymphom a among white men, aged 30 years or older, were ascertained from records of the Iowa State Health Registry and a special surveillance of records (N =780) from Minnesota hospitals and pathology laboratories. The diagnosis period for eligibility was March 1981 through October 1983 in Iowa and October 1980 through September 1982 in Minnesota. In Minnesota, non-Hodgk in' s lymphoma patients who resided in the cities of Minneapolis, St Paul, Duluth , or Rochester at the time of diagnosis were excluded. In Kansas, all cases of non-Hodgkin's lymphom a diagnosed from 1979 through 1981 among white men aged 2 1 years or older were identified through the University of Kansas Cancer Data Servic e, which has a register cove ring the state of Kansas. A random sample of 200 men was drawn from the 297 non-Hodgkin's lymphoma cases diagnosed in Kansas during the eligible time period.
The cases were reviewed by expert pathologists and classified according to what is called the Working Formulat ion (25-27). Analyses of follicular (Working Formulation catego ries B-D), diffuse (Working Formul ation categories A, E-G), and other (Working Formulation categories H-K) non-Hodgkin ' s lymphoma are prese nted. Onl y histologically confirmed cases were included. The number of confirmed cases was 220 in Nebraska and 172 in Kansas. In Iowa-Minnesota, the pathology revie w occurred after interviews were obtained for the non-Hodgkin' s lymphoma cases. Because the cases for which there was no interview did not undergo Scand J Work Environ Health 1993, vo1 19, no 2 patholo gy review, the total number of eligible histologically confirmed cases could not be determ ined.

Referents
The referents were randoml y selected from the same geographic areas as the cases, with frequency matching by race, gender, five-year age group, and vital status at the time of the interview. For living cases under 65 years of age, referents were selected by two-stage random digit dialing (28), For living cases aged 65 years or older, the referents were selected from record s of the Health Care Finan cing Administration (Medicare). For the deceased cases, the referents were selected from state mortality files with additional matching for year of death. Persons with the cause death listed as a malignancy under study or, in Kansas and Nebr aska, a malignancy of an illdefined site, homicide, suicide, or legal intervention were excluded. A total of 3379 referents (Nebraska 831, Kansas 1005, Iowa-Minnesota 1543) were identified.

Interview
Interviews were conducted with the subjects or their next-of-kin if the subjects were decea sed or incapacitated. The interviews were done by telephone in Nebraska and Kansas and in-person in Iowa-Minnesota. In Nebrask a, 201 non-Hodgkin's lymphoma patients and 725 referents were interviewed, yielding interv iew respon se rates of 9 1 and 87%, respectively. The overall referent response rate, which accounted for the 9 1% response rate in the household census phase of the random digit-dialing procedure, was 85%. In Kansas, 170 patients and 948 referents were interviewed, yieldin g inter view response rates of 96 and 94%, respectively. The random digit-d ialing household cen sus had a 92.3% response rate, a Pathology review occ urred after the interviews were condu cted. Cases without interview did not undergo pathology review . b The 96% response rate was based on 170 Interviews out of 172 confir med non·Hodg kin 's lymphoma cases initially diagnos ed as non-Hodgki n's lymphoma and five non-Hodqkln 's lymphoma cases initial · Iy diagnosed as other cancer types in the study (eg, Hodgkin's disease).
which made the overall referent response rate 90%. In Iowa-Minnesota, 780 presumptive non-Hodgkin 's lymphoma cases were ascertained, and an interview was obtained for 694 (89%). After pathology review of the cases with interviews 622 cases were confirmed as non-Hodgkin's lymphoma. Interviews were also obtained from 1245 referents (81%) in Iowa-Minnesota. The overall referent response rate , accounting for the 87.5 % household censu s response rate, was 78%. When the three studies were combined, interviews were obtained from 993 eligible cases and 2918 referents.
In each study, the interviews included detailed questions on agr icultural practices. In Nebra ska, the use of atrazine on the farm where the subjects lived or worked was ascertained via an open-ended question which did not name specific herbicide s, followed by a serie s of questions in which the interviewer named specific herbicides, includ ing atraz ine. Users were further que stioned concerning the years of atrazine use, the average annual number of days of use on the farm , and the average annu al numb er of days that atrazine was personally handled (ie, mixed or applied). In Kansa s, the use of atrazine was assessed by an open -ended question without any prompting for specific herbicides. Duration of use, days per year of expo sure (no further definition of exposure was given the subjects), and other information related to exposure was obtained for all he rbicides as a group , but not for atrazine specifically. In Iowa-Minnesota, the use of atrazine was assessed by a direct question naming atrazine. Users were asked the first and last year of atrazine use, the method of application , whether the subje ct had personally applied , mixed , or handled atrazine, and wheth er protective equ ipment was used.

Risk measurement
The measure of association was the odds ratio (OR). After the subjects from the three studies were combined, risk estimates for atrazine use were adjust ed for the effects of age «45, 45-64, and >64 years) and state (four) by stratification. To adju st for agricultural exposures other than atrazine, the anal yses were restricted to farmers and the stratification factors were expanded to include the use of 2,4-dichlorophenoxyacetic acid (2,4-0) or organophosphate insecticides. Maximum-likelihood estimates of the overall risk and 95% confidence interval s (95% CI) were obtained with the EPITOME program (29). Because the methods for obtaining duration and other use chara cteristics varied in the three studies , we did not combine these data for analy sis, but have presented the relevant result s for Nebraska separately.

Results
When the data from the three studies are combined, 101 non-Hodgkin ' s lymphoma patients and 214 ref-110 erents reported that atrazine was used on the farms where they lived or worked, yielding an OR of 1.4 (95% CI 1.1-1.8) (table 2). The OR value s ranged from a low of 1.2 (95% CI 0.8-1.8) in Iowa to a high of 2.7 (95% CI 1.2-5.9) in Kansas . There was no chan ge in the OR (OR 1.4) when the analy ses were restricted to the 94 patients and 190 referents who reported that they had person ally handled atrazine. For two states and for all of the states combined, the risks were greater among the farmers who used atrazine in their farming operations but did not handle it than among farmers who personally handled atrazine.
The percentage of farmers reporting atrazine use varied by study and appeared to be relat ed to the method s used to ascertain pesticid e histor y. In Kansas, 6% of the referents volunteered a history of atrazine use in response to an open-ended question. In Iowa-Minnesota, 16% of the referents reported atrazine use after being asked about atrazine specifically. In Nebraska , 6 % of the referent s volunteered a histor y of use and an additional 7% reported use when asked directly about atrazine , for a total of 13% of the referents. The risk of non-Hodgkin ' s lymphoma among the Nebraska subjects who volunteered a history of atrazine use (OR 1.5, 14 cases, 33 referents) was similar to that of the subjects who responded to a dire ct question about atrazine (OR 1.3, 15 cases, 36 referents).
When all of the states were combined, atrazine use was associated with OR values of 1.3 and 2.0 amon g the living and deceased subjects, respectivel y. For all of the subjects, adju stment for a history of hematopoietic or lymphatic cancer in a next-of-kin caused little change in the OR for atrazine use (OR 1.5,95 % CI 1.1-1.9). There was little variation in the risk of non-H odgkin ' s lymphoma by histological type, althou gh diffus e non-Hodgkin ' s lymphoma demonstrated the stronges t association with atrazine use (OR 1.6,95% CI 1.1-2.2) (table 3).
In addition to atrazine, the farmers could have been expo sed to other herbicides and insecticides, some of which are suspected of increa sing the risk of non-Hodgkin ' s lymphoma (30 ). In parti cular, associ ations between non -Hodgkin ' s lymphoma and organophosphate insecticides (OR = Nebr aska 2.0, Kansas 1.8, Minnesota 1.5, Iowa 1.3, total 1.5) and the herbicide 2,4-0 (OR =Nebraska lA, Kansas 2.3, Minnesota I.I , Iowa 1.0, total 1.2) were noted . To adjust for these other agricultural exposures, the analyses were restricted to farmers . Among the farmers with non-Hodgkiu ' s lymphoma, 57% had never used atrazine , 2,4-0 or organophosphate insectic ides, 17% had used at least one of the expo sures of intere st, 14% had used two of the three, and 12% reported using all three. Among the farmer referent s, 70% had never used atrazine , 2,4-0 or organopho sphate insecticides, 15% had used at least one, 7% had used two of the three, and 7% reported using all three. When the farmers who used atrazine were compared with Table 2. Number 01non-Hodgkin 's lymphoma cases and relerents and the odds ratios (OR) and 95% confidence int ervals (95% GI) accord ing to atrazine use among white men in eastern Nebraska, Iowa, Minnesota , and Kansas.

Referents
Non  Table 3. Num ber of non-Hodg kin's lymphoma cases and referents and odds ratios (OR)and 95% confidence intervals (95% Gil by his tolog ic type accor ding to atrazine use among men in eastern Nebraska, Iowa and Minnesota, and Kansas. the farme rs who did not, the OR, adjusted for age and state, was 1.5 (95% CI 1.1-1.9) (table 4). The risks were about the same after the farmers who used atrazine in their farming operation but did not handle it personall y were eliminated. When the OR values were also adjusted for the use of 2,4-D and organophosphate insecticides, the OR was reduced to 1.2 (95% CI 0.9-1.7). Adjust ment for 2,4-D use re-duced the OR values in Nebraska and Kansas, while adjustment for organophosphate insecticide use reduced the OR values in Nebraska, Iowa, and Minnesota. Table 5 presents the risk of non-Hodgkiu' s lymphoma associated with the characteristics of atrazine use in Nebraska , the state with the most detailed information on the duration and annual frequency of  atrazine use spec ifically . Farmers who had used atrazine for more than 15 years had a twofold increased risk of non-Hodgkin's lymphoma, whereas no excess risk was seen for short-term users. When the analyses were restricted to farmers to allow for adjustment for other agricultural facto rs, the OR values for at-razine use increased slightly. Adj ustment for the use of 2,4-D and orga nophosphate insectici des, particularly the latter, eliminated the risks associated with long-term atrazine use. Odds ratios fell belo w unity for all duration and frequency categories except the ;:::2 1 d per year category , which was based on only one exposed patient and one exposed referent. Exami nation of the Nebraska data on the first year of use of atrazine revealed greater risks associated with shorter latency. After adju stment for other pesticide s, the OR values were 0.4 and 1.0 for use before 1966 and afte r 1965, respectively .

Discussion
By co mbining the data from three cas e-referent studies conducted in Nebraska, Io wa-Minnesot a, and Kansas, we were able to evaluate the effects of agricultural pra ctice s for 993 non-Hodgkin ' s lymphoma cas es in white males and for 2918 referent s. Thi s large number of subjects and the detailed natu re of the study que stionnaires allowed us to examine the risks for spec ific agr icultural exposur es with sufficient power to adj ust for other agricultura l and nonagric ultura l ex pos ures .
In the combined data set, atrazine use was associated with a significant 40% exc es s risk of non -Hodgkin ' s lymphoma. However, adjustm ent for the use of 2,4 -D and organophosph ate insecticides, pesticides associated with non-Hodgkin's lymphoma risk in previous analyses of the data from the contributing studies (16-1 8), reduced the atr azin e assoc iat ion for three of the four states . Th e mo st dramati c reduction in risk was in Nebraska, whe re the OR dropped from 1.7 to 0.7. Nebraska was also the only state with det ailed data on duration , frequency , and calendar yea rs of use for at razine. The ageadj usted data showed tw o-to threefold exce sses in risk of non-Hodgkin ' s lymphoma for lon g-t erm and frequent atrazine users. When adjusted for other pesticides, the OR values were below unity for all but one duration or frequency category. Adjustment increased the risk onl y in Io wa , from 1.2 to 1.6 (95% CI 0.9-2.9). In two states and in all states combined, the risk to farme rs who personally handled atrazine was lower than that to farmers who used atrazine in thei r farm ing operatio n but did not handl e it themselves, contrary to what would ha ve been ex pected if th is chemical were cau sally assoc iated with non-Hodgkin' s lymphoma. In our j udg ment, these dat a provide little eviden ce that atrazi ne is assoc iated with non -Hod gkin ' s lymphom a among white men.
It has been suggested (30) that pe sticid es, and herbicides in particular, may be playing a ro le in the large increase in the occurrence of non-Hodgkiri' s lymphom a that has take n place over the last 15 yea rs in the United States (3 1), particularly in the Midwes t (3 2), and in other developed countries (33). The factors res po nsible fo r this increase must be cau sall y associated with non-Hodgkin ' s lymphom a and mu st have an inc rea sed prevalence of exposure. Although the use of atrazi ne has increased over thi s peri od , the result s presented in thi s paper suggest that it is unlik ely that atrazi ne ex pos ure ex plai ns an y apprec iable amount of the ob served increase in non-Hodgkin's lymphoma.