the process of cancer induction from environmental chemicals that bind-to DNA.

References

(12) BASSETT WB, WEISS RB: Acute leukemia following cisplatin for bladder cancer. J Clin Oncol 4:614, 1986 (13) HENNINGSH, SHORES R, WENK ML, ETAL: Malignant conversion of mouse

skin tumours is increased by tumour initiators and unaffected by tumour promoters. Nature 304:67-69, 1983 (14) HENNINGS H, SPANGLER EF, SHORES R, ET AL: Malignant conversion and

(/) LJTTERST CL, REED E: Platinum compounds. In Progressive Stages of Malignant Neoplastic Growth (Kaiser HE, ed). London: Alden Press, 1989, pp 85-97 (2) REED E, KOHN KW: Cisplatin and platinum analogs. In Cancer Chemotherapy—Principles and Practice (Chabner BA, Collins J, eds). Philadelphia: Lippincott. In press (3) REED E, OZOLS RF, TARONE R, ET AL: Platinum-DNA adducts in leukocyte

DNA correlate with disease response in ovarian cancer patients receiving platinum-based chemotherapy. Proc Natl Acad Sci USA 84:5024-5028, 1987 (4) KEMPF SR, IVANKOVIC S: Chemotherapy-induced malignancies in rats after treatment with cisplatin as single agent and in combination: Preliminary results. Oncology 43:187-191, 1986 (5) LEOPOLD WR, MILLER EC, MJLLER JA: Carcinogenicity of antitumor

(6)

(S) (9)

(10) REDMAN JR, VUGRIN D, ARLIN ZA, ET AL: Leukemia following treatment of

germ cell tumors in men. J Clin Oncol 2:1080-1087, 1984 (//) KERBRAT P, LE PRISE PY: Acute leukemia following testicular carcinoma. J Clin Oncol 3:1287, 1985

(15) O'CONNELL JF, KLEIN-SZANTO AJP, DIGIOVANNI DM, ET AL: Malignant

progression of mouse skin papillomas treated widi ethylnitrosourea, Nmethyl-AT-nitro-N-nitrosoguanidine, or 12-0-tetradecanoylphorbol-13acetate. Cancer Lett 30:269-274, 1986 (16) DEBOER JG, GLJCKMAN BW: Sequence specificity of mutation induced by the anti-tumor drug cisplatin in the CHO aprt gene. Carcinogenesis 10:1363-1367, 1987 (17) GARTJJ.KREWSKID, LEE PN, ETAL: Statistical methods in cancer research. Volume IH—the design and analysis of long-term animal experiments. IARC Sci Publ No. 79. Lyon: IARC, 1986, pp 1-219 (18) POKIER MC, REED E, ZWELLING LA, ET AL: Polyclonal antibodies to

quantitate cis-diamminedichloroplatinum(ll)—DNA adducts in cancer patients and animal models. Environ Health Perspect 62:89-94, 1985 (19) REEDE, SAUERHOFFS, POIRIER MC: Quantitationofplatinum-DNA binding

in human tissues following therapeutic levels of drug exposure. Atomic Spectroscopy 9:93-95, 1988 (20) QUINTANILLA M, BROWN K, RAMSDEN M, ET AL: Carcinogen-specific

mutation and amplification of Ha-rcu during mouse skin carcinogenesis. . Nature 322:78-80, 1986 (21) GREENHALGH DA, YUSPA SH: Malignant conversion of murine squamous papilloma cell lines by transfection with the/os oncogene. Mol Carcinog 1:134-143, 1988 (22) GREENHALGH DA, WELTY DJ, STRICKLAND JE, ET AL: Spontaneous Ha-

ras-gene activation in cultured primary murine keratinocytes: Consequences of Ha-ras gene activation in malignant conversion and malignant progression. Mol Carcinog 2:199-207, 1989 (23) POKIER MC, REED E, OZOLS RF, ET AL: DNA adducts of cisplatin in

nucleated peripheral blood cells and tissues of cancer patients. Prog Exp Tumor Res 31:104-113, 1987

Cancer Mortality in the U.S. Flour Industry Michael C. R. Alavanja,* Aaron Blair, Mary N. Masters ment (OR = 8.1) or in the elevator department (OR = 2.8) The mortality experience among 22,938 white males who were at particularly elevated risk of developing NHL, sugwere enrolled in the life insurance program of the American gesting that exposures in these departments should receive Federation of Grain Millers was assessed for the period 1955 further attention. [J Natl Cancer Inst 82:840-848,1990] through 1985 in a cohort mortality analysis and in a nested case-control analysis. Significantly fewer deaths were obControlling insects is an ongoing problem in the grain industry. served among this group than expected for all causes of death Grain, machinery, and facilities are treated chemically to minicombined [standardized mortality ratio (SMR) = 89] com- mize serious losses from insects in all segments of the industry, pared with the number of deaths observed among the general including farms, country (local) grain elevators, regional/inland population of U.S. white males of the same age. Excess risks for developing non-Hodgkin's lymphoma (NHL) (SMR = 149), leukemia (SMR = 136), and pancreatic cancer (SMR = 133) were restricted to workers employed in flour mills, Received August 15, 1989; revised February 23, 1990; accepted March 6, where pesticides are used more frequently than in other 1990. segments of the industry. In the nested case-control analysis, M. C. R. Alavanja (Epidemiology and Biostatistics Program), A. Blair excess risks for developing these cancers were also observed (Environmental Epidemiology Branch), Division of Cancer Etiology, National in these workers, but the relative risk for developing NHL Cancer Institute, Bethesda, MD. M. N. Masters, Westat, Inc., Rockville, MD. [odds ratio (OR) = 4.2] was approximately twice that for We thank John Quarantillo for programming support and Theresa P. Osborne developing pancreatic cancer (OR = 2.2) and that for develfor manuscript preparation. oping leukemia (OR = 1.8). Within the flour mills, the Correspondence to: Michael C. R. Alavanja, Dr. P.H., Executive Plaza workers who had ever worked in the maintenance depart- North, Rm. 543, National Institutes of Health, Bethesda, MD 20892.

840

Journal of the National Cancer Institute

Downloaded from http://jnci.oxfordjournals.org/ at Carleton University on May 10, 2015

(7)

cis-platinum(II) coordination complexes in the mouse and rat. Cancer Res 39:913-918, 1979 BARNHART KM, BOWDEN GT: Cisplatin as an initiating agent in two-stage mouse skin carcinogenesis. Cancer Lett 29:101-105, 1985 KALDOR JM, DAY NE: Aspects concerning the study of the quantitative carcinogenic effects of chemotherapeutic agents in man. IARC Sci Publ No. 70. Lyon: IARC, 1986, pp 327-335 KALDOR JM, DAY NE, BAND P, ET AL: Second malignancies following testicular cancer, ovarian cancer and Hodgkin's disease: An international collaborative study among cancer registries. Int J Cancer 39:571-585,1987 STEWART AL, WILKINSON PM: Rapid onset of acute myeloid leukaemia following radiotherapy and chemotherapy for metastatic semi no ma of the testis.J Cancer Res Clin Oncol 100:109-111, 1981

metastasis of mouse skin tumors: A comparison of SENCAR and CD-I mice. Environ Health Perspect 68:69-74, 1986

Downloaded from http://jnci.oxfordjournals.org/ at Carleton University on May 10, 2015

(subterminal) grain elevators, export terminals, flour and feed made up 7,927 members of the cohort but accounted for only 697 mills, and flour- or feed-processing plants. Chemical treatment deaths; this segment of the union population was excluded from may also be applied to grain prior to transport in trucks, railcars, the cohort because of insufficient numbers of deaths. The cohort barges, and ships. Some of the chemicals used are infrequently was followed from 1955, the inception year of the life insurance encountered in other components of U. S. agriculture, and, hence, plan, through 1985. their health effects have been inadequately assessed. A complete registration card for life insurance included name The pattern of chemical use in the U.S. grain industry has of the worker, Social Security number, union local number, changed dramatically during the past 40 years. The wide use of company name, date of union registration, date of termination 1 ,r-(2,2,2-trichloroethylidene)bis[4-chlorobenzene] (DDT) and from union, date of first employment, and date of termination of hydrogen cyanide in the 1940s and 1950s was curtailed by 1960. employment. (For a substantial portion of the cohort, the date of Use of the liquid fumigants ethylene dibromide, carbon tetrachlo- termination of employment was missing.) For a few subjects, the ride, and carbon disulfide began during the late 1940s and 1950s company name was also available from the death certificate. and continued until the mid-1980s. Methyl bromide, introduced Companies were assigned Standard Industry Codes (SICs) (4) shortly after World War II, and malathion and pyrethrum, based on company data, on information from Milling and Baking introduced in the mid-1950s, continue to be used today. Phos- News (a periodical published by the Sosland Publishing Co., phine and other phosphide formulations, introduced in the mid- Shawnee Mission, KS), and on information from other trade1960s, increasingly make up a larger portion of the fumigants association directories of the milling industry, the grain elevator used today, now that the liquid fumigants are not used. Dichlor- industry, and the animal feed industry. The vital status of each cohort member was assessed through union records, Social vos, introduced in 1970, is also used today in some facilities. A preliminary proportionate mortality study of 1,114 white Security Administration records, a credit bureau search, a mail male union members of the American Federation of Grain Millers survey of union members, and information supplied by the noted elevated mortality from pancreatic, lymphatic, and he- officers of union locals. Underlying and contributing causes of matopoietic cancers among employees of flour mills (7). It was death were coded by a nosologist using the International Classinot possible to identify specific occupational factors associated fication of Diseases Adapted for Use in the United States (ICD A), with this excess risk because information on work histories and eighth revision (5), according to the rules in effect at the time of exposure histories was limited; however, pesticides were sus- the subject's death. pected. In a more recent study, pesticide applicators who were Each union local decides as a group whether to participate in exposed to the grain fumigant phosphine were found to have a the American Federation of Grain Millers' life insurance plan, so significantly increased frequency of translocations in G-banded that the health status of an individual should not have influenced lymphocytes (2). Less stable aberrations, including chromatid participation in the insurance plan. Union locals participating deletions and gaps, were significantly increased only during the represented members employed at grain elevators, flour mills, application season and not when fumigation had been terminated animal feed-processing plants, soybean-processing plants, plants for the winter. This study of applicators and an earlier study (3) that blend flour and other ingredients, a beet sugar-processing reported that phosphine concentrations in the work environment plant, a prepared foods plant, and a potato-processing plant. exceeded national standards during grain fumigant applications. These facilities included small private companies, farmers' coopFor the study being reported here, our preliminary investiga- eratives, a state-owned facility, large grain-handling companies, tion was expanded with a cohort mortality study, a nested and manufacturers of various food products. Workers in small case-control study, and more detailed efforts to further evaluate country grain elevators were not included because they are cancer risks among workers in the grain industry. The cohort typically not organized by the American Federation of Grain mortality study examined the cause-specific mortality of grain Millers. Although the study cohort did not include a representaworkers, and the nested case-control study (for which more tive number of plants from all segments of the industry, flour complete employment information was available) attempted to mills were well represented (n = 57), including at least one mill relate occupational factors and exposure information with spe- from each of the nation's 10 largest (as of 1982) flour-milling companies. cific causes of death. Case-Control Study

Subjects and Methods Cohort Mortality Study Information on the cohort was assembled from the enrollment records of 40,247 current or former members of the American Federation of Grain Millers' life insurance program; this information is kept at union headquarters in Minneapolis, MN. A total of 22,938 white males had complete enrollment records, and they became the study cohort. Complete enrollment records are filled out 3 months after the worker's entrance into the industry. Excluded from the study were 9,058 workers with only temporary enrollment records and 324 workers for whom date of birth was unavailable from any source. All females and nonwhite males Vol. 82, No. 10, May 16, 1990

A nested case-control analysis was conducted using the employment information that we were able to obtain for a portion of the cohort. The analysis with flour millers showed elevated standardized mortality ratios (SMRs) for non-Hodgkin's lymphoma (NHL), leukemia, or pancreatic cancer. The analysis used two distinct control groups. For the first control group, seven white male controls were selected from eligible cohort members from any segment of the grain industry who were alive during the year of death of the index case; they were individually matched to each case with regard to date of birth (±1 yr). These controls may have died in any subsequent year or may have been alive at the end of the study. For the second control group, seven white male dead controls ARTICLES

841

were individually matched to each case with regard to date of birth (± 3 yr) and year of death. These controls were men who had died of any cause other than NHL, leukemia, or pancreatic cancer. Exposure Assessment

Survey of Current Members A demographic profile, an occupational history, and information on alcohol consumption and smoking habits among the workers in the grain industry were obtained through a survey of

Methods of Analysis In the cohort mortality analysis, the observed number of deaths from each cause was compared with the expected number, on the basis of 5-year age-specific and calendar-year-specific mortality rates among the U.S. white male population. SMRs were calculated as the ratio of observed to expected deaths, and the confidence intervals (CIs) and P values were determined by standard methods (6,7). The maximum likelihood method was used to compute odds ratios (ORs) (8,9). CIs for the ORs were computed by Gart's method (70). All maximum likelihood risk estimates were adjusted for the effect of age by stratification by five age groups, whereas conditional logistic regression was used to develop risk estimates for duration of follow-up response (8).

Results Cohort Mortality Analysis Table 1 shows the results of vital status follow-up for 22,938 study subjects, 19,270 of whom were alive and 3,668 of whom were dead at the end of follow-up. Death certificates were not available for 208 (6%) study subjects thought to be dead. Cause of death was listed as unknown in the SMR analysis for these workers. For another 324 white males, date of birth was unavailable, and these workers were excluded from the cohort; 322 were alive, and two died of non-cancer-related causes by the end of follow-up. Flour millers constituted the largest numbers of living workers (7,746) and of dead workers (1,914). The mean year of birth was 1911.6 for the 3,668 white male workers in the grain industry cohort who died during the follow-up period from 1955 through 1985; their median age at death was 65 years. Table 2 shows the mortality experience of the grain industry cohort according to major causes of death. Significantly fewer

Table 1. Number of workers in grain industry cohort by industry subdivision and vital status No. of workers dead Industry subdivision (SIC No.)

Flour milk (2041) Potato processing (2034) Soybean processing (2075) Animal feed processing (2048) Grain storage elevators (5153) All other facilities (2043, 2045, 2063) Unknown Total

842

Total No. of workers alive

Total No. of workers alive or dead

lota]

With death certificate

Without death certificate

7,746 3,493 2,227 1,810 680 2,714

1,914 356 320 294 120 427

1,828 314 298 277 114 400

86 42 22 17 6 27

9,660 3,849 2,547 2,104 800 3,141

600 19,270

237 3,668

229 3,460

8 208

837 22,938

Journal of the National Cancer Institute

Downloaded from http://jnci.oxfordjournals.org/ at Carleton University on May 10, 2015

Information on exposure was obtained from the following sources: (a) telephone interviews with 220 study cohort members with pesticide application experience who also had responded to the mail survey, (b) interviews with managers and industrial hygienists of 21 of the facilities included in this study, (c) discussion with the headquarters' staff at the American Federation of Grain Millers' union, (d) consultation with an experienced grain industry entomologist, and (e) walk-through surveys on industrial hygiene conducted at 10 plants. Work histories, including job title, department, and years of employment in the facility, were available from corporate employment records for only one fourth of the deceased subjects for the nested case-control study. We determined potential exposures to pesticides for each individual in this group by linking the employee's work history to exposure information given by the department. Job title and department were not available from corporate employment records for workers who were living or for three fourths of the workers who were dead. For both of these study subjects, potential exposures were imputed from the years the worker was employed in a facility and from the occupational chemicals used in that facility during those years (i.e., the exposure matrix categorized all employees that worked in a plant in the same yr as having the same occupational exposures). Industry subdivision and duration of follow-up were the only exposure-related variables available for analysis in the cohort mortality study. In the nested case-control study, the exposure variables were ever/never employed in an industry subdivision (e.g., flour mills, grain elevators, or soybean-processing plants), duration of follow-up since first employment in an industry subdivision, and employment in specific departments of the plant (e.g., maintenance department or warehouse department). For controls, the accumulation of duration of follow-up ceased in the year of death of the matched case.

all 6,578 active and retired members (in good standing) of the union's life insurance program in 1985 and 1986. The survey used a mailed self-administered questionnaire and a telephone interview of those not responding by mail. Up to two remailings and six telephone callbacks were attempted to maximize responses. The response rate to this questionnaire among the current participants in the life insurance plan in the cohort was 78%, with 5,084 workers responding among the 6,578 workers surveyed.

Table 2. Observed and expected numbers of deaths, SMRs, and 95% CIs among a cohort of white male workers in the grain industry No. of deaths Causes of death (ICDA No.) Expected

14 727 18 1,728 1,225 244 232 85 56 96 12 57 43 438 168 96 56 108 208 3,668

37.5 829.3 35.8 2,131.9 1,480.2 292.7 264.2 95.5 60.3 182.4 24.1 91.4 54.1 412.1 137.2 94.2 178.2 — 4,125.6

37* _ 88* 50* 81* 83* 83* 88 89 93 53*

95% CI

62* 79 106 122* 102

20-63 81-94 30-79 77-85 78-88 73-94 77-100 71-110 70-121 43-64 26-87 47-81 57-107 97-117 105-142 83-124

61t

52-70

89*

86-92

50t

*/> 25

5 49 4.1 0.8-22.1

. 8 64 9.4 1.4-61.5

21 211 4.2 1.2-14.2

6 89 1.2 0.4-4.1

12 142 1.9 0.7-4.9

33 358 2.2 1.1-4.3

3 21 3.6 0.6-20.5

4 57 1.1 0.2-6.1

25 285 1.8 0.8-3.9

*P linear trend = .01. t/* linear trend = .07. %P linear trend = 21.

Table 7. Risk of NHL, pancreatic cancer, and leukemia by department of employment within flour mills Exposed Department

OR

95% CI Cases

Elevator Milling Maintenance Warehouse Feed

2.8 0.5 8.1 0.6 0.8

Elevator Milling Maintenance Warehouse Feed

0.5 2.1 0.9 1.5

Elevator Milling Maintenance Warehouse Feed

0.7 0.2 4.4 0.7 1.8

NHL 3 3 4 4 1 Pancreatic cancer 2 10 2 8 0 Leukemia 1 1 2 1 1

Vol. 82, No. 10, May 16, 1990

Controls

9 32 7 35 9

0.6-12.2 0.1-2.1 1.4-47.7 0.2-2.2 0.1-6.6

22 50 16 46 8

0.1-23 0.6-6.8 0.2-4.3 0.5-4.5

9 19 6 9 4

0.1-7.1 0.02-1.9 0.4-51.2 0.1-73 0.2-18.2

= 4.4) were seen among those ever employed in the maintenance department compared with the risks for all other flour millers. Flour millers in the feed department had a nonsignificantly elevated risk of developing leukemia (OR = 1.8), whereas workers in the milling department (OR = 2.1) and warehouse employees (OR = 1.5) had a nonsignificantly elevated risk of developing pancreatic cancer. When study subjects had worked in more than one department, they were listed in each department for computation of ORs.

Downloaded from http://jnci.oxfordjournals.org/ at Carleton University on May 10, 2015

No. of cases No. of controls OR 95% CI

16-20

Ever a flour miller

Grain Industry Employee Profile To assess the possibility that the excess NHL risk may have been due to nonoccupational factors or previous occupational exposures, we surveyed by mail all current members of the union life insurance plan from these facilities (employees or retirees). Educational level has been found to be a relatively sensitive indicator of social class (11), and social class and education have been related to the development of NHL (72). In this study, workers in flour mills were found to have essentially the same level of formal education as workers in the other segments of the grain industry, with 10% of flour millers completing more than 12 ARTICLES

84:

Discussion Results from this cohort mortality investigation of 22,938 current and former white male members of the American Federation of Grain Millers' life insurance program showed nonsignificant excesses in the rates of NHL (SMR = 149), leukemia (SMR = 136), and pancreatic cancer (SMR = 133) among flour millers, compared with the expected rates among U.S. white males. Flour millers with 20 or more years of follow-up since first employment in the industry were at a statistically significant excess risk of developing NHL. The excess risk of NHL and pancreatic cancer displayed a significant trend with duration of follow-up. These results are generally consistent with an earlier proportionate mortality investigation that was performed on a portion of the entire cohort (/). The overall mortality experience for this group was less than that for U.S. white males (SMR = 85), as expected for an occupational cohort (7). This "healthy worker effect" was observed in every subdivision of the cohort except for workers in the potato-processing industry, who were at a slightly elevated overall risk of mortality, which increased the overall mortality for the combined category of workers in other grain industries to an SMR of 104. We have no explanation for the extremely low SMRs observed for NHL and pancreatic cancer among other workers in the grain industry. This deficit results in relative risk estimates from the case-control study that are much higher (over twofold) than those from the cohort mortality study. Internal comparisons are usually thought to be superior to external comparisons (7), because the populations being compared may be more comparable with regard to life-style and other factors than the general population. The lower than expected risk of developing NHL and pancreatic cancer observed among flour millers within 9 years of first employment suggests that flour millers and other workers in the grain industry were at similar risks for developing these cancers prior to employment in the grain industry. Nested case-control analyses within the grain industry cohort

46

indicated that the risk of developing NHL and pancreatic cancer among flour millers was significantly increased compared with that among either a living or a dead control group. However, the elevated risk of leukemia was not statistically significant. The risk of NHL also increased significantly with duration of followup in flour mills (P = .01), rising from an OR of 1.5 for those employed less than 15 years to an OR of 9.4 for those employed 25 or more years. Since the final date of employment could not be obtained reliably from union or company records, we could not assess the risk of cancer by duration of employment. The hypothesis that pesticides applied in flour mills may be the most likely agents responsible for the excess NHL risk found in these facilities was first proposed in an earlier study (7). Additional support was given to this hypothesis when a survey of current and retired workers in the cohort revealed that twice as many flour mill employees applied pesticides while working than workers in other subdivisions of the industry, and the excess mortality from NHL, leukemia, and pancreatic cancer appears more striking among flour mill employees. Alternative hypotheses for the cause of the excesses of NHL, leukemia, and pancreatic cancer include a role for grain or flour dust, extraction solvents, agricultural chemicals applied on the farm or on the path from the farm to the grain mill, or nonoccupational factors associated with these cancers. Some information is available to evaluate each of these alternatives. For example, grain dust occurs in most segments of the industry (e.g., flour mills, grain elevators, soybean-processing plants, and animal feedprocessing plants), yet the cancer excesses occurred only among employees of flour mills. Similarly, extraction solvents (e.g., hexane) are used almost exclusively in soybean-processing plants, where risks for these cancers are low, but they are used rarely, if ever, in flour mills, where risks are high. Pesticide application to the grain before it gets to the flour mill is also an unlikely explanation because workers in animal feed-processing plants and soybean-processing plants would incur a similar exposure, and they have not been found to have an elevated risk of developing NHL. The difference in risk for these cancers could not be accounted for by differences in the characteristics of the various groups of workers because employees in the flour industry did not appear to possess characteristics that differed from those of employees in other segments of the grain industry. NHL has been associated with exposure to agricultural chemicals as well as with living on farms (73) and handling herbicides {14). However, our survey of current and retired workers in the cohort indicated that residence on a farm was less prevalent among flour millers than among workers in other segments of the industry. Although a greater proportion of flour millers consumed alcohol than workers in other subdivisions of the grain industry, alcohol consumption is not known to be associated with the development of NHL or leukemia (75). The elevated risk of developing pancreatic cancer among flour millers compared with the risk among workers in other grain industries or the risk among U.S. males in general may be caused by excess alcohol consumption (16,17). However, alcohol abuse is not indicated in this industry, because mortality from cirrhosis of the liver was not elevated among flour millers or among workers in other grain industries. It is difficult to assess historical exposures to pesticides for individual workers in the grain industry. Actual measurements of Journal of the National Cancer Institute

Downloaded from http://jnci.oxfordjournals.org/ at Carleton University on May 10, 2015

years of education compared with 13% of workers in all other segments of the industry combined. The proportion of workers who ever smoked was slightly higher among flour millers: 70% for flour millers versus 65% for workers in all other segments of the grain industry. However, the percentage of workers who were current smokers was higher among workers in other segments of the grain industry (41%) than among flour millers (36%). Sixty percent of flour millers consumed at least one alcoholic drink per week, whereas only 48% of the workers in the other segments of the grain industry consumed that amount. Since the effect of occupational exposures to the grain industry could be confounded by exposures associated with living on a farm (73), the proportion of workers in each segment of the industry who ever lived on a farm was determined. Fewer flour millers (46%) lived on a farm than did workers in all other segments of the industry (61%). Occupational exposure to pesticides, on the other hand, was much more prevalent among flour millers than among workers in other grain industries. Thirty-one percent of flour millers applied pesticides while they were employed at these facilities, whereas only 16% of workers in all other segments of the grain industry did so. No association was found between previous occupations and work in the flour industry.

Vol. 82, No. 10, May 16, 1990

small. We were able to obtain from the companies work histories for only a small portion of the cases and controls, which limited analyses by department and job. Exposures to herbicides {14), cotton dust (75), wood dust (75), epoxy glues (79), and butadiene (20) as well as work in the textile industry (27) have previously been associated with NHL, but there is little reason to expect that grain workers would have unusual exposures to these substances. Although these limitations may have reduced somewhat our capabilities to fully evaluate associations, they are unlikely to substantially affect the overall association between work in a flour mill and NHL risk. Finally, the study could not assess the recent regulatory efforts of the U.S. Environmental Protection Agency, which banned the use of ethylene dibromide, carbon tetrachloride, and carbon disulfide as fumigants in the grain industry in the mid-1980s. Continued follow-up of this cohort, however, will make it possible to determine whether the occupationally associated lymphomas observed among flour millers from 1955 through 1985 continue to be an occupationally related hazard in the industry today. In summary, data from the current investigation uncovered a significant excess risk of developing NHL among employees of flour mills; this risk rose with increasing duration of follow-up. These findings are based on the analysis of a large cohort of workers, which included the entire work force of 57 flour-milling facilities nationwide. The NHL excess was observed in an SMR analysis where the U.S. white male population was used for comparison and in a nested case-control analysis. When a comparison was made with other workers in the grain industry by a case-control approach, the relative risk rose to more than ninefold after 25 years of follow-up. Although the appropriateness of using dead or living controls continues to be an actively discussed topic (22-25), the excess risk of developing NHL among flour millers was observed when comparisons were made with the risks among either a living or a dead control group in the nested case-control study. Data from a survey of more than 5,000 workers within the cohort indicated that these findings were unlikely to be due to other factors, such as socioeconomic status, smoking, or residence on a farm. Evidence for an association between employment in the flour industry and the development of leukemia and pancreatic cancer is less impressive than that for the development of NHL. Based on data from the case-control study, elevated risks observed for leukemia and pancreatic cancer among flour millers were considerably smaller than the risks observed for NHL among flour millers and did not show clear exposure-response patterns.

Downloaded from http://jnci.oxfordjournals.org/ at Carleton University on May 10, 2015

pesticide exposure are rare, and the records documenting worker application of pesticides were not available. The problem of exposure estimation is further compounded by the application procedure followed by most companies, where any worker, regardless of job title, could volunteer to assist in applying pesticides for overtime pay. During the period of this study, from 1955 through 1985, 10 chemicals were widely used in the grain industry to control insect infestation. These chemicals were used in a myriad of commercial formulations and often were used simultaneously with a number of pesticides at each facility. It was, therefore, impossible to use job titles or any other available record to retrospectively ascribe to any worker pesticide exposures from application. Cancer risks did differ somewhat by department of employment in flour mills. The pattern of risk by department appeared similar for NHL and leukemia, but it was distinctly different for pancreatic cancer, suggesting that if occupational agents are involved, they may differ for NHL and pancreatic cancer. A significantly elevated risk of developing NHL (OR = 8.1) and a nonsignificantly elevated risk of developing leukemia (OR = 4.4) were seen among flour workers who were ever employed in the maintenance department compared with the risks observed among all other flour millers. Nonsignificant increases occurred for NHL among workers employed in the elevator department of the mill (OR = 2.8) and for pancreatic cancer among workers employed in the milling department (OR = 2.1) and the warehouse department (OR = 1.5). Since workers in the maintenance department are responsible for maintaining and repairing machinery and equipment in every department of a flour mill, they would tend to have the most diverse occupational exposures in these facilities, and the occupational exposures encountered in any one department would not, therefore, be a good surrogate for the occupational exposures of maintenance workers. Chemical lubricants used by maintenance workers in the mills are unlikely etiologic agents, because these same chemicals would be expected to be used in every other segment of the grain industry to approximately the same extent as they are used in the flour mills. The excess risk of developing NHL among employees of the grain elevator department may reflect greater chemical exposure. In our survey, a greater number of employees from the elevator department than from other grain facilities indicated that they were involved more frequently in applying grain fumigants, including carbon tetrachloride, carbon disulfide, and phosphine. Since these chemicals tended to be applied during regular work hours, direct exposure may be more likely than for other fumigants used in flour mills, such as methyl bromide, ethylene dibromide, DDT, and hydrogen cyanide, all of which tended to be used when the plant was closed for a periodic general plant fumigation. Although this analysis is based on a small number of subjects, the relatively large variation in risk by department suggests that more detailed assessments of exposure by job function may help identify specific exposures that may be responsible for the observed cancer risks. A number of limitations to the current study frustrated attempts to identify the cause of the excess risk of neoplasms observed. NHL, leukemia, and pancreatic cancer are rare diseases, so that even in this large cohort, the actual number of cases is rather

References (/) ALAVANJA MCR, RUSH GA, STEWART P, ET AL: Proportionate mortality

study of workers in the grain industry. JNQ 78:247-252, 1987 (2) GARRY VF, GRIFFITH J, DANZL JJ, ET AL: Human genotoxicity: Pesticide

applicators and phosphine. Science 246:251-255, 1989 (3) ZAEBST DD, BLODE LM, BURROUGHS P, ET AL: Phosphine exposure in grain

elevators during fumigation with aluminum phosphine. Applied Ind Hygiene 3:146-150, 1988 (4) OFFICE OF MANAGEMENT AND BUDGET: Standard Industrial Classification

Manual. OMB Publ No. (PB) 87-100012. Springfield, VA: Natl Tech Inf Ser, 1987 (5) NATIONAL CENTER FOR HEALTH STATISTICS: International Classification of

Diseases, 8th rev. DHEW Publ No. (PHS) 67-1693. Washington, DC: US Govt Print Off, 1967

ARTICLES 847

(6) UDDELL FDK: Simple exact analysis of tbe standardized mortality ratio. J (17) HEUCH I, KVALEG, JACOBSENBK, ET AL: Use of alcohol, tobacco and coffee Epidemiol Community Health 38:85-88, 1984 and risk of pancreatic cancer. Br J Cancer 48:637-643, 1983 (18) SIEMIATYCKI J, RICHARDSON L, GERIN M, ET AL: Associations between (7) MONSON RR: Occupational Epidemiology. Boca Raton, FL: CRC Press, several sites of cancer and nine organic dusts: Results from an hypothesis1980, pp 107-133 generating case-control study in Montreal, 1979-1983. Am J Epidemiol (8) BRESLOW NE, DAY NE: Statistical Methods in Cancer Research, vol 1. 123:235-249, 1986 IARC Sci Publ No. 32. Lyon: IARC, 1980 (19) CARTWRIGHT RA, MCKTNNEY PA, O'BRIEN C, ET AL: Non-Hodgkin's (9) MANTEL N: Chi-square tests with one degree of freedom, extensions of the lymphoma: Case control epidemiological study in Yorkshire. Leuk Res Mantel-Haenszel procedure. J Am Stat Assoc 58:690-700, 1976 12:81-88, 1988 (10) GART JJ: Point and interval estimation of the common odds in the combination of 2 x 2 tables with fixed marginals. Biometrika 57:471-475, 1970 (20) DOWNS TD, CRANE MM, KIM KW: Mortality among workers at a butadiene facility. Am J Ind Med 12:311-329, 1987 (//) SUSSER MW, WATSON W: Sociology in Medicine. Oxford: Oxford Univ (2/) DELZELLE,GRUFFERMANS:Cancerandothercausesofdeathamongfemale Press, 1971, pp 121 textile workers, 1976-78. JNC1 71:735-740, 1983 (12) CANTOR KP, FRAUMENI JF JR: Distribution of non-Hodgkin's lymphoma in (22) GORDIS L: Should dead cases be matched to dead controls? Am J Epidemiol the United States between 1950 and 1975. Cancer Res 40:2645-2652,1980 115:1-5, 1982 (13) BLAIR A, MALKER H, CANTOR KP, ET AL: Cancer among farmers. Scand J Work Environ Health 11:397-407, 1985 (23) CALLE EE: Criteria for selection of decedent verses living controls in a mortality case-control study. Am J Epidemiol 120:635-642, 1984 (14) HOAR SK, BLAIR A, HOLMES FF, ET AL: Agricultural herbicide use and risk of lymphoma and soft-tissue sarcoma. JAMA 256:1141-1147, 1986 (24) MCLAUGHUN JK, BLOT WJ, MEHL ES, ET AL: Problems in the use of dead (15) SCHOTTENFELD D, FRAUMENI JF JR: Cancer Epidemiology and Prevention. controls in case-control studies. Am J Epidemiol 121:131-139, 1985 Philadelphia: Saunders, 1983, pp 728-738, 754-778, 1983 (25) FINKELSTEIN MM, LJSS GM: Selection bias in occupational case-control studies that use death registries to select subjects: A discussion and (16) CUZICK J, BABIKER AG: Pancreatic cancer, alcohol, diabetes mellitus and demonstration. Am J Ind Med 12:21-31, 1987 gallbladder disease. IntJ Cancer 43:415-421, 1989

Richard J. Sharpe, * H. Randolph Byers, Charles F. Scott, Susan I. Bauer, Theodore E. Maione

Downloaded from http://jnci.oxfordjournals.org/ at Carleton University on May 10, 2015

Growth Inhibition of Murine Melanoma and Human Colon Carcinoma by Recombinant Human Platelet Factor 4

vated killer cell activation and the induction of other cytoAlthough it is well established that angiogenesis is essential to kines. [J Natl Cancer Inst 82:848-853,1990]. tumor development, no human protein with high specificity and efficacy for prevention of angiogenesis has been characTumor development is a complex process that includes the terized. In a previous study, we demonstrated that recombi- establishment of a vascular network to provide nutrients and nant platelet factor 4 (rPF 4) inhibited angiogenesis in the oxygen to the growing tumor, as well as remove metabolic waste chicken chorioallantoic membrane. In the present study, we products. It is now recognized that solid tumors are unable to have extended that finding to the use of recombinant human grow beyond a few cubic millimeters in the absence of new platelet factor 4 (rHuPF 4) to inhibit solid tumor growth in the capillary formation (/). This angiogenic process, which is highly mouse. rHuPF 4 effectively suppressed the growth of the regulated in normal tissues, is promoted by tumor-derived prodB16-F10 murine melanoma in syngeneic C57BL/6J hosts and ucts that stimulate local endothelial cells to divide and migrate prevented the growth of primary tumors of both B16-F10 toward the developing neoplasm, producing the dense vasculamurine melanoma and HCT 116 human colon carcinoma in semisyngeneic CByB6F1/J female athymic nude mice. These two transformed cell lines were completely insensitive to rHuPF 4 in vitro at levels (50 ng/mL) that extensively inhibit Received December 14,1989; revised February 2,1990; accepted February 28, normal endothelial cell proliferation. The migration of hu- 1990. R. J. Sharpe, C. F. Scott, S. I. Bauer, T. E. Maione, Repligen Corporation, man endothelial cells was also inhibited at these concentraCambridge, MA. tions of rHuPF 4, suggesting a second mechanism by which R. J. Sharpe, Department of Dermatology, Harvard Medical School and Beth rHuPF 4 may modulate capillary development. The observed Israel Hospital, Boston, MA. antitumor effects of rHuPF 4 might be due to the inhibition of H. R. Byers, Department of Pathology, Harvard Medical School and Massaangiogenesis. This finding could have implications for the chusetts General Hospital, Boston, MA. We thank Dr. Salvadore E. Luria for his thoughtful comments, advice, and development of novel therapeutic approaches to angiogenic never-ending encouragement. diseases. Alternative, and possibly concurrent, mechanisms Correspondence to: Richard J. Sharpe, M.D., Repligen Corporation, One of the rHuPF 4 antitumor effect include lymphokine-acti- Kendall Square, Bldg. 700, Cambridge, MA 02139.

848

Journal of the National Cancer Institute