Agricultural Herbicide Use and Risk of Lymphoma and Soft-Tissue Sarcoma Shelia K

Home , Alachlor

Agricultural Herbicide Use and Risk of Lymphoma and Soft-Tissue Sarcoma Shelia K. Hoar, ScD; Aaron Blair, PhD; Frederick F. Holmes, MD; Cathy D. Boysen; Robert J. Robel, PhD; Robert Hoover, MD; Joseph F. Fraumeni, Jr, MD

A population-based case-control study of soft-tissue sarcoma (STS), Hodgkin's agents for leather and textiles, and disease (HD), and non-Hodgkin's lymphoma (NHL) in Kansas found farm medications.1 For these reasons, a pop¬ herbicide use to be associated with NHL (odds ratio [OR], 1.6; 95% confidence ulation-based case-control study was interval [CI], 0.9, 2.6). Relative risk of NHL increased significantly with number conducted to clarify whether agricul¬ of of herbicide and Men to herbicides tural use of herbicides and insecticides days exposure per year latency. exposed and NHL in than 20 had sixfold increased risk of NHL 95% affects risk of STS, HD, more days per year a (OR, 6.0; the United States. CI, 1.9, 19.5) relative to nonfarmers. Frequent users who mixed or applied the herbicides themselves had an OR of 8.0 (95% CI, 2.3, 27.9) for NHL. Excesses METHODS were associated with use of phenoxyacetic acid herbicides, specifically Kansas, a major wheat-producing acid. Neither STS nor HD was associated with 2,4-dichlorophenoxyacetic state, was chosen as the location for pesticide exposure. This study confirms the reports from Sweden and several the study, since herbicides have been US NHL states that is associated with farm herbicide use, especially used more frequently than other pesti¬ phenoxyacetic acids. It does not confirm the case-control studies or the cohort cides on wheat. 2,4-Dichlorophenoxy- studies of pesticide manufacturers and Vietnam veterans linking herbicides to acetic acid (2,4-D) has been the most STS or HD. commonly used herbicide in Kansas; (JAMA 1986;256:1141-1147) substantial amounts of 2,4,5-trichlorophenoxyacetic acid (2,4,5-T) have also been used, along with myriad other EPIDEMIC-LOGIC studies from Swe¬ benzofurans. There have also been sev¬ chemicals.11 den have suggested that workers eral reports of increased STS and NHL exposed to phenoxyacetic acid herbi¬ among workers producing phenoxy¬ Cases cides and are at excess acetic acid herbicides4"7 and chlorophenols among All newly diagnosed cases of STS, risk of soft-tissue sarcoma farmers.8"10 (STS), HD, and NHL among white male Kan¬ and non- Hodgkin's disease (HD), sas 21 or For all residents, aged years older, Hodgkin's lymphoma (NHL).13 See also p 1176. from 1976 through 1982, were identi¬ three cancers, risks were increased fied through the University of Kansas fivefold to of wheth¬ sixfold, regardless Concern over possible carcinogenic (Kansas City) Cancer Data Service, a er exposures were contaminated by risks from phenoxyacetic acids and population-based registry covering the dibenzodioxins or di- polychlorinated chlorophenols is heightened by the state of Kansas. Cancer reporting, potential for widespread exposure. In mandated by Kansas law, is considered From the Epidemiology and Biostatistics Program, addition to herbicide formulations used complete, as evidenced by a higher National Cancer Institute, Bethesda, Md (Drs Hoar, Blair, in and in the Vietnam war, Hoover, and the Cancer Data Service, agriculture annual incidence rate for STS (4.1/ Fraumeni); occur in blue stain University of Kansas Medical Center, Kansas City (Dr these chemicals 100 000) than reported by the nearby Holmes and Ms Boysen); and the Division of Biology, retardants used in sawmills, slime con¬ Iowa National Cancer Institute-spon¬ Kansas State University, Manhattan (Dr Robel). trol preparations in paper and pulp sored Reprint requests to Epidemiology and Biostatistics Surveillance, Epidemiology, and Program, National Cancer Institute, Landow 4C16, manufacturing, cutting oils and fluids, End Results registry (3.4/100 000). Bethesda, MD 20892 (Dr Hoar). wood preservatives, waterproofing There were 200 men diagnosed with

Downloaded from jama.ama-assn.org at University of Waterloo on January 13, 2012 Table 1.—Soft-Tissue Sarcoma, Hodgkin's Disease, and Non-Hodgkin's Lymphoma in Relation to Duration and Frequency of Herbicide Use

Soft-Tissue Sarcoma Hodgkin's Disease Non-Hodgkin's Lymphoma Odds Ratio Odds Ratio Odds Ratio No. of No. of (95% Confi- No. of (95% Confi- No. of (95% Confi- Controls Cases dence Interval) Cases dence Interval) Cases dence Interval) Nonfarmers 286 38 1.0 50 1.0 37 1.0 Farmers 662 95 1.0 (0.7, 1.6) 71 0.8 (0.5, 1.2) 133 1.4 (0.9, 2.1) Duration of herbicide use, y 0 470 73 1.1 (0.7, 1.8) 43 0.7 (0.5, 1.2) 93 1.3 (0.8, 2.1) 1-5 53 0.7 (0.2, 2.0) 0.9 (0.4, 2.2) 9 1.3 (0.6, 3.1) 6-15 58 0.7 (0.2, 1.9) 0.8 (0.3, 1.9) 12 1.6 (0.7, 3.4) :16 57 11 1.4 (0.6, 3.1) 10 1.2 (0.5, 2.6) 16 2.0 (1.0, 4.0) Unknown 24

X for trend 0.13 0.11 2.06 P (one-tailed) .45 .46 .02 Frequency of herbicide use, d/y 0 497 74 1.1 (0.7, 1.7) 45 0.7 (0.4, 1.1) 94 1.3 (0.8, 2.0) 1-5 102 0.6 (0.3, 1.5) 16 1.0 (0.5, 1.9) 19 1.4 (0.7, 2.6) 6-10 29 0.5 (0.1, 2.3) 0.4 (0.1, 1.7) 1.6 (0.5, 4.3) 11-20 13 2.9 (0.8, 9.5) 1.3 (0.2, 6.8) 2.6 (0.8, 8.8) ;21 12 0.8 (0.04, 6.5) 1.0 (0.2, 4.1) 6.0 (1.9, 19.5) Unknown

X for trend 0.002 -0.56 3.33 P (one-tailed) .50 .29 .0004

STS and 173 men with HD. A random 92.3% of the working residential tele¬ farm locations and sizes, herbicides sample of 200 men was drawn from the phone numbers called, a person an¬ and insecticides used, years and acres 297 men diagnosed with NHL from swered who was willing to provide the treated, names and locations of compa¬ 1979 through 1981. names and ages of household members nies where pesticides were purchased, Pathology specimens for 87% of the aged 21 to 64 years and, if someone in method of application, days per year cases were reviewed by a panel of three the household was selected as a control, exposed, and use of protective equip¬ pathologists to confirm the diagnoses the household address. For deceased ment. Information on other crops was and to standardize the subgroup termi¬ patients, the controls were selected not obtained; however, in 1978, these nology.12,13 Specimens for the remaining from Kansas state mortality files, with four crops constituted 94% of Kansas cases either could not be obtained the additional matching factor of year farm acreage and 87% of acres treated (11%) or were of poor or insufficient of death. Persons with a cause of death with herbicides.1116 All but four sub¬ quantity to allow review (2%). For of STS, HD, NHL, a malignancy of an jects who lived or worked on farmland STS, HD, and NHL cases, the confir¬ ill-defined site (International Classifi¬ as adults grew at least one of the four mation rates were 81%, 85%, and 90%, cation of Diseases code 195), homicide, specified crops. respectively. Among the eligible histo- or suicide were excluded. One half of Interviews were obtained from 133 logically confirmed cases, 139 were the patients with STS and NHL died patients with STS, 121 with HD, 170 STS, 132 HD, and 172 NHL. before the initiation of the study, while with NHL, and 948 controls, which only one third of the patients with HD represented 95% of the eligible sub¬ Controls had died. The next of kin were inter¬ jects (patients, 96%; controls, 94%). The controls were white men from viewed for deceased subjects. The same The overall response rate, a weighted the general population of Kansas. controls were used for comparisons average accounting for the initial Three controls (N=1005) were matched with the three different cancer series refusals in the random digit-dialing to each patient on age ( ± 2 years) and when of comparable age. control selection process, was 93%. vital status. For living patients, con¬ trols aged 65 years or older were Interview selected from the Health Care Financ¬ The patients and controls, or their Pesticide Suppliers' Survey ing Administration file (Medicare), next of kin, were interviewed by tele¬ To evaluate potential observation whereas controls aged 64 years or phone between December 1982 and bias, corroborative evidence of the self- younger were selected by telephone, January 1984. The questions on farm¬ reported pesticide exposures was using a two-staged random digit-dial¬ ing practices covered the calendar sought for a sample of 130 subjects ing technique.14 Control selection was years working or living on farmland with farming experience. Suppliers for not biased by differences in telephone during which any of four specific crops 110 subjects were located and provided coverage for urban (91%) and rural (wheat, corn, sorghum, or pasture) information on the subject's crops and (90% ) white Kansan households.15 For were grown or livestock was raised, the herbicide and insecticide purchases.

Downloaded from jama.ama-assn.org at University of Waterloo on January 13, 2012 in Relation Herbicide Table 2.—Non-Hodgkin's Lymphoma to Group risk of STS or HD were seen with either duration or of herbi¬ No. of No. of Odds Ratio (95% frequency Herbicide Group Cases Controls Confidence Interval) cide use. More detailed analyses showed no association between agricul¬ Never farmed 37 286 1.0 tural factors and the occurrence of STS Ever use* or HD, so the remainder of this report Phenoxyacetic acids 24 78 2.2 (1.2, 4.1) will describe only the NHL results. Triazines 14 43 2.5 (1.2, 5.4) Subjects who reported usually mix¬ ing or applying the herbicides them¬ Amides 22 2.9 (1.1, 7.6) selves (OR, 1.9; 95% CI, 1.1, 3.3) had Benzoics 4.0 (0.1, 62.6) higher risks for NHL than those who that someone else Carbamates 5.6 (0.6, 45.9) reported performed these functions (OR, 1.1). Indeed, the Trifluralin 12.5 (1.6, 116.1) trends in the OR with increasing fre¬ Uracils 19 114 1.3 (0.7, 2.5) quency and duration of use derived mainly from the workers who mixed or Nonspecified 11 5.8 (1.9, 17.2) applied the herbicides themselves. For Hierarchical uset example, men who mixed or applied Phenoxyacetic acids 24 78 2.2 (1.2, 4.1) the herbicides and who were exposed for one to six to 11 to and Uracils; no phenoxyacetic acids 13 101 1.0 (0.5, 2.1) five, ten, 20, more than 20 days per year had ORs of Triazines; no phenoxyacetic 1.4 1.5 acids or uracils 11 2.2 (0.4, 9.1) (95% CI, 0.7, 3.0), (95% CI, 0.5, 4.6), 1.8 (95% CI, 0.4, 7.4), and 8.0 (95% Amides; no phenoxyacetic acids, CI, 2.3, 27.9; seven patients, nine con¬ triazines uracils, or trols). Farmers who did not use protec¬ tive such as rubber 'Farmers could report more than one group. equipment, gloves or had a OR associated tFarmers were assigned to only one group. masks, higher with herbicide use (OR, 2.1; 95% CI, 1.0, 4.2) than those who protected them¬ selves (OR), 1.5; 95% CI, 0.7, 3.1). Risk Measurements ghum, or pasture) was reported by 22 Higher risks for herbicide use were The measure of association between patients with STS, 28 with HD, and 40 also seen among farmers who used pesticide exposures and risk of cancer with NHL, compared with 192 controls, backpack or hand sprayers (OR, 2.3; was the odds ratio (OR). All estimates yielding ORs of 0.9 (95% CI, 0.5, 1.6), 95% CI, 1.0,5.2), an application method were adjusted for the effects of age by 0.9 (95% CI, 0.5, 1.5), and 1.6 (95% CI, with greater potential for personal stratification. Maximum likelihood es¬ 0.9, 2.6), respectively. There was a exposure than other methods.20 After timates of the overall risk and the 95% significant trend (P=.02) in risk of excluding persons who used backpack confidence interval (CI) were computed NHL with increasing years of herbicide or hand sprayers, the ORs for tractor- by Gart's method.17 Matched analyses use and with number of days of herbi¬ mounted or mist-blower spraying and yielded results similar to those pro¬ cide exposure per year (P=.0004) (Ta¬ aerial application of herbicides were vided by the unmatched approach. The ble 1). Persons exposed to herbicides both 1.5. Too few subjects remained for unmatched analyses are presented more than 20 days per year had an OR evaluation of the group that applied since they allowed control of additional of 6.0 (95% CI, 1.9,19.5). There was no herbicides only in the soil. factors that were not matching factors association with years of herbicide use A logistic regression analysis was and increased power by pooling con¬ after adjustment for annual days. On performed to examine risk accounting trols for each case series. For duration- the other hand, adjustment for years of for all pesticide exposure variables response relationships, significance herbicide use changed the OR for the simultaneously. The results showed was assessed by means of Mantel's five patients and six controls exposed that age and annual days of herbicide one-tailed linear trend test.18 Logistic more than 20 days per year from 6.0 to exposure were significantly related to regression was also performed to eval¬ 7.4 (95% CI, 1.6, 38.9), relative to the NHL risk. Restricting the analysis to users. The risk of NHL uate simultaneously several compo¬ least exposed persons exposed to herbicides and nents of pesticide exposure.19 associated with herbicide use did not using categorical variables yielded ORs change significantly when restricted to of 1.1 (95% CI, 0.3,4.0), 1.3 (95% CI, 0.3, RESULTS exposures incurred before 1976, the 6.0), and 10.3 (95% CI, 2.1, 49.5) for Ninety-five patients with STS, 71 earliest possible date of death for the persons exposed to herbicides six to with HD, and 133 with NHL reported pooled control series. The NHL risk ten, 11 to 20, and more than 20 days per having worked or lived on farmland, also rose with increasing time since year, respectively, relative to persons compared with 662 controls, yielding first exposure. Farmers who started exposed one to five days per year. ORs of 1.0 (95% CI, 0.7, 1.6), 0.8 (95% using herbicides after 1965, from 1956 Increased risk was also associated with CI, 0.5, 1.2), and 1.4 (95% CI, 0.9, 2.1), through 1965, 1946 through 1955, and use of a backpack or hand sprayer (OR, respectively (Table 1). No trend with before 1946 had ORs of 1.3,1.7,1.7, and 1.3; 95% CI, 0.5, 3.7) and failure to use years spent working or living on a farm 3.3, respectively. This trend was dimin¬ protective equipment (OR, 2.0; 95% CI, was observed. Risks did not vary by ished by controlling for frequency of 0.8, 5.1). Total period of herbicide use crop or farm acreage. herbicide use, but farmers who began was not associated with NHL when the use before 1946 still had an excess risk other pesticide exposure variables were Herbicides (OR, 2.2). No association was seen with controlled. The effect of personally Farm herbicide use on any of the number of acres treated with herbi¬ mixing or applying herbicides could not four specific crops (wheat, corn, sor- cides. No consistent patterns of excess be evaluated; the overwhelming major-

Downloaded from jama.ama-assn.org at University of Waterloo on January 13, 2012 Table 3.—Non-Hodgkin's Lymphoma in Relation to Duration, Frequency, and Latency of ity of subjects involved in the logistic Acid Use 2,4-Dichlorophenoxyacetic analysis either mixed or applied the herbicides themselves. No. of No. of Odds Ratio (95% Controls Confidence Interval) Significant excesses were associated with ever use of phenoxyacetic acids, Never farmed 37 286 1.0 triazines (eg, atrazine, cyanazine, me- Duration of use, * y tribuzin, prometone, propazine, terbu- 1-5 16 1.3 (0.3, 5.1) tryn), amides (eg, alachlor, propach- 6-15 22 2.5 (0.9, 6.8) lor), trifluralin, and nonspecified herbi¬ cides, such as "liquids," "sprays," and 16-25 15 3.9 (1.4, 10.9) "dusts" (Table 2). Most farmers re¬ >26 17 2.3 (0.7, 6.8) ported use of chemicals in several of the herbicide subgroups. Since the a X for trend 3.560 priori hypotheses dealt with phenoxy¬ P (one-tailed) .0002 acetic acids, we assessed risks associ¬ ated with herbicides ranked in a hier¬ Frequency of use.t d/y 1-2 17 2.7 (0.9, 8.1) archical manner (Table 2). In the absence of phenoxyacetic acid expo¬ 3-5 16 1.6 (0.4, 5.7) sure, the NHL risk associated with 6-10 16 1.9 (0.5, 6.7) triazine exposure was reduced to 1.9 11-20 3.0 (0.7, 11.8) (95% CI, 0.4, 8.0) and the risk with uracil herbicides (eg, bromacil, terba- >21 7.6 (1.8, 32.3) cil) was reduced to 1.0 (95% CI, 0.5, X for trend 3.733 2.1). In this acid P (one-tailed) .0001 study, phenoxyacetic herbicide use was essentially synony¬ First year of uset mous with use of 2,4-D (OR, 2.3; 95% 1966 or later 21 1.9 (0.6, 6.0) CI, 1.3, 4.3), since only three patients 1956-1965 23 2.9 (1.1, 7.2) and 18 controls had used 2,4,5-T, and all but two of these controls had also 1946-1955 24 2.1 (0.8, 5.6) used 2,4-D. Use of 2,4-D only, ie, elimi¬ Before 1946 6.2 (0.6, 65.3) nating 2,4,5-T users, was associated with an of 2.6 X for trend 3.561 OR (95% CI, 1.4, 5.0). There were significant, although incon¬ P (one-tailed) .0002 sistent, increases in NHL risk in rela¬ tion to the and 'Five controls had missing data. duration, frequency, of use The tOne patient and ten controls had unknown frequency of exposure. latency 2,4-D (Table 3). more $First available for use in 1942. 2,4-D users exposed to herbicides

Table 4.—Non-Hodgkin's Lymphoma in Farmers in Relation to Use of Herbicides and Insecticides

Never Used Insecticides Ever Used Insecticides Adjusted for Insecticide Use Odds Ratio' Odds Ratio Odds Ratio No. of No. of (95% Confi- No. of No. of (95% Confi¬ No. of No. of (95% Confi¬ Annual Days of Use Controls dence Interval) Controls dence Interval) Controls dence Interval) Herbicides 0 90 474 1.0 21 1.1 (0.3, 3.5) 94 495 1.0 1-5 11 57 1.1 (0.5, 2.3) 43 1.2 (0.5, 2.9) 19 100 1.1 (0.6, 2.1) >6 18 2.1 (0.7, 5.9) 12 35 2.3 (1.0, 4.9) 18 53 2.1 (0.9, 4.9) Insecticide use, adjusted for herbicide use 1.1 (0.6, 2.2)

Never Used Herbicides Ever Used Herbicides Adjusted for Herbicide Use Odds Ratio' Odds Ratio* Odds Ratio' No. of No. of (95% Confi- No. of No. of (95% Confi¬ No. of No. of (95% Confi¬ Cases Controls dence Interval) Cases Controls dence Interval) Cases Controls dence Interval) Insecticides 0 90 474 1.0 17 75 1.3 (0.7, 2.4) 107 549 1.0 1-2 12 0.9 (0.1, 4.2) 30 1.7 (0.7, 4.1) 10 42 1.2 (0.5, 2.8) 1.5 (0.2, 8.1) 12 48 1.7 (0.8, 3.5) 14 57 1.4 (0.6, 3.1) Herbicide use, adjusted for insecticide use 1.4 (0.8, 2.4)

'Odds ratio relative to farmers, no herbicide or insecticide use.

Downloaded from jama.ama-assn.org at University of Waterloo on January 13, 2012 than 20 days per year had an OR of 7.6 insecticide use from 1.5 to 1.1 (95% CI, and yards was not associated with (95% CI, 1.8, 32.3). However, these 0.6, 2.2). Similar analyses based on NHL. The OR associated with ever variables cannot be determined with years of exposure also suggested that smoking at least 100 cigarettes was complete accuracy because the ques¬ risk was more strongly associated with slightly below 1 (OR, 0.7; 95% CI, 0.5, tionnaire elicited dates and frequency exposure to herbicides than to insecti¬ 1.0), as it was for lifetime consumption of use of any herbicide on each farm cides. The number of subjects was too of at least 100 cups of coffee (OR, 0.8; instead of dates and frequency for each small for a more detailed stratified 95% CI, 0.5, 1.4). Consumption of raw, specific herbicide. Therefore, the actual analysis of herbicide or insecticide use. unpasteurized milk products had no years or days of 2,4-D use may be less In a logistic analysis of all pesticide effect on NHL risk (OR, 1.1; 95% CI, than that reported for all herbicides on variables, insecticide exposure was not 0.8, 1.6). Eight patients with NHL had a farm. found to be significantly related to risk. diabetes, half the expected number Risk associated with herbicide expo¬ The important variables remained age (OR, 0.5; 95% CI, 0.2, 1.2). No subjects sure was also examined by subgroups and annual days of herbicide expo¬ had systemic lupus erythematosus, of NHL. There were no differences in sure. celiac disease, or immunodeficiency the ORs associated with herbicide use The elevation in NHL risk (OR, 1.3) syndromes or had received immuno- when cases were categorized by histo¬ seen among farmers who did not report suppressive drugs. Seven patients with logie types (follicular, diffuse, and not herbicide use may be due to other NHL reported previous radiation treat¬ specified) and by grade (prognostic exposures encountered in farming ac¬ ment (OR, 0.9; 95% CI, 0.4, 2.2). Sub¬ groupings of low, intermediate, and tivities or to biased misclassification of jects reporting a family history of high), as classified by the National exposure status. However, it is not cancer had a significant risk of NHL Cancer Institute Working Formula¬ likely that more patients than controls (OR, 2.3; 95% CI, 1.6, 32). Three tion.13 In addition, no significant differ¬ underreported herbicide exposure. patients and four controls reported a ences in risk for any tumor were Slightly fewer "nonexposed" farming relative with lymphoma (OR, 4.0; 95% associated with either age at diagnosis patients with NHL (52% ) than controls CI, 0.7, 22.2). or-vital status at interview. Deceased (55%) were deceased, with interviews subjects had just slightly lower risks of supplied by next of kin who might COMMENT NHL associated with overall herbicide underreport exposures. Approximately This investigation confirms the re¬ exposure (OR, 1.5; 95% CI, 0.7, 3.3) or the same percentage of "nonexposed" sults of the case-control study by Har- long-term exposure (OR, 5.5; 95% CI, farming patients (4%) and controls dell et al3 that initially suggested an 0.7, 41.3) than did living subjects (OR, (5% ) were usually employed in agricul¬ association between herbicide use and 1.7; 95% CI, 0.8, 3.7; and OR, 5.6; 95% ture. NHL. Our finding of a sixfold increase CI, 1.1, 28.9; respectively). of NHL among farmers exposed to Fungicides herbicides more than 20 days per year Insecticides Thirty-two patients with NHL and is consistent with the sixfold excess Insecticide use on crops or animals 105 controls treated seeds with fungi¬ risk associated with exposure to either was reported by 54 patients with NHL cides (OR, 2.1; 95% CI, 1.2, 3.7). Risk phenoxyacetic acids or chlorophenols and 275 controls, yielding an OR of 1.5 was elevated among both herbicide in the Swedish study. In both Sweden (95% CI, 0.9,2.4). There was no associa¬ users (22 patients, 68 controls; OR, 2.3; and Kansas, risk was elevated among tion with increasing years of insecti¬ 95% CI, 1.2, 4.3) and farmers who had persons exposed to phenoxyacetic cide use. Risk increased significantly never used herbicides (ten patients, 37 acids, eg, 2,4-D, not likely to be contam¬ but inconsistently with days of expo¬ controls; OR, 1.9; 95% CI, 0.8, 4.4). No inated by dioxins.21,22 These findings sure per year. Men exposed to insecti¬ other information on fungicide use was are consistent also with a number of cides more than six days per year had a collected to allow further evaluation of descriptive studies suggesting in¬ 2.8-fold increased NHL risk (95% CI, this association. creased lymphoma risk among agricul¬ 1.2, 6.5). Farmers who started using tural workers, particularly in regions insecticides after 1965, from 1956 Pesticide Suppliers' Data where herbicide use is common.8"10,23'24 through 1965, 1946 through 1955, and Suppliers usually reported less pesti¬ Cohort studies of pesticide manufac¬ before 1946 had ORs of 1.5, 0.7,1.5, and cide use than subjects. Agreement on turing workers and applicators have in 1.7, respectively. No association was specific years of exposure was better general been too small to examine seen with number of acres treated with for insecticide use than herbicide use. adequately the risk of NHL. Danish insecticides. Other exposure variables, There were no consistent differences workers manufacturing 2 methyl-4 such as mixing and applying insecti¬ between agreement rates for patients chlorophenoxyacetic acid had seven cides, application method, and insecti¬ and controls. Multiplying the number NHL deaths with 5.4 expected, a non¬ cide type, showed little or no associa¬ of patients and controls who reported significant 30% excess.25 No NHL cases tion with NHL risk. any herbicide use by the percentage of have been observed in five other cohort verified herbicide use yielded a recalcu¬ studies of exposed workers, but the Herbicides vs Insecticides lated OR of 1.8 for NHL, which was total number of workers involved was Table 4 shows that NHL risk slightly higher than the OR based on only 2705.26"29 Also, the Vietnam veter¬ increased with annual days of herbi¬ interview data only (1.6). Agreement an experience offers little evidence to cide exposure, but was not further between suppliers and subjects im¬ date for an association between NHL increased by simultaneous use of insec¬ proved when pesticide use during the and herbicide exposure. The mortality ticides. Adjustment for insecticide ex¬ last ten years was considered. study of 1247 men who applied herbi¬ posure did not alter the ORs for herbi¬ cides in Project Ranch Hand reported cide use. On the other hand, risk Nonfarming Exposures no deaths due to lymphoma, as of increased only slightly with annual Nonfarming exposures did not con¬ December 1982,30 but less than one such days of insecticide exposure within found the association between NHL death was expected in this small strata of herbicide use. Adjustment for and agricultural use of herbicides. Non- cohort. herbicide use decreased the OR for farming pesticide use in home gardens An unusual presentation of NHL of

Downloaded from jama.ama-assn.org at University of Waterloo on January 13, 2012 the scalp was reported in two men of ers of the contaminants was not avail¬ nities for more than one pesticide 158 who developed chloracne while able in the early time period.21 supplier per subject. The evidence sug¬ employed in a British plant manufac¬ The Swedish case-control studies and gests that the errors introduced by turing pentachlorophenol.31 The ex¬ several of the industrial cohort studies these factors are likely to be similar for pected number of cases of cutaneous have been criticized on a variety of cases and controls. This random mis- NHL was far less. A similar associa¬ grounds, including possible inaccurate classification of exposure would tend to tion of cutaneous NHL with occupa¬ diagnoses, observation and/or recall dilute risk estimates, rather than pro¬ tions involving spraying phenoxyacetic bias, and lack of control for confound¬ duce spurious associations. While this acid herbicides was found in Sweden.32 ing variables.5,46 The current study was may have led us to miss an association In our study, only one patient had designed to address a number of these between pesticide use and STS or HD, cutaneous NHL, but he reported no concerns. The cases were drawn from it is unlikely to have created the herbicide exposure. all incident cases in a defined popula¬ association with NHL. If inaccurate An association between phenoxy¬ tion. Included for analysis were only exposure assessment occurred, we acetic acid herbicides and NHL may cases histologically confirmed as NHL would have underestimated the magni¬ have biologic plausibility through the by an expert review panel. The tude of the risks involved. If the risks relationships between dioxin contami¬ response rate for both cases and con¬ reported are accurate, and if they nants and the immune system, as trols was high. The relationship be¬ reflect a true causal relationship, then postulated by Hardell et al.3 Dioxin tween NHL and herbicide use was the amount of NHL in the current suppresses cell-mediated immunity33 specific: there was no relationship study attributable to herbicide expo¬ and causes thymic involution in labora¬ between such use and the two other sure would be 11%.M tory animals,34 while NHL occurs cancer sites studied, and the associa¬ Despite the limitations of the expo¬ excessively in various immunodeficien¬ tion was with herbicide and not insecti¬ sure information, the sixfold excess cy states of humans.35"37 Dioxin is also a cide use. Indeed, the association was risk of NHL among high-intensity potent animal carcinogen, with the risk limited to two subcategories of herbi¬ users of herbicides is cause for concern, seen for squamous cell carcinomas of cides, phenoxyacetic acids and tria¬ particularly since the association was the lung, hard palate/nasal turbinâtes/ zines. This degree of specificity argues mainly for phenoxyacetic acids and tongue, hepatocellular carcinomas, fol- against a significant role for either was apparent using several different licular cell thyroid adenomas, and skin observation or recall bias in the asso¬ measures of exposure. Since over 42 fibrosarcomas.38"43 However, it should ciation. In addition, independent as¬ million pounds of phenoxyacetic acid be noted that 2,4-D, the herbicide most sessments of herbicide exposure (as herbicides were applied to US farm¬ frequently used by subjects in this reported by self or next of kin, and by lands in 1976,59 the carcinogenic effects study, has not been shown to be carcin¬ pesticide supplier) yielded similar esti¬ suggested by this study and others ogenic in animals44 or immunosuppres- mates of risk. Finally, while the origins have important public health implica¬ sive21,45 (S. Wong, PhD, written commu¬ of NHL in the general population are tions. nication, Dec 12, 1985). 2,4-Dichloro- largely unknown, some known risk phenoxyacetic acid does not contain factors (eg, immune-altering conditions 2,3,7,8-tetrachlorodibenzo-p-dioxin, the and drugs, family history)35,37,47"54 as The study was supported in part by contract most carcinogenic dioxin isomer, al¬ well as several speculative factors (eg, N01-CP-ES-11027 from National Cancer Institute. less toxic isomers coffee We thank Ariel B. Baker and Naomi E. Wash- though other may cigarette smoking, consumption, burn for assisting the study manager; Rita King- occur.21,22 The increased risk for ionizing radiation)5,52,55"57 were assessed ma, Lois I. Murphy, and Jeanine M. Peterson for farmers first 2,4-D before 1946 and found not to confound the herbi¬ interviewing; Vesta Gee and Claudia J. Oblak for using Hobbs and Melanie Wall for may indicate the presence of carcino¬ cide association. coding; Shelley formula¬ Several circumstances secretarial assistance; Joan L. Van Nice, Pros genic impurities in the early offered oppor¬ Bunag, and Dolores Wiesmann for identifying and tions, with subsequent improvements tunities for inaccuracies in the assess¬ locating study subjects; Fritz Lin, MD, and H. in the manufacturing process. Alterna¬ ment of pesticide exposures. These Clarke Anderson, MD, for reviewing pathology Edward lia tively, the high risk of NHL among include the memories of subjects, the samples; Brown, Dave, and Ko Jam of relevant Liu for computer programming; and Richard H. persons exposed 40 years ago may knowledge practices by next Adamson, PhD, Susan M. Sieber, PhD, and Eliza¬ reflect a long latency period. The tech¬ of kin, the vagueness of some questions beth K. Weisburger, PhD, for reviewing the nology necessary to identify the isom- concerning exposure, and the opportu- manuscript.

References 1. Hardell L, Sandstrom A: Case-control study: cet 1981;1:1370. fication Project: NCI sponsored study of classifica- Soft-tissue sarcomas and exposure to phenoxy- 7. Johnson FE, Kugler MA, Brown SM: Soft-tissue tion of non-Hodgkin's lymphomas: Summary and acetic acids or chlorophenols. Br J Cancer 1979; sarcomas and chlorinated phenols. Lancet 1981; description of a working formulation for clinical 39:711-717. 2:40. usage. Cancer 1982;49:2112-2135. 2. Eriksson M, Hardell L, Berg NO, et al: Soft\x=req-\ 8. Cantor KP: Farming and mortality from non\x=req-\ 14. Waksberg J: Sampling methods for random tissue sarcomas and exposure to chemical sub- Hodgkin's lymphoma: A case-control study. Int J digit dialing. J Am Stat Assoc 1978;73:40-46. stances: A case-referent study. Br J Ind Med Cancer 1982;29:239-247. 15. Census of the Population: 1970, vol 1, General 1981;38:27-33. 9. Burmeister LF, Everett GD, Van Lier SF, et al: Housing Characteristics, pt 18, Kansas. US Bureau 3. Hardell L, Eriksson M, Lenner P, et al: Malig- Selected cancer mortality and farm practices in of the Census, 1973. nant lymphoma and exposure to chemicals espe- Iowa. Am J Epidemiol 1983;118:72-77. 16. Kansas Crop and Livestock Reporting Service: cially organic solvents, chlorophenols and phenoxy 10. Buesching DP, Wollstadt L: Cancer mortality Pesticides: 1978 Pesticide Usage. Topeka, Kan, US acids: A case-control study. Br J Cancer 1981; among farmers. JNCI 1984;72:503. Dept of Agriculture, Economics, Statistics, and 43:169-176. 11. Waldron AC, Park EL: Pesticide Use on Major Cooperatives Service. 4. Honchar PA, Halperin WA: 2,4,5-T,trichloro- Crops in the North Central Region 1978, 17. Gart JJ: Point and interval estimation of the phenol, and soft-tissue sarcoma. Lancet 1981; research bulletin 1132, Wooster, Ohio, Ohio. . . Agri- common odds ratio in the combination of 2X2 1:268-269. cultural Research and Development Center, 1981. tables with fixed marginals. Biometrika 1970; 5. Cook RR: Dioxin, chloracne, and soft-tissue 12. International Classification of Disease for 57:471-475. sarcoma. Lancet 1981;1:618-619. Oncology. Geneva, World Health Organization, 18. Mantel N: \g=x\2 test with 1 degree of freedom, 6. Moses M, Selikoff IJ: Soft-tissue sarcomas, 1976. extension of the Mantel-Haenzel procedure. Am phenoxy herbicides, and chlorinated phenols. Lan- 13. Non-Hodgkin's Lymphoma Pathologic Classi- Stat Assoc J 1963;58:690-700.

Downloaded from jama.ama-assn.org at University of Waterloo on January 13, 2012 19. Engelman L: Stepwise Logistic Regression in phoma of the scalp in workers exposed to dioxins. Brooks AFB, Tex, US Air Force Occupational and BMDP STATISTICAL SOFTWARE, Dixon WJ Lancet 1981;2:369. Environmental Health Laboratory, 1978. (ed). Berkeley, University of California Press, 32. Olsson H, Brandt L: Non-Hodgkin's lympho- 45. Blakely BR, Schiefer BH: The effect of topi- 1983, pp 330-344. ma of the skin and occupational exposure to cally applied n-butylester of 2,4-dichlorophenoxy- 20. Libich S, To JC, Frank R, et al: Occupational herbicides. Lancet 1981;2:579. acetic acid on the immune response in mice. J Appl exposure of herbicide applicators to herbicides 33. Vos JG, Moore JA, Zinkl JG: Effect of 2,3,7,8\x=req-\ Toxicol, in press. used along electric power transmission line right\x=req-\ tetrachlorodibenzo-p-dioxin on the immune sys- 46. Cole P: Direct testimony before the Adminis- of-way. Am Ind Hyg Assoc J 1984;45:56-62. tem of laboratory animals. Environ Health Per- trator of the Environmental Protection Agency of 21. International Programme on Chemical Safety: spect 1973;5:149-162. the United States of America in re. The Dow Environmental Health Criteria 29: 2,4-Dichloro- 34. Harris MW, Moore JA, Vos JG, et al: General Chemical Company, et al. FIFRA Docket No. 415, phenoxyacetic Acid (2,4-D). Geneva, World Health biological effects of TCDD in laboratory animals. et al, Washington, DC, Oct 23,1980. Organization, 1984. Environ Health Perspect 1973;5:101-109. 47. Li FP, Fraumeni JF Jr, Mantel N, et al: Cancer 22. Cochrane WP, Singh J, Miles W, et al: Deter- 35. Hoover R, Fraumeni JF Jr: Risk of cancer in mortality among chemists. JNCI 1969;43:1159\x=req-\ mination of chlorinated dibenzo-p-dioxin contami- renal transplant recipients. Lancet 1973;2:55-57. 1164. nants in 2,4-D products by gas chromatography- 36. Kinlen LJ, Sheil AGR, Peto J, et al: Collabora- 48. Olin GR: The hazards of a chemical laboratory mass spectrometric techniques. J Chromatogr tive United Kingdom-Australasian study of can- environment: A study of the mortality in two 1981;217:289-299. cer in patients treated with immunosuppressive cohorts of Swedish chemists. Am Ind Hyg Assoc J 23. Stubbs HA, Harris J, Spear RC: A propor- drugs. Br Med J 1979;1461-1466. 1978;39:557-562. tionate mortality analysis of California agricultur- 37. Filipovich AH, Spector BD, Kersey J: Immu- 49. Searle CE: Epidemiologic study of the mortali- al workers, 1978-1979. Am J Ind Med 1984;6:305\x=req-\ nodeficiency in humans as a risk factor in the ty of British chemists. Br J Cancer 1978;38:192\x=req-\ 320. development of malignancy. Prev Med 1980;9:252\x=req-\ 193. 24. Pearce NE, Smith AH, Fisher DO: Malignant 259. 50. Monson RR, Fine LJ: Cancer mortality and lymphoma and multiple myeloma linked with 38. Crow KD: Soft tissue sarcomas and chlori- morbidity among rubber workers. JNCI 1978; agricultural occupations in a New Zealand cancer nated phenols. Lancet 1981;2:369. 61:1047-1053. registry-based study. Am J Epidemiol 1985; 39. Kociba RJ, Keyes DG, Beyes JE, et al: Results 51. Cantor KP, Fraumeni JF Jr Distribution of 121:225-237. of a two-year chronic toxicity and oncogenicity non-Hodgkin's lymphoma in the United States 25. Lynge E: A follow-up study of cancer inci- study of 2,3,7,8-tetrachlorodibenzo-p-dioxin between 1950 and 1975. Cancer Res 1980;40:2645- dence among workers in manufacture of phenoxy (TCDD) in rats. Toxicol Appl Pharmacol 1978; 2652. herbicides in Denmark. Br J Cancer 1985;52:259\x=req-\ 46:279-303. 52. Kalant N, Seemayer T: Malignant lymphoma 270. 40. Kociba RJ, Keyes DG, Beyer JE, et al: Long\x=req-\ in spontaneously diabetic rats. N Engl J Med 26. Zack JA, Suskind RR: The mortality experi- term toxicologic studies of 2,3,7,8-tetrachlorodi- 1979;300:737. ence of workers exposed to tetrachlorodibenzo- benzo-p-dioxin (TCDD) in laboratory animals. 53. Green JA, Dawson AA, Walker W: Systemic dioxin in a trichlorophenol process accident. J Ann NY Acad Sci 1979;320:397-404. lupus erythematosus and lymphoma. Lancet 1978; Occup Med 1980;22:11-14. 41. Van Miller JP, Lalich JJ, Allen JR: Increased 2:753-756. 27. Ott MG, Holder BB, Olson RD: A mortality incidence of neoplasms in rats exposed to low 54. Douglas AP: Celiac disease and lymphoma of analysis of employees engaged in the manufacture levels of 2,3,7,8-tetrachlorodibenzo-p-dioxin. intestines and brain. N Engl J Med 1977;296:821. of 2,4,5-trichlorophenoxyacetic acid. J Occup Med Chemosphere 1977;10:625-632. 55. Vianna NJ, Polan A: Lymphomas and occupa- 1980;22:47-50. 42. National Toxicology Program (Report 80-31): tional benzene exposure. Lancet 1979;2:1394-1395. 28. Cook RR, Townsend JC, Ott MG, et al: Mortal- Carcinogenesis Bioassay of 2,3,7,8-Tetrachlorodi- 56. Paffenbarger RS, Wing AL, Hyde RT: Charac- ity experience of employees exposed to 2,3,7,8\x=req-\ benzo-p-dioxin (CAS No. 1746-01-6) in Osborne\x=req-\ teristics in youth predictive of adult-onset malig- tetrachlorodibenzo-p-dioxin (TCDD). J Occup Med Mendel rats and B6C3F1 mice (Gavage Study), nant lymphomas, melanomas, and leukemias. 1980;22:530-532. technical report series 209, National Institutes of JNCI 1978;60:89-92. 29. Riihimaki V, Asp S, Hernberg S: Mortality of Health publication 82-1765. Research Triangle 57. Krikorian JG, Burke JS, Rosenberg SA, et al: 2,4-dichlorophenoxyacetic acid and 2,4,5-trichloro- Park, NC, Dept of Health and Human Services, Occurrence of non-Hodgkin's lymphoma after phenoxyacetic acid herbicide applicators in Fin- 1982. therapy for Hodgkin's disease. N Engl J Med land: First report of an ongoing prospective cohort 43. National Toxicology Program (Report 80-32): 1979;300:452-458. study. Scand J Work Environ Health 1982;8:37\x=req-\ Carcinogenesis Bioassay of 2,3,7,8-Tetrachlorodi- 58. Cole P, MacMahon B: Attributable risk per- 42. benzo-p-dioxin (CAS No. 1746-01-6) in Swiss\x=req-\ cent in case-control studies. Br J Prev Soc Med 30. Project Ranch Hand II: An Epidemiologic Webster Mice (Dermal Study), technical report 1971;25:242-244. Investigation of Health Effects in Air Force series 209, National Institutes of Health publica- 59. Eichers TR, Andrilenas PA, Anderson TW: Personnel Following Exposure to Herbicides, tion 82-1757. Research Triangle Park, NC, Dept of Farmers' Use of Pesticides in 1976, agricultural Baseline Mortality Study Results Brooks Air Health and Human Services, 1982. economic report 418. National Economic Analysis Force Base, Tex, Epidemiology Division, Data 44. Young AL, Calcagni JA, Thalken CE, et al: Division, Economics, Statistics, and Cooperatives Sciences Division, US Air Force School of Aero- The Toxicology, Environmental Fate, and Human Service, US Dept of Agriculture, 1978. space Medicine, 1983. Risk of Herbicide Orange and Its Associated 31. Bishop CM, Jones AH: Non-Hodgkin's lym- Dioxin, US Air Force technical report 78-92.

Downloaded from jama.ama-assn.org at University of Waterloo on January 13, 2012 Frequency of Elevated' Serum Pesticide Metabolite Levels in Exposed Dairy-Farm Families to the list of possible battery retrieval Compared With National Background Data methods. Use of a magnet to remove batteries from the stomach has been Prevalence of Elevated Serum Analytes (%) described,2 but, to our knowledge, a Heptachlor magnet has not been used for removal Population Epoxide Oxychlordane Transnonachlor of batteries from other orifices. A Exposed farm-family members 9/39t (23.1) 9/39f (23.1) 12/39t (30.8) magnetized screwdriver worked very well in the following case. Participants in NHANES lit 118/3172(3.7) 126/3174(4.0) 205/3174(6.5) Report of a Case.\p=m-\A9-year-old girl ' Greater than or equal to 1 ppb. awoke with acute otalgia in the left ear tP<.01 by Fisher's exact test (two tailed). immediately after her 5-year-old play\x=req-\ II indicates Second National Health and Nutrition Examination ÍNHANES Survey. mate inserted a miniature toy watch battery in her ear canal. Her intense pain continued and she was examined ranging up to 89.2 ppm (fat basis). not show increased prevalence of in our clinic about 90 minutes after the Subsequently, thousands of gallons of abnormally elevated LFT results or incident. The battery was lodged mid\x=req-\ milk and other dairy products were any consistent pattern of subclinical way in the canal and efforts to retrieve taken off of local groceries' shelves differences in mean LFT results, com¬ the battery with alligator forceps were after laboratory analyses showed hep¬ pared with a reference population of unsuccessful and painful. After an tachlor epoxide contamination levels in healthy, unexposed persons.5 intense hunt for a magnet, a hospital the milk as as 12.6 maintenance several high ppm (fat Comment.—The purpose of this person brought These levels were over ten times basis). was to evaluate the of magnetized tools to the clinic. The first than the Food and study magnitude greater Drug and the of acute attempt with a magnetized screwdriver Administration action level of 0.1 exposures presence and removed ppm and subacute health effects in a group successfully painlessly the Examination of the ear (fat basis).3 of high-risk persons; it was not battery. canal showed moderate of Study.—To assess the potential ex¬ intended to address long-term, chronic erythema the canal without or erosion. tent of human exposures and health disease risks (such as carcinogenesis). swelling No further was effects, we studied a group of people at On the basis of these data, we con¬ treatment necessary. highest risk of exposure, namely, cluded that, although there was evi¬ Comment.—Since physicians are in¬ dairy-farm families who consumed un¬ dence of somewhat elevated serum creasingly likely to be faced with a diluted raw milk products known to be concentrations of heptachlor metabo¬ miniature battery as a foreign body in contaminated. Using a standardized lites, these high-risk persons did not a body orifice, we propose that a questionnaire, we elicited from study exhibit related metabolic effects. Thus, magnet should be tried first to retrieve participants basic demographic data we were able to offer reassurance to the battery. (The presence of a metal and information about their consump¬ persons in the general community that tympanotomy tube or similar metal tion of dairy products. We also col¬ they are likely to be in no danger, since appliance would probably be the only lected serum samples for pesticide they were at even lower risk of expo¬ contraindication.) Perhaps medical of¬ analyses and for both routine liver sure than were the members of the fices and emergency rooms should be function tests (LFTs) and specific dairy-farm families whom we tested. equipped with a permanently magne¬ assays for hepatic microsomal enzyme Paul A. Stehr-Green, DrPH tized probe or a magnetized screwdriv¬ induction. Rebecca J. Schilling, DVM er! We compared results of analyses for Virlyn W. Burse Karen K. Steinberg, PhD Gregory L. Landry, MD the 45 study participants (representing Centers for Disease Control M. Bruce Edmonson, MD 13 families) with national background Atlanta University of Wisconsin Center data collected in the for Health Sciences population Second Wendy Royce, RNP Madison National Health and Nutrition Exami¬ James C. Wohlleb, MS nation and with Arkansas Department of Health 1. Kavanagh KT, Litovitz T: Miniature battery foreign Survey' unexposed Little Rock bodies in auditory and nasal cavities. JAMA 1986;255:1470\x=req-\ reference from 1472. populations neighbor¬ H. MD states. As shown in the there Denny Donnell, 2. Yasuo I, Noriyuki I, Satoru S: Magnetic removal of ing Table, Missouri Department of Health alkaline batteries from the stomach. J Pediatr Surg was a statistically significant increased Jefferson City 1985;20:250-251. of serum levels prevalence pesticide 1. Casarett LJ, Doull J: Toxicology: The Basic Science of elevated above 1 ppb among the farm- Poisons. New York, Macmillan Publishing Co Inc, 1976. members. Both before and after 2. Consolidated heptachlor/chlordane hearings. Federal family Register 1976;41(Feb 19):7552-7585. adjustment for age and sex differences, 3. Heptachlor and heptachlor epoxide in milk: Revision of of the farm- regulatory level. Federal Register 1982;47(Oct 26):47466. using analysis covariance, 4. C: Residues metabolites of Murphy R, Harvey and Incorrect Table Title. \p=m-\An error occurred family members had statistically sig¬ selected persistent halogenated hydrocarbons in blood a in the article entitled Herbi- nificant higher mean levels of serum specimens from general population survey. Environ "Agricultural ±0.94 Health Perspect 1985;60:115-120. cide Use and Risk of Lymphoma and Soft\x=req-\ heptachlor epoxide (0.81 ppb), 5. Hoffman RE, Stehr-Green PA, Webb KB, et al: Health Tissue Sarcoma," published in the Sept 5 serum oxychlordane (0.70 ±0.75 ppb), effects of long-term exposure to 2,3,7,8-tetrachlorodibenzo\x=req-\ JAMA 1986;255:2031-2038. issue of The Journal (1986;256:1141-1147). and serum transnonachlor ±0.60 p-dioxin. (0.79 On page 1144, the title of Table 3 should ppb) compared with an unexposed have read as follows: "Table 3.\p=m-\Non-Hodg- urban population from this same Attractive Method for Battery kin's Lymphoma in Relation to Duration, region of the United States (Diane Removal Frequency, and Latency of Herbicide Use Rowley, MD, written communication, To the Editor.\p=m-\We appreciated the Among 2,4-Dichlorophenoxyacetic Acid April 1, 1986). In logistic regression article by Kavanagh and Litovitz1 enti- Users" (not ". Duration, Frequency, and and of covariance in tled "Miniature Battery Foreign Bodies Latency of 2,4-Dichlorophenoxyacetic. . Acid analyses analysis Use"). which we adjusted for age and sex, the in Auditory and Nasal Cavities." We high-risk farm-family members did would like to add the use of a magnet Downloaded from jama.ama-assn.org at University of Waterloo on January 13, 2012