Document 7Mvrp1B6G4xXRREoJ5KbEYq5V

__~ ~ ~ 194 Mortalii among Petroleum Workers Honda Jl - .. Y ,..a th c'l H An Updated Study of Mortality among Workem - m P', at a Petroleum Manufacturing Plant \b ft. le Yasushi Honda, MD Elizabeth Delzell, DSc Philip Cole, MD T his report is a further investigation of This study evaluated mortality among 9796 white male workers at a petroleum-manufacturing plant. The main pybose was to examine recent patterns i n leukmia mortality, for which an increase had been reported in a n earlier investigation. Compared to U.S. white m,the c o h d had a n excess of k k e m i a in 1940-1979 (38 obsen,ed/23 expected; standardized mortality ratio = 168; 95 5% confidence interval = 119-230). In t h 1980s, there was a deficit of IpuP (8 observed/l4 expected; standardized mortality ratio = 55; 95 o conjidence i n t m a l = 24-108). However, this was balanced @ an excess of myelofbrosis and myelavsphia (4 observed, < I expected). These results indicate that any-occupational leukmnogenic exposures at the plant have been reduced to a point at which thty are insuficient to cause leukemia. Hourly workers also had a n excess of deaths jkom mesothelioma in the 1980s (8 observed, about 2.5 expected), possibly because of exposure to asbestos in the past. cility began operations in 1918 and currently includes crude oil refining., fuel, diesel fuel, lubricating oils, as. phalt, and various petrochemical products, including benzene. Disease patterns among workers at the plan1 were evaluated previously in th(; retrospective follow-up in a case-control study of leuke a prospective investigation o era1 m ~ r b i d i t ya, n~d in a prog medical surveillance for l e u k e q among active and retired e m p l q ees5 The most recent retrospective fol- low-up study, which examined mor- tality from 1940 through 1984, found that the cohort had a 68% increase in leukemia during 1940-1979, basal on 38 observed and 23 deaths.* The excess appear minish in 1980-1984, but and expected numbers of were too small to be conclusive. present study was undertaken marily to evaluate more thorou the apparent disappearance of the earlier leukemia excess and to pro- vide an update of the mortality pat- terns among the workers. \c t1t $1' 1111 From the Department of Epidemiology. School of Public Health, University of Alabama at Birmingham. Birmingham, Alabama. Dr Honda's present address is Environmental Epidemiology Section. Environmental Health Sciences Division. National Institute for Environmental Studies, Tsukuba, Japan. Address correspondence to: Dr Elizabeth Delzell, Department of Epidemiology, UAB School of Public Health. 209 Tidwell Hall. University Station. Birmingham, AL 35294-0008. 1076-2752/95/3702-0194$3.OoiO Copyright by American College of Occupational and Environmental Medicine d. Methods The study included white who had worked at the plant least 6 months before 1990 an were either hourly and actively ployed at any time during I JOEM Volume 37,Number 2, February 1995 I. through 1989 or salaried and actively employed at any time through 1989. Hourly workers who left employ- ment before 1942 and salaried em- ployees who terminated before 1940 \here not included because their records had not been retained consis- iently by the company. Data files developed for previous f(dlow-up studies of this cohort'.: served as the primary source of iden- , tifying and work history information. ?be files contained the name, Social security number, race, gender, birth date, hire date, termination date, em- ployment status (active, retired, ter- minated, transferred), and payroll status (hourly, salaried) of subjects employed in or before 1983. We used computerized data from the company's health surveillance sys- tem to update through 1989 the work history of subjects actively employed at the end of 1983 and to identify and obtain information on persons hired shce 1983. To determine the vital status of , subjects as of January 1, 1990, we t used data from the previous fol- low-up study,* from company W r d s , from the Social Security Ad- ministration's Death Master file, and );om the National Death Index (XDI).and records of the Illinois and ql5souri divisions of motor vehicles. I h t h certificates of decedents were ':'!med from state vital records of- 'IQb or from the company and were ;(ded by a nosologist according to ninth revision of the International "lassification of Diseases (ICD). 'uhlWs were classified as presumed iiv% if they were actively employed active pension or division of -1 vehicle records in or after year in which the NDI was w b s h e d , and if they did not have tn ND1 or other death record. Sub- W s not classified as deceased or as "tsumed living were considered as ?[to follow-up on their employternination date. All subjects w follow-up had terminated em- before 1979. we CWtpared the mortality rates ;ri:Or..*_. subjects with the rates of the white male U.S. population, using the standardized mortality ratio (SMR) as the measure of association. The SMR is the ratio of the observed number of deaths among cohort members to the expected number, multiplied by 100. We computed expected numbers of deaths by multiplying the age- and calendar timespecific person-years (PY) of the study cohort by the corresponding U.S. population rates. For always salaried workers, PY accumulation began 6 months after the hire date 0% on January 1, 1940, whichever was later. For ever hourly workers, PY accumulation began 6 months after the hire date or on January 1, 1942, whichever was later. Follow-up ended on January 1, 1990, on the death date or on the date when the subject was lost to follow-up, whichever was earliest. We calculated 95% confidence intervals (CI) of the SMRs assuming a Poisson distribution for the observed numbers of deaths. The most recent version of a program developed by Monson6 was used for most analyses. This program was supplemented with leukemia described in our type-specific mortality rates from the Surveillance, Epidemiology and End Results (SEER) program to estimate expected numbers for the 19801989 time period. Results The Overall Cohort The total number of PY of follow-up and the average number of years per subject were 300,991 and 31, respectively, with the present update adding 30,997 PY to the previous study. Of the total 9796 subjects, 5895 (60%) were classified as presumed living, 3627 (37%) were deceased, and 274 (3%) were lost to follow-up. Of the 3627 deaths, 747 (31%) occurred during the update period. The death certificate and, hence, the cause of death were obtained for 96%of all decedents. Sev- 195 enty-nine percent of the cohort was ever hourly, 75% had a potential follow-up period of at least 30 years, and 53% had 10+ years of employment. The SMR for all causes during the period 1940-1989 was 79, based on 3627 observed deaths; and the SMR for all cancers combined was 88. based on 868 observed deaths (Table 1). Both of these SMRs represent statistically s i m c a n t deficits. For most noncancer disease categories, the observed numbetsf deaths was appreciably lower than expected. The cohort did not have a statistically significant excess of any specific type of cancer. The leukemia SMR for the overall study period was 123 (95% CI = 90-164). There was an excess of leukemia in 1940-1984 (44 observedl29 expected; SMR = 149, 95% CI = 108-200), as reported in the previous investigation of this cohort, but a deficit occurred in 19851989 (2/7.7; SMR = 26, 95% CI = 3-94). Among workers with relatively long-term employment at the plant (ie, 10+ years), mortality patterns were similar to those observed for the overall cohort (all causes SMR = 76, 95% CI = 73-79; all cancer SMR = 84,95%CI = 78-92). Also, the SMRs of men with relatively long potential induction time (ie, 20+ years since hire) did not indicate any unusual mortality experience (all causes S M R = 82, 95% C1 = 80-85; all cancer SMR = 90, 95% CI = 84-96). Houfly Workers Hourly workers had lower than expected mortality from most causes of death (Table 2). For the entire follow-up period, there was only a 9% excess of leukemia deaths (SMR = 109, 95% CI = 76-152). The overall result reflects a 36% excess of leukemia in the 1940-1984 time period and a deficit in 1985-1989, when hourly workers had no leukemia death, compared to 6.3 expected. Mortality among Petroleum Workers Hon TABLE 1 Observed Numbers of Deaths and SMRs for the Overall Cohort, 1940-1989 obs All causes 3627 All cancer Buccal cavity and pharynx Digestiveorgans Esophagus Stomach Large intestine and rectum Liver Pancreas Larynx Lung Skin Prostate Bladder Kidney Central nervous system All lymphopoietic cancer Lymphosarcoma Hodgkin's disease Leukemia Other lymphopoietic tissue Other cancer 868 8 215 15 28 116 12 35 10 280 19 74 26 23 22 104 14 9 46 29 87 Benign and unspecified neoplasms Diabetes mellitus Blood diseases Nervous system diseases Circulatory diseases Respiratory diseases Digestivesystem diseases Genitourinary system diseases External causes Other specified causes Unknown causes 95% confidence interval of SMR. 14 51 10 30 1822 203 122 39 249 77 142 SMR 79 95% CI' 76-81 a8 82-94 29 12-57 80 69-91 66 37-108 60 40-86 97 80-116 56 2S99 68 47-94 73 35-134 86 76-97 114 69-179 93 73-117 86 56-126 97 61-145 86 54-131 114 93-138 89 49-149 106 48-200 123 90-164 99 66-142 105 84-129 113 62-190 78 58-102 86 41-157 64 43-91 77 74-81 63 54-72 62 52-74 56 39-76 73 64-82 41 33-52 (4% of total deaths) -&A c * 'f TP I 1 i P Further details pertaining to leuke- an inverse relationship with duration occurred in 1940-1984 and mia mortality patterns are presented of employment. Larynx cancer 1985-1989. An approxim in a later section. deaths were increased among men pected number of Although there was a logdeficit with fewer t h h 20 years of ernploy- can be obtained b of cancer during the overall fol- ment (10/5.7), but a deficit (0/6.2) sum of the expected numbe low-up period (SMR-= 90;95% CI was present among men who worked phosarcoma, Hodgkin's disease I = 84-97), the cancer SMR was for 20 or more years. Review of the kemia, and other lymphopoietic tishigher in 1985-1989 (SMR = 102) death certificates of the 25 men who than in 1940-1984 (SMR = 87). In 1985-1989, there were moderate increases in deaths from larynx cancer (V1.9; SMR = 264, 95% CI = 85-617) and from cancer of other died of cancer of other and unspecified sites in 1985-1989revealed four deaths from endocrine cancer, one death from connective tissue cancer, two deaths from mesothelioma, and number of deaths in the all lympho poietic cancer category. The appmimate expected number of MF,/MDs deaths was less than 0.50 both feu - '. and unspecified sites (25/15; SMR = 18 deaths from unspecified types of 172, 95% CI = 112-255). Further cancer. examination of larynx cancer mortal- The "all lymphopoietic cancer" ity patterns indicated that the in- category in Table 2 included six creases during 1985-1989 probably observed deaths from myelofibrosis were not related to occupational ex- or myelodysplastic syndrome (MF/ posure, because there was, overall, MDS) in 1940-1989, two of which I JOhl Volume 37,Number 2, February 1995 -4 '' TABLE2 Observed Numbers of Deaths and SMRs for Ever Hourly Subjects,1942-1989 Obs All causes 3257 All cancer I Buccal cavity and pharynx Digestive organs Esophagus Stomach Large intestine and rectum Liver Pancreas Larynx Luns skin prostata Bladder Kidney Central nervous system All lymphopoieticcancer Lymphosarcoma Hodgkin's disease Leukemia i Other lymphopoietic tissue Other cancer Miand unspecified neoplasms Diabetes mellitus Blood diseases Nervous system diseases Circulatory diseases Respiratory diseases Digestivesystem diseases G e n i i n a r y system diseases Extemal causes Other specified ca Unknown causm 763 6 191 14 23 106 10 30 10 252 15 64 22 21 17 90 13 8 35 28 75 13 47 10 27 1661 190 105 35 213 71 122 (4%) - SMR 82 90 25 82 72 56 102 54 67 85 91 109 91 83 104 80 115 96 112 109 112 106 123 83 98 68 80 67 63 57 76 45 197 95% CI 79-85 84-97 9-55 71-94 39-120 35-04 83-123 26-100 441-156 60-103 61-179 70-1 16 52-126 64-158 47-128 93-1 42 51-164 48-221 7 6 152 75-163 83-1 33 65-21 0 61-110 47-181 45-98 77-84 58-77 51-76 39-79 66-87 35-56 two hourly workers, dying in 1982 and 1989, had "myeloproliferative disease" (ICDcode 238.7), specified hy the ICD classification as a "neoplasm of uncertain behavior." Finally. three other men had MF (n = 2 ) or preleukemia (n = 1)reported as a contributing, rather than underlying, cause of death. All of the men who died of MFMDS or myeloproliferative disease had started working iit the plant in or before 1945. The subject with preleukemia was hired in 1955. The finding of two mesothelioma decedents among hourly workers with "other and unspecified cancer" Prompted us to review the death certificates of all other cohort members to determine if there were addi- tional mesothelioma deaths. coded to other ICD categories. We found a total of nine mesothelioma deaths, eight of which occurred in or after 1980. The expected number of deaths could not be estimated because national mortality rates are not available for this type of cancer. However, using 1973-1980 and 1985-1988 SEER mesothelioma incidence rates and assuming that rates before 1970were one-half the 19731980 rates or lower, the expected number of cases was about 0.5 for the 1942-1969 time period. 1.5 for 1970-1979. and 2.5 for 1980-1989. The SMR was 200 (Cl = 92-380) overall and 320 (CI = 138-630) for the 1980s. These results may be un- derestimates of the true SMRs be- cause in the general population, disease incidence rates are higher than mortality rates. All of the subjects with mesothelioma were hourly; all but two started working at the plant before 1950; and several had jobs that may have entailed exposure to asbestos (eg, pipefitter, insulator, carpenter) (Table 3). Salaried Workers Salaried workers had fewer deaths than expected for all major disease categories. There was a small increase in lymphopoietic cancer deaths, because of a statistically significant excess of leukemia (11 ob- served/5.1 expected; SMR = 216, 95% C1 = 108-386), which was e---- TABLE 3 Characteristics of Subjects with Mesothelioma Year of death Am at death IC0 Code' 1972 56 199.1 1981 63 199.1 1984 68 228.0 1986 67 163.9 1986 73 162.9 1987 67 199.1 1987 a2 1987 1988 70 57 * Code assigned by nosologist. 162.9 163.9 199.1 Mortality among Petroleum Workers Honda Employment dates 1937-1967 1956-1981 1942-1971 1941-1980 1938-1966 1947-1954 1935-1962 1945-1946 1953-1958 4c i1 I` Plant job titles Pipefitter ---. 1 Carpenter, maintenance Instrument repairman Yardman, insulator Various jobs, lube comp. and ship: laborer and yardman, engineering field: drum reconditioner and- i ``ic other jobs, compounding Fireman helper, boiler house; extra man, lube filters: extra man, clay handler and clay burner; opera- tor helper, lube-oil filters Tester, control lab, expenmental I lab Fireman helper; boiler washer Pipefitter helper a I restricted to the 1940-1984 time pe- For the overall cohort, all of the riod (9 observedl3.7 expected; SMR leukemia excess occurred before = 242, 95% CI = 110-460). 1980 (SMR = 168; 95%CI = 119- may be attributable in part to Leukemia, Mortality, Patterns As shown in Table 4, the slight overall increase in leukemia among hourly workers was concentrated 230). For the period 1980-1989, there were fewer observed than expected leukemia deaths ( S M R = 55, 95%CI = 24-108). The deficit was due to the occurrence of only four employment of relatively healthy individuals by the industry and to socioeconomic benefits associated with employment. However, the persis- among subjects who were under 45 leukemia deaths among hourly work- periods of follow-up and the con& years of age at the time of death (SMR = 211, 95% CI = 77-460), who died before 1970 (SMR = 188, 95%CI = 115-290), who were hired before 1940(SMR = 123,95%CI = 74-192), who worked at the plant for 10+ years (SMR = 123, 95%CI = 81-173, and who died 10-39 y5ars after starting work (SMR = 154, 95% CI = 102-222).-All six of the ers, compared to 12 expected. Salaried workers had slightly more leukemia deaths than expected in 19801989 (4/2.4), but this difference was based on small numbers. When leukemia mortality in 1980-1989 was examined by cell type, again, there were no excesses. Cell type-specific observed/expected numbers were 3/4.5 for lymphocytic leukemia (in- tency and strength of the observed mortality deficits suggest that petro. leum refinery and related workers are not exposed to strong or pem sive health hazards. This study found persuasive dence that the leukemia served among the workers period 1940-1979 was I men who were under 45 years of age cluding one subject with hairy cell present in the 1980s. In fa at the time of death died before 1970, leukemia), 4/6.1 for myelocytic leu- of leukemia was observed for the five of them before 1960 (SMR = kemia, 4/41 for acute myelocytic latter time period among hourl! 303, 95% CI = 98-707). The in- leukemia, and U3.9 for unspecified workers. The recent deficit does not crease in leukemia among salaried leukemia. appear to be attributable to inax& workers was concentrated among plete vital status ascertainment, ~IW men who died at age 45 and older (SMR = 256, 95%CI = 128-459), Discussion much as there was no consist* pattern of lower SMRs in 1980- who were hired before 1950 (SMR The results of the present study 1989, compared with 1940- = 275, 95% CI = 132-505), who and of other investigations of petro- Instead, the apparent leukemi worked for IO+ years (SMR = 306, leum industry workers consistently cit may be due to chance 95% CI = 123-630), and who died indicate that such workers have over- problems with certifying le 10+ years after hire (SMR = 236, all mortality rates that are moder- the underlying cause of 95% CI = 118-422). ately lower and cancer rates that are discussed later in relation 3t at -5 P Volume 37,Number 2,February 1895 Expected Leukemia Deaths among White Men by Selected Characteristics ObscKved/expected deaths Salaried Overall Age at death t45 45-54 55-64 65+ Period of death (1960 1960-1969 1970-1979 1980-1989 W2.8 a.7 m.4 17118 94.6 1V6.1 1m.2 4/12' 010.8 310.8 211.3 6/2.2' 210.5 010.8 511.4' 412.4 Period of hire 4940 1940-1 949 1950+ Years worked 0.5-9 10-19 20+ Years since hire <lo 10-19 . * 20-29 - 30-39 40+ 0.05. ** 19/15 12/13 m.3 8/10 7/59 2Wl6 V1.6 7B.1 8/59 139.2 6/12 w1.8' 411.9 111.4 412.8 310.9 411.4 0/0.5 W0.7 211.o 311.4 411.6 199 Total 46/37 613.6 714.5 10/8.7 23120 1115.1. 1116.9 16/11 8/14 25/17 16115 514.7 12/13 1016.8 24117 112.1 9/3.8* 10/6.9 16111 10114 $IDS.On balance, the results for itukeemia imply that any occupafHMal leukemogenic exDosures that i'\ %ht have bee: presen; at the plant have been reduced to the point at 1 I' they are insufficient to cause 'g i b m i a . I1 follow-up studies among pe' 'deUm workers, in addition to this increased leuke- statistically signifiy one of the studies (Mor- Wong, unpublished observaight additional studies have leukemia mortality rates lar to or lower than the e general population unpublished observa- tions).12-'* None of the other studies with positive results for leukemia covers the entire 1980-1989 decade. Therefore, it is not possible to determine if the trend of reduced leukemia mortality in the 1980s observed in the present study is generalizable to other groups of petroleum industry workers. Some petroleum industry studies have reported excesses of various cancers other than leukemia, including central nervous system cancer, kidney cancer, skin cancer, lymphosarcoma, and "other lymphopoietic tissue" cancer.' The present investigation did not find any consistent excess of any of these cancers. Kidney cancer deaths were slightly elevated among hourly workers in 1985-1989,but the increase was lim- ited to short-term workers. The present study confirms an ear- lier report,' which suggested that the employees have a higher rate of MF/ MDS than expected. The increase appears to be related to the same time period of employment as the leukemia excess, but the reason for the excess is unknown. MF/MDSs share some clinical and pathological features with certain types of leukemia, and some cases of MF/MDS have been reported to coexist with or to progress to l e ~ k e m i a . ' ~T*hu~s, MFMDS and leukemia may be part of a single disease process. In addition, there may have been a local shift in diagnosis or reporting practices in the 1980s, with some persons i .L i Mortality among Petroleum Workers Honda designated as having MF/MDS, who. if diagnosed in earlier time periods, would have been designated as having leukemia. This interpretation is petroleum industry workers (Hornstra, unpublished observations). 's Acknowledgment ried workers. 4rn J Eprdemrol 1983:tla 516-541. 10. Divine BJ, Barron V, Kaplan SD. mortality study. I. Mortality among ery, petrochemical and research w ~ ' I i ,e ' supported by the fact that in the This study was supported by a contract J Occup Med. 1985;27:445-447. 1980s, the excess of MF/MDS (4 with the Shell Oil Company. 11. Bertaizi PA, Pesaton AC observed/O expected) was closely balanced by the deficit of leukemia (8 observed/l4 expected) during the same decade. That is, if leukemia 1. McCraw DS, Joyner RE, Cole P. Excess leukemia in a refinery population. J Oc- al. Mortality study of oil refinery workers. Environ Health. 1989 12. Rushton L,Alderson MR. An logical survey of eight od and MF/MDS are considered as a cup Med. 1985;27:220-222. Britain. Br J Ind Med. 1981; single disease category, the combined observed number of deaths is approximately equal to the expected number. Another follow-up study of workers from various petroleum re- 2. Wongsrichanalai C, Delzell E. Cole P. Mortality from leukemia and other diseases among workers at a petroleum refinery. J Occup Med. 1989;31:106-111. 3. Austin H, Cole P, McCraw DS. A casecontrol study of leukemia at an oil refin- 13. Rushton L. Alderson MR. Epldemi,+ iBcnaltasiunr.vBeyr JofIenidghMteddi.st1ri9b8u3ti;o4n0:3ce3n1)te-3n3 14. Hank NM, Shallenberger LG. Dona& DL, et al. A retrospective mortality S& of workers in three major U.S. refin& ,I Lbl,,r:( t '.,, ~ 1 ('o' fineries reported 3 observed compared to 1.5 expected MF deaths in ery. J Occup Med. 198628:1169-1173. 4. Tsai SP, Dowd CM, Cowles SR. Ross and chemical plants. Part I. Cornpartma I h l t s i t h U.S. population. J Occup tvd ' Ill(. the time period before 1981.'' Other CE. Morbidity patterns among employ- 1985;27:283-292. investigations of petroleum industry ees at a petroleum refinery. J Occup Med. 1991;33:lq76 -1080. workers have not reported results 5. Cowles SRdBennett JM. Ross CE. Med- separately for MF or MDS. ical surveillance for leukemia at a petro- 43~6-17. .I S Whereas only one death from me- chemical manufacturing complex: four- 16. Nelson NA, Van Peenen PFD, sothelioma occurred among the em- year summary.J Occup Med 1991;33: AG. Mortality in a recent ( rnr Cod edit1 ployees in the period 1940-1979, there were eight such deaths, constituting an apparent excess, in the 1980s. The job histories of the men with this cancer indicate that several of them had potential for exposure to asbestos during their employment, 808-812. 6. M o w n RR. Analysis of relative survival and proportional mortality. Compur Biomed Res. 1974;7325-332. 7. Delzell E, Austin H, Cole P. Epidemiologic studies of the petroleum industry. Occup Med State Art Rev. 1988;3:455474. cohort. J Occup Med. 1987;29. 17. Marsh GM, Enterine PE, M Mortality patterns among pe finery and chemical plant worke Ind Med. 1991;19:29-42. 18. Kaplan SD. Update of a m of workers in petroleum Occup Med 1986;28:514-5 in i 10Ui Tt rtn dnd \ion The and the observed excess may be due to this agent, which is an established cause of mesothelioma. Two other investigations also have reported what appears to be an unusual number of mesothelioma deaths among 8 . Divine BJ, Barron V. Texaco mortality study. 111. A cohort study of producing and pipeline workers. Am J Ind Med. 198%11:189-202. 9. Wen 0,Tsai SP, McClellan WA, et ai. Long term mortality study of oil refmery workers. I. Mortality of hourly and sala- 19. Goasguen E,Bennett JM. Class and morphologic features of the dysplastic syndromes. Semin 1992;19:4-13. 20. List AF, Jacobs A. Biology genesis of the myelo-dysp dromes. Semin Oncol. 1992;l sigh [her( A st'rvt initi. mitt pres< sion 1'992 patir LIgo I Idea. ilten "en