Document Gm5wa9rKay8k16XVNKBp8ZBmv

American Journal of Industrial Medicine 22:209--229 (1992) A Retrospective Mortality Study Within Operating Segments of a Petroleum Company A. Robert Schnatter, Dr ph, msc, Gilles Theriault, md, Dr ph, Arnold M. Katz, md, F. S. Thompson, rn, Donna Donaleskl, mph, and N. Murray, bs, ctH This retrospective mortality study was conducted among 34,597 oil industry workers in diverse operating segments. Employees were traced through Statistics Canada, and overall mortality (SMR = 0.85) was lower than general population rates and similar to other petrochemical cohorts. The most notable finding was a significant excess of malignant melanoma [observed deaths (N) = 16, SMR = 1.87, 95% Cl = 1.07,3.04], which concentrated among upstream workers (N = 6, SMR = 6.00, 95% C3 = 2.19, 13 06), and was directly related to employment duration and latency. Specific sub stances or hydrocarbon (HC) streams could not be implicated, although possible expla nations include dermal HC exposure, ultraviolet light exposure, or a synergistic effect between these two factors. Marketing/transpartation workers showed a non-significant excess of multiple my eloma (SMR = 1.81), which was also related to employment duration, latency, and commencement of employment before 1950. Lymphatic cancer, skin cancer, and kidney cancer mortality was not elevated in refinery workers, a finding at odds with some previous refinery worker studies. Although the malignant melanoma and possibly the multiple myeloma mortality patterns are consistent with an occupational link, further studies are needed to investi gate the relationship of these diseases with particular exposures. 1992 wiiey-Liss, Inc *y words: hydrocarbon, oil, gasoline, malignant melanoma, multiple myeloma, aortic aneurysm, cancer, occupational exposures, refinery workers, female mortality INTRODUCTION Mosr studies of the oil industry have been restricted to refinery workers [Delzell et al., 1988; Wong and Raabe, 1989) and have not examined other operating seg ments such as marketing^distribution, pipeline, or production workers. Exposures are clearly different for workers in different petroleum operating segments [Runion, Occupational Health and Epidemiology Division, Exxon Biomedical Sciences. Inc., East Millstone, NJ (A.R.S.., D D ), School of Occupational Health, McGill University, Montreal, Canada (G.T). Occupational Health Division, Imperial Oil, Limited, Toronto, Canada (A.M.K., F.S.T., N M ). Address reprint requests to Dr A Robert Schnatter, Exxon Biomedical Sciences, Inc,, Mettlcrs Road, CN 2350. East Millstone. NJ 08875-2350 Accepted for publication December 26, 1991.. 1992 WUey-Liss, Inc. EXX-MOR-003410 210 Schnatter et al. 1988], Only Divine and Barron [1987] and Rushton and Alderson [1983] have studied these latter operating segments. The International Agency for Research on Cancer (IARC) has recently classi fied refinery operations as probably carcinogenic to humans [IARC, 1989] based on mortality findings for leukemia and skin cancer. Kidney cancer is also an issue, due to findings for wholly vaporized unleaded gasoline in male tats [Kitchen, 1984], but this may be due to a species and sex-specific protein not present in humans [Health Effects Institute, 1988]. The major objective of this study was to examine the mortality experience of a large Canadian cohort of oil industry workers in the context of other similar studies. In particular, the risk of leukemia and other lymphatic cancers, malignant melanoma, and kidney cancer was focused upon. A second study objective was to describe mortality in relation to the different exposures prevalent in operating segments within the petroleum industry. A subsequent report will focus on hydrocarbon exposure. METHODS Population Definition This study is a retrospective cohort mortality study. Some study subjects were previously studied by Hanis et a], [1919]. The study population consists of a) all active employees and living annuitants (i.e., retirees) on 1 January 1964 and b) all new regular employees hired between I January 1964 and 31 December 1983. One year of work was necessary before an employee, male or female, was included in the cohort. This study population differed from that of Hanis et al. [1979] in that all employees are included instead of males only, and a 1 year employment criterion is imposed in lieu of 5 years for terminated workers and 1 year for active workers. Work History Summarization Work histories for each cohort member were obtained from computer records, generally available from 1960-19641 forward. For employees who worked before this time (55%), the first computerized job entry was extrapolated back to their employment date. For annuitants who terminated before 1960-1964, the last job was coded and is extrapolated back to the employment date. The work histories were used to derive operating segment (e.g., marine, marketing, etc.) and presumed exposure to hydrocarbons (HCs) for each employee. Operating Segment Definition Three experienced industrial hygienists reviewed all unique location/department combinations from the work histories. They assigned each entry to one of the fol lowing segments: a) refinery, b) agricultural chemicals, c) petrochemical, d) market ing-transportation, e) marine, f) exploration/drilling/production, g) pipeline, h) coal and minerals, i) office, j) fabricated building products subsidiary, and k) other/ unknown. These were chosen to exploit the differences in exposure patterns particular to each segment. For example, pipeline workers and drilling workers can be dermally 11960-1964 are approximate dates for the start of work histories. A small percentage of employees had computerized work history before this dare, while another very small percentage started after 1964.. Mortality in Petroleum Operating Segments 211 exposed to crude oil and drilling muds Most marketing terminal workers are nor exposed to crude oil but can be exposed to finished petroleum products on an inter mittent basis, most likely through inhalation. Refinery workers can be exposed to a wider range of hydrocarbons including crude oil, intermediate streams associated with processes such as alkylation or reforming, and/or finished product [see Runion, 19SS for more detail]. Hydrocarbon Exposure Definition Eight industrial hygienists supplied a score for exposure to HCs for each unique work history entry in their region of responsibility. Scores were defined as exposed to HC, not exposed to HC, and unknown. This resulted in a matrix of scores for each location/job/department which was extrapolated to each cohort member. HC exposure represents dermal and inhalation exposure to many different streams and/or products, and includes combustion products. For this paper, analyses were conducted on an ever/never exposed basis. If an employee was ever assigned to a job/location/depart ment combination deemed exposed, all subsequent person-years are counted as ex posed. For most analyses, the HC-exposed subcobort shows mortality patterns similar to that of all workers. For brevity, we will not routinely present results for the HC-exposed subcohort, but will note where mortality patterns in this group are different from the total cohort. Mortality Tracing Mortality information was obtained through three sources: a) the Statistics Can ada (SC) Mortality Data Base; b) internal data sources (including provincial registrars used in the previous study); and c) the National Death Index,, Records on all em ployees doi actively employed on 31 December 1983, nor receiving company benefits were sent to Statistics Canada (SC), SC identified 3,764 total deaths (96.3% of the total), and the U.S. National Death Index (NDI) also identified 19 deaths (0.5%). Another 128 decedents were identified through company sources or the first study. Company records for five employees indicated that they were alive and receiv ing benefits, yet they were identified as decedents by SC. Two of these potential decedents with relatively low SC matching scores [see Smith and Newcombe, 1980] were counted as withdrawals; their date last employed was counted as the date last observed alive. Thus, of the 3,764 deaths identified by SC, 3,762 were counted as decedents. Employee records not matched to SC or NDI files were assumed alive until the end of the follow-up period. The validity of this critical assumption was assessed previously [Schnatter et al.., 1990J. This test showed that SC blindly ascertained 97.6% of the known Canadian deaths from this cohort. Importantly, there were no strong determinants of death ascertainment such as province or age of death, indi cating a uniform underascertainment of 2.4%. This number provides the best estimate of negative bias on the SMRs due to death under ascertainment. Determination of Underlying Cause of Death The revision of the International Classification of Diseases (ICD) coding scheme in effect al the time of death was used to assign the underlying cause of death. For the 3,762 decedent records in which an SC death code was available, it was used. 212 Schnatter et tl. TABLE I. Total Cohort of Ail Operating Segments in a Petroleum Company by Gender and Vital Statue on 12/31/83 Vital status Gender Mates Females Total Alive 22,644 8,042 30,686 Dead 3,702 207 3,909 Withdrawn 2 0 2 Total 26.34S 8,249 34,597 For the 128 records in which only a code from the First study was available, it was used. For the 19 death certificates obtained through NDI, a certified nosologist determined the underlying cause of death. Records for 39 (i.0%) decedents with no underlying cause of death are included as deaths for total mortality but are not assigned to a cause specific category. Owing to the presence of asbestos and vinyl chloride monomer at particular study locations, an SC-appointed nosologist reviewed all death certificates with codes in the Appendix for any mention of mesothelioma and/or angiosarcoma. Mortality Analyses Employees contributed person-years from 1 January 1964 or date of employ ment plus l year, whichever was later, Person-years were enumerated via the Monson program [Monson, 1974] to the earliest of the following dates: date of death, the end of the follow-up period (31 December 1983), or date of last employment for em ployees designated as withdrawals. Canada's Laboratory Centre for Disease Control (LCDC) supplied mortality rates and ICD/LCDC rate translation tables. If an em ployee ever worked in one of the operating segments, subsequent person-years and mortality are counted in that segment. Person-years can be counted in more than one segment. This is generally preferred to other strategies which examine most frequent, first, or last assignment [Checkoway et aL, 1989], since it maximizes the personyears in each segment. However, this strategy can lead to misclassification for short term employment in an operating segment. Expected deaths were calculated by multiplying Canadian population mortality rates by the study population person-year distributions specific for gender, quinquen nia of age, and calendar period. SMRs were computed provided that either the observed or expected number of deaths was at least 5.0. Ninety-five percent confi dence intervals were calculated for the SMR according to the formula of Byar, referenced in Rothman and Boice [1979]. For some analyses, we computed SMRs and exact confidence intervals according to Fisher [Rothman and Boice, 1979], when observed and expected deaths were less than 5 0. RESULTS Overall Cohort A total of 34,597 employees contributed 428,190 person-years from 1964 through 1983. The distribution of persons by gender and vital status is displayed in Table I. For the total cohort, 20,928 or 60% of employees entered after 1964, indicating that this is a relatively young cohort There are 19,565 additional employ ees and 346,133 additional person-years at risk (PYAR) in this cohort, compared to the previous study. Mortality in Petroleum Operating Segments 213 A total of 3,909 deaths occurred compared to 4,604 expected deaths correoonding to an all-cause SMR of 0.85 [95% confidence interval (Cl) = 0.82, 0.88; able II]. For the overall cohort, there are significant deficits for most broad nonmalignant disease categories, as well as all cancers. Significant excesses are present for malignant melanoma [observed deaths (N) = 16, SMR = 1.87, 95% Cl = 1.07, 3.04] and aortic aneurysm (N = 69, SMR = 1.43, 95% Cl = 1.11, 1.81). A larger, less precise SMR is present for benign neoplasms (N = 8, SMR = 1.96, 95% Cl = 0.84, 3.86). For female employees (Table HI), all-cause mortality is significantly low (N = 207, SMR = 0.73). No cause of death category is significantly elevated but there are slight excesses of lung cancer, cirrhosis, and diseases of the arteries and capillaries. There are deficits in most major disease categories, while cancer mortality is about as expected (N = 80, SMR = 0.96). The balance of this paper focuses on male employees. No angiosarcomas and seven mesotheliomas (six pleural, one unspecified) were identified. Based on Canadian rates available from the literature [McDonald and McDonald, 1980] the expected number of deaths in this cohort is 1.24, The resulting SMR of 5.65 is significant (95% Cl = 2.27, 11.63). The average starting employment year for these seven decedents is 1933 (range 1927-1943), the average duration of employment is 35 years (range 29-42), and all decedents died between 1970 and 1980. Examination of the job histories suggests many (e.g., pipefitter and mechanic) in which asbestos exposure would be likely, although we did not estimate past asbestos exposure. Five of these seven cases worked at a single refinery.''petrochemical plant complex, and six of the seven worked in refineries at some point in their careers. Mortality patterns by employment duration are shown in Table IV for selected causes of death. Table IV presents mortality patterns by employment duration for selected causes of death. Latency periods of 10 years (shown) and 20 years (not shown) were considered. For malignant melanoma, SMRs are higher for longer employment duration subgroups, especially the 15-24 year subgroup. SMRs are very mprecise for the 35 + employment group, however. The patterns are similar when a 20 year latent period is considered. For leukemia, kidney cancer, and nonHodgkin's lymphoma, however, there are no apparent trends with employment du ration, and considerable overlaps in the 95% confidence intervals. Operating Segments ' The following operating segments are not discussed in detail since very few deaths (N) occurred: agricultural chemicals (N = 13), building products (N = 3), coal/minerals (N -- 2), and unknown (N = 30). Office workers contributed 384 deaths, but no disease categories showed significant excesses, while significant def icits were present for most major nommalignant cause of death categories, as well as total malignancies. Refinery locations. This operating segment is the largest, consisting of 9,276 male workers, of which 7,437 (80%) were ever employed in an HC-exposed job. Overall mortality is slightly below national rates (N = 1805, SMR = 0.91, 95% Cl = 0 86, 0.95, see Table V), as are most major disease categories. Significant deficits of ''other heart disease" (SMR = 0 65) and acute myocardial infarction (SMR = 214 Schnatter et >1. TABLE It Mortality Remits 1964-1983 in Total Cohort of All Operating Segments in a Petroleum Company (N = 34.5971____________ ____ __________ ______ ___________________ _____ 95% Conf. Int. Cause of death [1CD codes. 8th) Alt causes (001-999) Infectious Dx1 (001-136) Endocrine Dx (240 -279) Blood Dx (280-289) Circulatory Dx (390-458) Respiratory Dx (460-519) Digestive Dx (520-577) Genitourinary Dx (580-629) Symptoms/ill-defined (780-796) Accidents'violence (800-999) Malignant neoplasms (140 -209) Esophagus Cab (150) Stomach Ca (151) Large Intestine Ca (153) Rectum Ca (154) Liver Ca (155) Pancreas Ca (157) Broncfcus/Lung Ca (162.1) Bone Ca (170) Malignant melanoma (172) Breast Ca (174) Cervix uteri Ca (180) Corpus uteri Ca (182.0) Ovary Ca (183 0) Prostate Ca (185) Kidney Ca (189.0-189 2) Bladder Ca (188) Brain (malignant) Ca (191) Nervous Svstem (ail) (191-192. 225. 238-239) Rcticulum-ccll sarcoma (200.0) Lymphosarcoma (200. U Non-Hodgkin's lymphoma (200.202) Multiple myeloma (203) Leukemias (204-207) Benign neoplasms (210-228) Diabetes (250) Cerebrovascular Dx (430--138) Mvocardia! infarction (410) Other heart Dx (420-429' Dx of artcries/capillary (440--148) Aortic aneurysm (441) Cirrhosis (571) Dx of pancreas t577) Kidney diseases (580-593' Observed 3.909 17 65 10 2.085 250 143 48 26 239 911 22 57 94 39 8 48 243 6 16 20 3 3 6 89 23 26 25 32 9 6 30 12 39 8 53 333 700 78 156 69 62 10 33 Expected 4,603.6 28.6 86.7 ` 11.2 2.309.0 321.5 189.5 64.1 36.8 421.9 1,025.9 22.3 83.1 93.7 41 4 8.6 57,7 283.1 44 8.6 2! 3 4.0 1.9 57 8S.7 23.4 32.9 25.5 32.5 62 8.7 27.7 13.8 37.7 4.1 70.9 374.5 816.8 109.2 137.0 48.3 86,2 86 46.7 SMR 0.85 0 60 0.75 0 90 0-90 0.78 0.76 0-75 0-71 0.57 0.89 0.99 0 69 1.00 094 0.93 0-83 0.86 1.37 1.87 0 94 -- -- 1 05 1.04 0.98 0.79 0.98 0.98 1.45 0.69 1.08 0.87 1.03 1 96 0.75 0,89 0 86 0.72 1.14 1 43 0.72 1.16 0.71 Lower limit Q..S2 0.35 0.58 0.43 0.87 0.68 0.64 0.55 0 46 0.50 0 S3 0.62 0.52 0.81 0.67 0.40 0,61 0.75 0.50 1.07 0.57 -- -- 0.38 0,83 0 62 0.52 0.64 0 67 0 66 0.25 0.73 0,45 0 74 0.84 0.56 0..80 0,80 0.57 0.97 1.11 0.55 0.56 0.49 Upper limit 0.88 0.95 0.96 1.65 0.94 0 88 0.89 0.99 1.04 0.64 0 95 1.50 0.89 l 23 1.29 1.83 1 10 0.97 2 97 3.04 I 45 -- -- 2.28 1.28 1.47 1.16 1.45 1.39 2.75 1.51 1.55 1.52 1 41 3.86 0.98 0 99 0 92 0.89 1.33 1.81 0.92 2.14 0.99 aDx = Disease/Diseases hCa = Cancer Q 89) were offset by a statistically significant excess for diseases of arteries, capil laries and veins (SMR = 1 34), This category subsumes aortic aneurysms, which are also elevated iN = 32. SMR = 1-34, 95% Cl = 0 98, 2.02). No malignant EXX-MOR-003415 Mortality in Petroleum Operating Segments 215 TABLE III. Mortality Results 1964-1983 in Ail Females of All Operating Segments In a Petroleum Company (N = 8,249) Cause of death CICD Codes, 8th) AJ1 causes (001-999) Infectious Dx (001-136) Endocrine Dx (240-279) Blood Dx (280-279) Circulatory Dx (390-458) Respiratory Dx (460-519) Digestive Dx (520-577) Genitourinarv Dx (580-629) Symptoms/ill-dcfined (780-796) Accidems/violence (800-999) Malignant neoplasms (140-209) Esophagus Ca (150) Stomach Ca (151) Large intestine Ca (153) Rectum Ca (154) Liver Ca (155) Pancreas Ca (157) Bronchus/lung Ca (162 I) Bone Ca (170) Malignant melanoma (172) Breast Ca (174) Cervix uteri Ca 080) Corpus uteri Ca 082.0) Ovarv Ca (183.0) Kidney Ca (IB9.0-J89..2) Bladder Ca 088) Brain (malignant) Ca (191) Nervous system (ail) (191--192, 225, 238-239) Reticulum-ccll sarcoma (200.0) Lymphosarcoma (200.1) Non-Hodgkin's lymphoma (200,202) Multiple myeloma (203) . Leukemias (204--207) Benign neoplasms (210-228) Diabetes (250) Cerebrovascular Dx (430-438) Myocardial infarction (410) Other heart Dx (420-429) Dx aneries/capillary (440-448) Aortic aneurysm (44!) Cirrhosis (571) Dx of pancreas (577) Kidney diseases (580-593) Observed 207 0 4 0 78 7 12 3 2 11 80 2 2 8 2 0 3 11 0 1 20 3 3 6 1 1 2 2 1 1 3 1 3 1 3 20 15 5 8 3 7 0 3 Expected 285.0 2.1 8.2 1.0 114.1 13.2 12.6 3.9 2.6 32.2 83.5 0.8 3.9 8.8 2.5 0.5 36 7.9 0.3 1.1 20.1 40 1.9 5.-J 1.3 0.9 2.6 34 0.5 0.7 2.4 L.O 3.2 0.6 6.6 27.4 33.2 7.3 6.5 1.3 54 D.6 3.3 SMR 0.73 -- 0.49 -- 0.68 0.53 0.95 -- -- 0.34 096 -- -- 0.90 -- -- -- 1.39 -- -- 1.00 -- -- 3.05 -- -- -- -- -- -- -- -- -- -- 0 46 0.73 0 45 0.69 J .23 -- 1 29 -- -- 95% Conf. lnt. Lower limit Upper limit 0.63 -- 0.13 -- 0.54 0.21 0.49 -- -- o.n 0.76 -- -- 0.39 -- -- -- 0.69 -- -- 0.6 __ 10.83 1.24 __ 0 85 1.09 1.66 __ -- 0.61 1 19 -- -- 1.78 __ _ -- 248 -- 1.54 __ 0.38 -- -- -- -- -- _ -- -- -- 0.09 0.45 0.25 0.22 0.53 -- 0.52 -- -- 2.28 _ __ -- __ -- -- -- -- -- -- 1.33 1.13 0 75 1.60 2 42 -- 2.66 -- -- neoplasm subsite was statistically elevated, except mesothelioma (N = 5, SMR -- 1111, 95% Cl = 3-61, 25.93). The benign neoplasms category shows an elevated SMR of 3.24, but it is very EXX-MOR-003416 216 Schnatter et al. TABLE rV. Mortality Patterns by Latency and EmpAoyment Duration for Selected Causes of Death In Male Workers of All Operating Segments of a Petroleum Company (N= 26,348)* Cause of death Malignant melanoma Leukemia Kidney cancer Non-Hodgldn's lymphoma Years employed 0-4 5-14 15-24 25-34 35 + 0-4 5-14 15-24 25-34 35 + 0-4 5-14 15-24 25-34 35 + 0-4 5-14 15-24 25-34 35 + Observed 0 0 9 3 2 0 I 8 14 8 0 1 7 9 4 1 2 5 14 5 Expected 0.28 0.77 1.96 2.32 1..04 ' 0.57 I 96 8,42 13.57 7.09 0.27 1.10 5.79 9.15 5.08 0.45 L56 6.45 9 63 5.31 SMR 0.0 0,0 4.59 1.29 1.92 0.0 0 51 0.95 1 03 1.13 0.0 0.91 1.21 0.98 0.79 2.22 1.28 0.78 1.45 0.94 95% Conf. Ini. 0-13.18 0-4.79 2.10-8.72 0.27-3.78 0.23-6.95 0-6.47 0 12-2 84 0.41-1.87 0.56-1 73 0.48-2.22 0-13.66 0.23-5 07 0.48-2.49 0.45-1 87 0.22-2.02 0,56-12.38 0.16-4.63 0.25-1.81 0.79-2.44 0.30-2.20 'All results for a latency period of 10 years. imprecise based on a wide 95% Cl (1 04, 7.55) and is not significant (N -- 3, SMR = 2.52, 95% Cl = 0.52, 7.37) in exposed workers. The anatomic sites included two large intestine tumors and one unspecified tumor of the digestive system, but the malignant large intestine cancer SMR (0.91) was not elevated in refinery workers. For cancers of interest from previous studies, leukemia mortality is not elevated (SMR = 0.84). Reticulum cell sarcoma (RCS) rates are elevated (N = 6, SMR - 2.44, 95% Cl = 0.89, 5.31), but no patterns with latency or duration of employment are evident. Interestingly, there are no lymphosarcomas (LSA), which are commonly grouped with RCS in mortality studies. Non-Hodgkin's lymphoma (viz. RCS, LSA, and unspecified lymphomas) shows a moderate, non-significant elevation (N = 14, SMR = 1.27, 95% Cl = 0.69, 2.13). There is also a deficit of multiple myeloma (SMR = 0.51). There were very few observed and expected malignant melanoma deaths, resulting in an imprecise SMR of 1.01. The kidney cancer SMR is not elevated (N = 9, SMR = 0.91, 95% Cl = 0.42, 1.73). Marketing/transportation locations. This operating segment consists of 6,823 male employees, of whom 5,041 (74%) are classified as ever exposed (Table VI). These workers are potentially exposed to finished product (viz. gasoline, diesel, and heating fuel), primarily via inhalation. Mortality relative to the general popula tion is again significantly low (N = 1157, SMR = 0,88, 95% Cl = 0.83, 0.93) as are most major disease categories. There was an excess of aortic aneurysms (N = 25, SMR = 1.78, 95% Cl = 1.15, 2.63), which was decidedly more pronounced in non-exposed (including unknown exposure) marketing workers (N = 13, SMR = 2 64 , 95% Cl = 1.40, 4.51). The SMR was 1.31 among exposed workers and did not show trends with employment duration and latency. Mortality in Petroleum Operating Segments 217 TABLE V. Mortality Results 1964-1983 in Refinery Workers of a Petroleum Company (N = 9,276) 2ause of death (1CD Codes. 8th) . AL causes (001-999) Infectious Dx (OGl-136) Endocrine Dx (240-279) Blood Dx (280-289) Circulatory Dx (390-458) Respiratory Dx (460-519) Digestive Dx (520-577) Genitourinary Dx (580-629) Svmpcoms/ill-defined (780-796) Accidents/violcnce (800-999) Malignant neoplasms (140-209) Esophagus Ca (150) Stomach Ca (151) Large intestine Ca (153) Rectum Ca (I54j Liver Ca (155) Pancreas Ca (157) Bronchus/lung Ca (162.1) Bone Ca (170) Malignant melanoma (172) Prostate Ca (185) Kidney Ca (189 0-189.2) Bladder Ca (188) Brain (malignant) Ca (192) Nervous system tali) (191-192. 225. 238-239) Reticulum-cell sarcoma (200.0) Lymphosarcoma (200.1) Non-Hodgkin's lymphoma (200,202) Multiple myeloma (203) Leukemias (204-207) Jehign neoplasms (2)0-228) Diabetes (250) Cerebrovascular Dx (430-438) Myocardial infarction (410) Other Heart Dx (420-429) Dx arteries/capillary (440-448) Aortic aneurvsm (441) Cirrhosis (571) Dx of pancreas (577) Kidney diseases (580-593) Observed 1.805 8 37 8 1,008 133 61 22 17 82 388 7 29 36 22 4 20 115 3 3 46 9 14 9 i2 6 0 14 3 13 5 34 160 315 31 87 32 27 5 24 Expected 1.994.3 12 1 36.6 4.8 1,047.4 150.4 77.7 29.8 15.5 147.0 430 9 98 37.4 39.6 28.2 3.6 24.8 122-8 1.8 3.0 43.1 9.9 15 6 92 11.8 2.5 3.6 11.0 5.9 15 5 1.5 30.3 1725 353.6 47.4 65.2 22 4 32,6 34 20.9 SMR 0 91 0.66 1.01 1.66 0.96 0.88 0.78 0.74 l 10 0.56 0.90 0.71 0.78 0.91 1.21 -- 0.81 0 94 -- -- 1.05 0.91 0.90 0.98 1.02 2.44 -- 1.27 0.51 0.84 3.24 1.12 0,93 0.89 0 65 1.34 1.34 0,83 -- 0.67 95% Conf. Int. Lower limit Upper limit 0.86 0.2E 0 71 071 0 90 0.74 0.60 0 46 0.64 Q 44 0 81 0.29 0.52 0 64 0.76 -- 0 49 0.77 -- -- 0.77 0 42 0 49 048 0 95 1.30 1.39 3.27 1 02 1.05 1.01 1.12 1.76 069 0.99 1.47 1.11 I 26 1.83 -- 1.24 1.12 -- -- 1.40 1.73 1.51 1.86 0.53 0.89 -- 0.69 0..10 0.45 1.04 0.78 0..79 0.79 0.44 1.07 0,98 0.55 -- 0.37 1.78 5.31 -- 2.13 1.49 1.44 7.55 1.57 1.08 0 99 0.93 1.65 2.02 1.21 -- M2 Mortality for all malignaui neoplasms is below that of the comparable genera] population segment (N = 254, SMR = 0.89, 95% Cl ~ 0.78, 1.01). Two cancer subsites are moderately elevated: malignant melanoma (H = 6, SMR = 2.56, 95% Cl = 0.94, 5 57) and multiple myeloma (N = 7, SMR = 1.81, 95% Cl = 0.73, 3.73). There was also a multiple myeloma death in a female marketing worker, bringing the multiple myeloma SMR for all marketing workers to 1.89 (N = 8, 95% C] = 0.81, 3.73). EXX-MOR-003418 218 Schnatter et al. TABLE VI. Mortality Results 1964-1983 in Marfcetlng/Tranaportatlon Workers of a Petroleum Company (N = 6,823) __________._ 95% Coaf. Ibl of death 2odes, 8th) Observed Expected SMR Lower limit Upper limit Ail causes (001-999) Infectious Dx (001-136) Endocrine D* (240-279) Blood Dx (280-289) Circulatory Dx (390-458) Respiratory Dx (460--519) Digestive Dx (520-577) Genitourinary Dx (580-629) Symptoms/ill-defmed (780-796) Accidcnts/violence (800-999) Malignant neoplasms (140-209) Esophagus Ca (150) Stomach Ca (151) I^rge intestine Ca (153) Rectum Ca (154) Liver Ca (155) Pancreas Ca (157) Bronchus/lung Ca (162 1) Bone Ca (170) Malignant melanoma (172) Prostate Ca (185) Kidney Ca (189.0-189 2) Bladder Ca (188) Brain (malignant) Ca (191) Nervous system (all) (191-192. 225 , 238-239) Reticulum-cell sarcoma (200.0) Lymphosarcoma (200 1) Non-Hodgkin's lymphoma (200,202) Multiple myeloma (203) f-eukemias (204-207) n neoplasms (210-228) - tes (250) _..ebrovascular Dx (430--438) Myocardial infarction (410) Other Heart Dx (420-429) Dx arteries/capillary (440 -448) Aortic aneurysm (441) Cirrhosis (571) Dx of pancreas (577) Kidney diseases (580--593) 1,157 6 13 3 651 67 35 13 10 72 254 9 18 31 6 2 15 64 2 6 28 9 5 7 8 l 2 7 7 14 1 9 90 230 28 46 25 14 2 9 1,317.2 8.3 24.0 3.1 670.0 93 7 54.8 iS 6 10.5 118.8 285.8 6.5 24.1 25.7 118 2.5 16.4 82.8 1.2 2.3 25.9 67 97 7.1 90 17 2.5 7,7 39 10.6 11 19.6 106.4 237.5 31.0 39.9 14 1 25 4 2.5 13.4 0.88 0.74 0.54 -- 097 0.72 0 64 0 70 0.96 0 61 0.89 1.38 0.75 1.20 0.51 -- 0.92 0.77 -- 2.56 1 08 ! .34 0.52 0.99 0.89 -- -- 0.91 1.81 I..33 -- 0 46 0.85 0.97 0.91 1.15 1.78 0.55 -- 0.67 0.83 0.27 Q..29 -- 0.90 0.55 044 0..37 0.46 0.47 0.78 0.63 0.44 0.82 0.19 -- 0.51 0.60 -- 0 94 0.72 0.61 0.17 0 40 0.38 -- -- 0.36 0.73 0 73 -- 0.21 0.68 0 85 0.60 0.84 1.15 0.30 -- 0.31 0 93 l 60 0.93 -- 1.05 0.91 0.89 1.20 1 76 0.76 1 01 2.63 1.18 1.71 1,10 -- 1.51 0.99 -- 5.57 3.56 2.54 1.21 2.03 1,75 -- LB7 3,73 2,23 -- 0 87 1.04 1.10 1.31 1.54 2 63 0 93 -- 1.28 We next examined the mortality patterns for these two diseases in exposed marketing workers by employment duration, latency, year of death, and hire date (see Tables VILA, and VHB), For multiple myeloma, higher SMRs were evident for work ers with longer latencies and employment. The risk concentrates in employees hired before 1950. No such trends were apparent for malignant melanoma, although work ers hired between 1940-1949 seemed to be at highest risk. The detection of trends is hindered by extremely small numbers. EXX-MOR-003419 Mortality in Petroleum Operating Segments TABLE VIIA. Multiple Myeloma Mortality Analyses in Marketing Workers of a Petroleum Company . nev s) j--4 5-14 15-24 25-34 35 + Deaths obs. 0 0 0 3 2 SMR 0 0 0 4.27 1.59 Employment (years) 0-4 5-14 15-24 25-34 35 + Deaths obs. 0 0 0 4 } Year of death Deaths obs. SMR Hire year Deaths obs. 1964 00 <1930 1965-1969 1 2.42 1930-1939 1970-1974 2 3,48 1940-1949 1975-1979 1 1.39 1950 + 1980-1983 l 1.31 2 1 2 0 219 SMR 0 0 0 3 66 2.09 SMR 2.54 2.04 1.57 0 TABLE VUB. Malignant Melanoma Mortality Analyses in Marketing Workers of a Petroleum Company Latency (yeajs) Deaths obs. SMR Employment (years) Deaths obs. 0-4 5-14 15-24 25-34 35 + 00 1 2.94 1 2 15 1 2.59 1 2.60 0-4 5-14 15-24 25-34 ' 35 + 0 I 2 1 0 Year of death Deaths obs SMR Hire year Deaths obs. 1964 00 <1930 1965-1969 2 7.54 1930-1939 1970-1974 1 2.86 1940--1949 `^75-1979 0 0 1950 + 0-1983 1 1.92 0 0 3 1 SMR 0 2.96 3.99 2.22 0 SMR 0 0 6 99 1 16 Mortality for other lymphopoietic cancers is generally consistent with expecta tion (see Table VT). Leukemia mortality is moderately elevated (N = 14, SMR -- 1.33, 95% Cl - 0.73, 2.23), but the SMR decreases in exposed marketing workers (N = 7, SMR = 0.98, 95% Cl = 0.39, 2.02). Mortality for the myeloid and lymphatic cell types are about as expected for ail marketing workers (N = 7, SMR = 0.96). This implies an SMR of 2.14 for seven leukemias which were not myeloid or lymphatic (not statistically significant). The vast majority of these would be ex pected to be leukemias unspecified as to cell type. Kidney cancer morality is of particular interest in this operating segment due to toxicologic findings on totally vaporized gasoline. Kidney cancer is moderately ele vated in exposed marketing workers (SMR = 1.53), but the excess is not statistically significant (95% Cl = 0,61, 3.16). Marine locations. Marine workers have potential exposure to finished prod uct, most often during loading and unloading operations, and are not generally ex- 220 Schnatter et al. posed to crude oil. This exposure partem most closely resembles that of the marketing and transportation workers. There are relatively few employees (1,660) and deaths (176) in this group, although all-cause and external cause SMRs are significantly low. ' - cancer subsite is significantly elevated. There were no deaths due to malignant anoma and only one from multiple myeloma; these two causes were elevated in .arketing/transportation workers. There were five central nervous system cancers (benign and malignant), resulting in an SMR of 2.41 which is not significant (95% Cl = 0.78, 5.63). This cause was not elevated in marketing workers (SMR = 0.87, Table VI). Upstream (production/drilling/exploration and pipeline) locations. There were only 176 pipeline workers in this study. The vast majority of pipeline operations applicable to the study time period carried crude oil. Thus, pipeline workers were grouped with production, drilling, and exploration workers to form an upstream operating segment. There were a total of 4,443 male upstream workers, but a lower percentage of workers was judged as potentially exposed to HC (2,465 of 4,443, or 55%) compared to the other operating segments. All cause mortality is significantly low in this group (SMR = 0.74, 95% Cl = 0.66, 0.84), as are most non-maiignant categories (see Table VHI). One cancer subsite was significantly elevated: malignant melanoma (N = 6, SMR = 6.00). Though this SMR is quite high, it is very imprecise, based on the exact 95% Cl of (2.19, 13.06). None of the lymphopoietic cancer sites, or kidney cancer, were elevated. Since most upstream workers (87.3%) resided in Alberta, we used Alberta reference rates in lieu of Canada rates. Table DC shows a slight SMR increase to 6.59 (95% Cl = 2.42, 14.41) when using these rates. Thus, factors unique to Alberta do not explain the excess melanoma mortality. Of the 16 total melanomas in the entire cohort, 6 occurred on the trunk, while 9 were unspecified. Unfortunately, all six upstream cases were unspecified. Table DC shows that malignant melanoma mortality remained statistically elevated for exposed upstream workers (SMR = 5.48, 95% Cl = 1.10, 16.03). Despite extremely small numbers, there was a strong upward trend in SMRs for 'ployment duration and latency (see Table X). The SMR for employees with at least rears service and a 20-year latent period is 13 .33, based on three deaths. Although ^rnplete computerized work histories are not available, two of the three cases worked in drilling and gas production, while the third decedent was a gas plant operator. Petrochemical locations. The single petrochemical location contributed too few employees (1,396) and deaths (52) for detailed analyses. However, SMRs for all causes (0.52), circulatory disease (0.55), and accidents/violence (0.28) are signifi cantly reduced. There was only one cancer subsite with more than two deaths (lung cancer SMR -- 0.52). No angiosarcoma deaths were observed despite the presence of vinyl chloride monomer at this location. Although some authors have noted an increased incidence of brain tumors in petrochemical workers [Reeve et al., 1983; Austin and Schnatter, 1983], no such deaths were observed in these workers. DISCUSSION This study of petroleum workers simultaneously examined mortality according to different operating segments and exposure to hydrocarbons A strength of this Mortality in Petroleum Operating Segments 221 TABLE Vm, Mortality- Results 1964-1983 in Upstream Workers of a Petroleum Company (N = 4,443) 95% Conf. Ini. ause of death rlCD Codes. 8th) Lower Upper . Observed . Expected SMR limit limit All causes f001-999) Infectious Dx (001-136) Endocrine Dx (240- 279) Blood Dx (280-289) Circulatory Dx (390-458) Respiratory Dx (460-519) Digestive Dx (520-577) Genitourinary Dx (580--629) SymptomsdJl-defined (7B0-796) Accidents/violence (800 -999) Malignant neoplasms (140-209) Esophagus Ca (150) Stomach Ca (151} Large intestine Ca (153) Rectum Ca (154) Liver Ca (155) Pancreas Ca (157) Bronchus/lung Ca (162 1) Bone Ca (170) Malignant melanoma (172) Prostate Ca (185) Kidney Ca (189.0-189.2) Bladder Ca (188) Brain (malignant) Ca (191) Nervous system (all) (191-192, 225. 238-239) Reticulum-cell sarcoma (200.0) Lymphosarcoma (200.1) Non-Hodgkin's lymphoma (200,202) Multiple myeloma (203) Leukemias (204-207) rnign neoplasms (210-228) Diabetes (250) Cerebrovascular Dx (430-438) Myocardial infarction (410) Other Heart Dx (420--429) Dx arteries/capillary (440-448) Aortic aneurysm (441) Cirrhosis (571) Dx of pancreas (577) Kidney diseases (580-593) 275 1 3 0 118 8 18 3 4 36 80 3 7 6 2 O 3 20 0 6 7 2 4 4 6 0 0 2 0 2 0 2 19 48 3 5 2 7 3 2 369.6 2.4 6.3 08 162.8 20.0 18.2 3-6 3.1 58.0 84 0 1.9 6.2 6.9 3.2 0.8 4.8 26.9 04 1.0 4.6 2.2 2.1 30 3.7 0.6 0.8 2.7 1.1 3.3 0.4 50 20.4 72.8 8.1 73 3.4 10 6 09 3.0 0.74 -- -- -- 0.73 0 40 0.99 -- -- 0 62 0,95 -- 1.13 0.87 -- -- -- 0 74 -- 6 00 1.54 -- -- 1.60 -- -- -- -- -- -- -- 0.93 0.66 0.37 0.69 -- 0 66 -- -- 0.66 -- -- 0.60 0.17 0.59 -- -- 0.43 0.76 -- 0.45 0.32 -- -- -- 0.45 -- 2.19 0.62 -- -- -- 0.59 -- -- -- -- -- -- -- 0.56 0 49 0 07 0.22 -- 0.26 -- -- 0.84 -- -- -- 0.89 0 79 1.56 -- -- 0.86 1.19 -- 1.25 1.88 -- -- -- 1.15 -- 13.06 3.16 -- -- 3.49 -- -- ~ -- -- -- -- 1.45 0.88 1.08 1.60 -- 1.36 -- -- study is its breadth; ii examines different operating segments and different work environments with different exposure patterns. Stratification by operating segments allows for the assessment of mortality patterns for several fairly homogeneous work environments Another strength of the study is its size; to our knowledge it has the largest number of employees of any single-company mortality follow-up study re ported in the petroleum industry [see Wong and Raabe, 1989). Thus, the precision of EXX-MOR-003422 222 Schnatter et al. TABLE IX. Mafignaot Mdanoeoa SMRs in Upstream Workers of a Petrotenm Company 1964-19S3 Referral population Observed Expected SMR 95% ci Canadian population Alberta population 6 6 1.00 0.91 6.00 2.19-L1.06 6.59 2.42-14.41 Exposure sums Observed Expected SMR 95% a Exposed Not exposed Unknown 3 0.55 5 48 1.13-16.03 2 0.30 6.62 0.80-23.92 1 0.15 6.64 0.17-36.90 TABLE X. Malignant Melanoma SMRs by Latency and Employment Duration for Upstream Workers Id a Petroleum Company 1964-1983 Latency (years) Observed deaths Expected deaths SMR 0-4 5-14 14-24 25-34 35 + 0 0.06 0 0 0.14 0 1 0.17 5.92 1 0.14 7.25 l 0.04 23.80 Employment duration (years) Observed deaths Expected deaths SMR 0-4 5-14 15-24 25-34 35 + 0 0.11 0 0 0 13 0 1 0 18 5.56 1 0.12 8.33 1 0.02 62.50 the SMRs is enhanced without sacrificing data comparability that is inherent in inter-industry studies. Among the study shortcomings are: a) the relatively young cohort examined (only 11% are deceased and 54% have less than 20 years of follow-up); b) the relatively short follow-up period; c) the incomplete job history information for some workers; and d) the absence of data on life-style risk factors such as smoking, alcohol use, and socio-economic status. The incompleteness of job histories necessitated extrapolation of last job or first known job to earlier time periods. This could result in misclassification of exposure and latency/duration in long-term exposed employ ees. The relatively short follow-up time and young cohort translates into less precise SMRs for longer employment duration and latency times, given the cohort's size,, For example, 30% of the 109,336 person-years in refinery workers are for workers with less than 5 years worked and 10 years latency. Further follow-up of the cohort would thus be desirable. Overall Cohort Overall mortality and most non-malignant causes of death were significantly lower than general population rates in the overall cohort, as well as the major oper Mortality in Petroleum Operating Segments 223 ating segments. This is most likely due to the attendant survival advantages for selection into and maintenance of employment This healthy worker effect is repeat edly observed in petroleum Industry and other occupational cohort studies, lncom- ete death ascertainment is not a likeJy explanation, since a blinded test for known jeaths-revealed nearly 98% death ascertainment based on the identifiers supplied for linkage to Statistics Canada's Mortality Data Base (MDB) [Schnatter et al., 1990]. Malignant Melanoma The most noteworthy finding is the significant excess of malignant melanoma overall, particularly for the upstream operating segment (SMR = 6.0). The excess persists when Alberta rates are used (SMR = 6 6) and for HC-exposed upstream workers (SMR -- 5.5). Although we analyzed results by operating segment to detect potential risks applicable to a fairly homogeneous work environment* the upstream operating seg ment remains relatively diverse. For example, drilling workers are primarily exposed to crude oil and drilling muds (made up of complex HC mixtures) via dermal ab sorption, but production workers can be exposed to a wider range of HCs, including natural gas and other liquid streams.. Exploration workers are less likely to be exposed to hydrocarbons. Many upstream workers spend a significant part of their occupationi time out-of-doors. Thus, at this time, wc are unable to relate the increased melanoma SMR in upstream workers to a specific agent or HC stream. Various HC fractions have produced squamous cell carcinomas in mice, includ ing light and heavy vacuum and catalytically cracked distillates [IARC, 2989). How ever, these findings may not be etiologically relevant to melanomas, which arise from pigment-producing melanocytes. Previous experimental and population-based studies give little indication that substances which may have been encountered by the upstream workers can cause melanoma. Other exposures, such as polychlorinated biphenyls [Bahn et al., 1976], gold mining [Armstrong et al., 1979], rubber industry processes [Holmberg et al., 1983], vinyl chloride [Heldas et al., 1984], aircraft worker activities [Costa et 'L, 1989], electrical jobs and electromagnetic fields [Olin et al., 1985; Vagero et al,., 185; Swerdlow, 1983], and the work activities of chemists [Austin et al., 1981; Hoar _jid Pell, 1981) have been associated with melanoma, but none of the findings are conclusive. Most refinery worker cohort studies do not show an excess of skin cancer and/or melanoma [Deizell et al., 1988], except for two [Rushton and Alderson, 1981; Nelson et al., 1987]. In experimental studies, 7,I2-dimethyIbenz(a)anthracene (DMBA) [Takizawa et al., 1985; Berkelhammer and Oxenhandler, 1987] and ure thane [Vesselinovitch et al,, 1970; Toth et al., 1961] have been shown to cause melanomas in mice and hamsters, respectively. In summary, however, the literature lends little support that hydrocarbon streams encountered in the upstream operating segment can cause melanoma. Melanoma mortality is increasing rapidly [Osterlind and Jenson, 2 986; Glass and Hoover, 1989] and is thought to be due to a real rise in incidence, rather than improved diagnosis for the disease [Elwood and Lee, 1974]. Concurrent 5-year intervals were used for observed and expected deaths in this study, so a temporal effect should be ruled out as. an explanatory factor for the excess. Several investigators have suggested that the background melanoma increase is due to ultraviolet (UV) light exposure [Fitzpatrick and Sober, 1985; Lee, 1982). and. 224 Sdmatter tt ad. specifically, UV overexposure, or the number of sunbuming episodes [Rhodes et al., 1987; Mackie and Atchison, 1982], especially during childhood [Gallagher et al., 1985). Owing to the amount of occupational time spent outdoors by these workers, LTV light exposure must also be considered a candidate etioiogic factor for these melanomas. A siinligbt/exposure interaction is also a plausible explanation for this excess. Indeed, experimental evidence for this phenomenon exists [Epstein et al., 1967]. Hairless, pigmented mice exposed to DMBA alone developed blue nevi precursor lesions. Mice treated further with mid-UV energy light developed melanomas, while no melanomas developed in mice treated with DMBA alone, UV light alone, or neither factor. The applicability of these findings to our study is uncertain without more specific information on exposure to HC streams and UV light. Other known risk factors for melanoma include congenital moles, dysplastic nevi syndrome, a history of cutaneous melanoma in close relatives, and the skin's sun-sensitivity [Rhodes et al., 1987]. We did not collect information on any of these risk factors, thus they remain as possible explanations of the excess. A factor more common in Alberta, where most of the upstream workers reside, is not a likely explanation, since the use of Alberta reference rates did not alter the excess.. The only other study of upstream workers did not report a melanoma or skin cancer excess [Divine and Barron, 1987], These workers were primarily from lower latitudes, where cumulative sunlight exposure would be expected to be higher. A greater opportunity for intermittent sunlight exposure in Alberta, local work environ ment differences, or the presence of other potential confounders mentioned above may explain the differences in the two studies. Although the malignant melanoma excess concentrated in employees ever as signed to the upstream segment, the melanoma SMR was also significant for all males in the total cohort and was elevated, though not significantly so (N = 6, SMR = 2.56) in marketing/transportation workers, and in exposed marketing/transportation workers (N ~ 4, SMR = 2.37). Trends with latency and duration of exposure were not evident, as in the upstream segment. While these data indicate that this finding could be due to chance, further follow-up efforts on this cohort would be useful to monitor melanoma mortality and incidence. Aortic Aneurysm Aortic aneurysm mortality was also significantly elevated in the male cohort as a whole (SMR = 1.41). There are few population-based studies of aortic aneurysms, and fewer studies which examine the relationship of the disease to occupation. Mor tality due to aortic disease is subject to variation since other cardiovascular diseases such as atheroscierodc heart disease (ASHD), hypertensive disease, and cerebrovas cular disease often coexist, and sometimes take priority in death certificate coding [Daien, 1987], Melton et al. [1984] stress that improved diagnostic capabilities (roentgeno grams and ultrasonography) have seemed to fuel a rise in the incidence of this disease, particularly the smaller, asymptomatic abdominal aneurysms. The majority (35 of 40 specified, 87.5%) of the aortic aneurysms in this study were abdominal aneurysms, which are more strongly influenced by this trend. If employees in this study are more likely than the general population to undergo the diagnostic procedures mentioned above, an improved detection rate for aortic aneurysms could result. This has been Mortality in Petroleum Operating Segments 225 labeled a "diagnostic sensitivity bias" for workers with better access to sophisticated diagnostic procedures [Greenwald et al., 1981]. Other factors that cannot be ruled out xplanatory factors are trauma, and dietary and lifestyle factors related to byperiOn and arteriosclerosis. Operating Segments Refinery workers. Several studies of refinery workers have been conducted and have been reviewed by Delzel) et al. [1988], Wong and Raabe [1989], and IARC [1989]. The latter group classified refinery operations as probably carcinogenic largely based on findings regarding leukemia, and also on skin cancer. The findings in this study do not support this conclusion. Both of these death categories showed deficits in these refinery workers. The only significant finding ir refinery workers was for benign neoplasms. However, this excess was weakened in ever-exposed workers and was not based on tumors of a common anatomical site. Furthermore, the corresponding malignant cancer sites were not elevated. While the combined category of all lymphopoietic cancers show an SMR close to unity, mortality patterns for different lymphopoietic cancers show considerable variation. Lymphosarcoma (LSA) and multiple myeloma show very low SMRs, while reticulosarcoma (RCS) shows an elevation (SMR = 2.44). The combined RCS and LSA SMR is 0.98 (N = 6), consistent with previous refinery worker studies [see Delzell et al., 1988] which commonly group these two diseases. Possible diagnostic overlap [Percy et al., 1981; Heath, 1982] could explain disparate SMRs for lym phopoietic subtypes [Delzel] et al., 1988], but non-optimal groupings of these sub types cloud even this explanation. Future petroleum worker studies could evaluate RCS and LSA separately as a test of the elevated RCS rate found in this study. Alternately, a grouped category of "non-Hodgkin's lymphoma" is preferred because it accounts for all lymphomas which are not distinctly Hodgkin's disease. In other studies, unspecified lymphomas are often grouped with multiple myeloma, poly cythemia vera, and myelofibrosis into an "other lymphatic cancer" category. This rubric consists of diseases with different etiologies; if the study size is sufficient, it uld seem advisable to at least evaluate unspecified lymphomas with other nonJgkin's lymphomas, and multiple myeloma separately. Marketing/transportation workers. Marketing/transportation workers show significantly elevated mortality for aortic aneurysms, which is stronger in non-exposed workers. In the only other study among similar employees, Rushton and Alderson [1983] report results for "diseases of the arteries" which includes aortic aneurysms as well as atherosclerosis. This cause was not elevated based on 81 observed deaths (SMR = 0.91). Multiple myeloma mortality was elevated, though not significantly, in exposed as well as total marketing workers in the present study. This excess concentrated among employees with long employment duration and latency, and most of the decedents died within 5 years of last employment. Rushton and Alderson [1983] reported a multiple myeloma SMR of 1.17, based on 11 cases. This does not support our finding, although additional analyses by duration of employment were not per formed for this cause. Rushton and Alderson [ 1983] do report a significant excess of another lymphopoietic disorder, myelofibrosis, although we did not observe any such deaths. Rushton and Alderson [1983] also found a non-significantJy elevated SMR for Hodgkin's disease (1.39), but we found a deficit (N = 2, SMR = 0.84). 226 Schnatter et al. Some finished products such as gasoline contain benzene, a known cause of leukemia. Leukemia mortality was moderately raised in all marketing workers (N 14, SMR = 1.33) but was lower for exposed marketing workers (N = 7, SMR = ' 98).. We have no knowledge of how many workers were actually exposed to ben- ne, nor the level at which they may have been exposed. However, Rushton and Alderson [1983] also report results (SMR = 1 04) similar to our exposed workers. Other investigators have suggested that multiple myeloma may be related to benzene exposure [Rinsky et ai., 1987; Decoufle et al., 1983]. The leukemia results described above would argue against a benzene relationship, although Rinsky et al. [1987] suggest that benzene exposure at lower levels may cause multiple myeloma. Without more specific job histories relating to possible benzene exposure, we cannot evaluate this further. Different lymphopoietic cancers share possible etiologies (e..g., aromatic sol vents and radiation), and even diagnostic characteristics [e.g,, lymphocytic leukemias/iymphomas, Heath, 1981]. Theoretically, an insult to a multipotent stem cell could manifest as different lymphopoietic cancers, which justifies the examination of an aggregate category of lymphopoietic cancers. However, several investigators ar gue against this [Crane et al., 1989; Linet, 1985], pointing out different etiologies even within specific leukemia subtypes (e.g., genetic factors, radiation, chromosomal aberrations). Although the authors of this paper share different views on the issue, we have examined a category of combined lymphopoietic cancer for reader convenience. For non-Hodgkin*s lymphoma, multiple myeloma, leukemias, and myelofibrosis, a slight, non-significant (N = 21, SMR = 1.22) excess was observed in exposed marketing workers. A similar finding can be calculated for the U.K. workers (N = 57, SMR - 1.15) (Rushton and Alderson, 1983), suggesting no, or at most a slight, relationship in the total marketing work environment. Renal carcinogenicity has been reported in male rats exposed to totally vapor ized gasoline [Kitchen, 1984], although subsequent studies ascribe this to accumu lation of a protein not found in humans. Exposure potentials of marketing/distribution workers are characterized by intermittent gasoline vapor inhalation, mostly during loading/unloading operations. For all HC-exposed workers in this segment, the kid'ey cancer SMR shows a 34% elevation, which is not statistically significant. Rushjn and Alderson [1983] also report an elevation (SMR = 1.21), based on 23 deaths in marketing workers which is not statistically significant. While neither study alone is persuasive of an effect, a small effect cannot be ruled out. Ongoing studies in different downstream populations should shed further light on whether exposures encountered in this setting might be related to kidney cancer mortality. Upstream workers. With the exception of malignant melanoma, upstream workers did not show other significantly elevated SMRs. Specifically, there were no thyroid cancer deaths or deaths from benign or unspecified neoplasms as in Divine and Barron [1987], However, the upstream group in this study is relatively small, and the study has insufficient statistical power for excluding risks as large as five-fold for these rare causes of death. ACKNOWLEDGMENTS The authors would like to thank Martha Fair and Elizabeth Coppock of Statistics Canada for mortality foilow-up efforts, Eugene Sales for nosological coding, Hans Mortality in Petroleum Operating Segments 227 Siegel and other company industrial hygienists for IH support, Celia Milano and Gail Jorgenson for programming, Dr. John Acquavella for coordinating efforts and helpful Mnments, and Dr. Nancy Hanis for beginning the early phases of the study. REFERENCES ` Armstrong BK, McNulty JC, Levitt U, Williams KA. Hobbs MST (1979): Mortality in gold and coal miners in Western Australia with special reference to lung cancer. Br J lnd Med 21:619-623. Austin DF. Reynolds PJ, Snyder MA, Biggs MW, Stubbs HA (1981): Malignant melanoma among employees of the Lawrence Livermore National Laboratory. Lancet 2:712-716. Austin SG, Schnatier AR (1983): A cohort mortality study among petrochemical workers. J Occup Med 25:304--312. Bahn AX, Rosenwaike 1, Hermann N, Stellman J. O'Leary K (1976): Melanoma after exposure to PCB's. N Eng) J Med 295:450. ' Berkelhammer J. Oxenhandler RW (1987): Evaluation of premajignam and malignant lesions during the induction of mouse melanomas. Cancer Res 47:1251-1254. Chcckoway H, Pearce N, Dement JM (1989): Design and conduct of occupational epidemiology studies. II. Analysis of cohort data. Am J lnd Med 15:375-394. Costa G, Merietti F, Segnan N (1989): A mortality cohort study in a north Italian aircraft factory. Br J lnd Med 46:738-743. Crane MM, Keating Ml, Trujillo JM, Labarthe Dr, Frankowski RF (1989): Environmental exposures in cytogenetically defined subsets of acute nonlymphocytic leukemia. JAMA 262:634-639. Dalen J (1987): Diseases of the aorta. In Brauowald E, Issclbacber KJ, Petersdorf RG, Wilson JD, Martin JB, Fauci AS (eds): "Harrison's Principles of Internal Medicine '' New York: McGraw-Hill Co. Decouflf P, Blatmer WA, Blair A (1983): Mortality among chemical workers exposed to benzene and other agents. Environ Res 30:16--25. Delzell E, Austin H. Cole P (1988): Epidemiologic studies of the petroleum industry. Occupational Med: State of the Art Reviews 3:455--474. Divine BJ. Barron V (1987): Texaco mortality study IE. A cohort study of producing and pipeline workers. Am J lnd Med 10:371-381. El wood JM. Lee JAH (1974): Trends in mortality from primary tumors of the skin in Canada. Can Med Assoc J 110:913-915. Epstein JH, Epstein WL, Nakai T (1967): Production of melanomas from DMBA-mduced blue nevi in hairless mice with ultraviolet light, J Natl Cancer Inst 38:19-30. Citzpatrick TB, Sober AJ (1985): Sunlight and skin cancer, N Engl J Med 313:818--820. UagherRP, Elwood JM, Hill GB (1985): Risk factors for cutaneous malignant melanoma: The Western Canada melanoma study.. Recent Results Cancer Res 102:38-55. Glass AG, Hoover RN (1989): The emerging epidemic of melanoma and squamous ceil skin cancer. JAMA 262:2097-2100, Greenwald P, Friedlander BR, Lawrence CE, HearneT, Earle K (1981): Diagnostic sensitivity bias: An epidemiologic explanation from an apparent brain tumor excess J Occup Med 23:690-694. Hanis NM, Stavraky KM, Fowler JL (1979): Cancer mortality in oil refinery workers J Occup Med 21:167-174. Health Effects Institute (19B8): "Gasoline Vapor and Human Cancer. Evaluation of Existing Information and Recommendations for Future Research." Cambridge, MA: HE1. Heath CW (1982): The Leukemias. In Schottenfeld D, Fraumeni JF (eds): "Cancer Epidemiology and Prevention." Philadelphia: W. B. Saundere. Heldas SS, Langard SL, Andersen A (1984): Incidence of cancer among viny) chloride and polyvinyl chloride workers Br J lnd Med 41:25-30, Hoar SK, Pell S (1981): A retrospective cohort study of mortality and cancer incidence among chemists. J Occup Med 23:485 -494. Holmberg B, Westerholm P, Massing R, Kestrup L, Gumaelius K, Holralund L, Englund A (1983): Retrospective cohort study of two plants in the Swedish rubber industry Scand J Work Environ Health 9 [Suppl 2):59-68. IARC (1989); "1ARC Monographs on ib: Evaluation of Carcinogenic Risk of Chemicals to Man: 228 Scimatter et al. Occupational Exposures in Petroleum Refining; Crude Oil and Major Petroleum Fuels." Lyon, France: 1ARC, Vol 45. ^lichen DN (1984): Neoplastic renal effects of unleaded gasoline in Fischer 344 rats In Mehlman MA, Hemstrcet GP. Thorpe IJ, Weaver NK (eds): "Renal Effects of Petroleum Hydrocarbons: Ad vances in Modem Environmental Toxicology VII." Princeton: Princeton Scientific Publishers, pp 65-72. . Lee JA (1982): Melanoma and exposure to sunlight. Epidemiol Rev 4:110--136. Linet 1985: "The Leukemias: Epidemiologic Aspects." New York: Oxford University Press. Mackie RM. Aichison T (1982): Severe sunburn and subsequent risk of primary cutaneous malignant melanoma in Scotland. Br J Cancer 46:955-960. McDonald AD. McDonald JC (1980): Malignant mesothelioma in North America. Cancer 46:1650 1656. Melton LI. Bickentaff IJ, Hoilier LH, Van Peenen HJ, Lie JT. Paiiolero PC, Cherry KJ, O'Fallon WM (1984): Changing incidence of abdominal aortic aneurysms: A population-based study. Am J Epidemiol 120:379-386. " Monson RR (1974): AnaJysis of relative survival and proportional mortality. Comp Biomed Res 7: 325-332, Nelson NA, Van Peeneo PFD, Blanchard AG (1987): Mortality in a recent oil refinery cohort. J Occup Med 29:610-612. - Olin R, Vagcro D, Ahlbom A (1985): Mortality experience of electrical engineers. Br J Ind Med 42:211-212. Osteriind A. Jensen OM (1986): Trends in incidence of malignant melanoma of the skin in Denmark 1943--1982. Recent Results Cancer Res 102:8-17. Percy C, Stanek E, GlocckJer L (1981): Accuracy of cancer death certificates and its effect on cancer mortality statistics. Am J Public Health 71:242-250.. Reeve GR. Bond GO, Lloyd JW. Cook RR, Waxweiler RJ, Fishbeck WA (1983): An investigation of brain turnon among chemical plant employees using a sample-based cohort method J Occup Med 25:387-393 Rhodes AR, Wetnstock MA. Fitzpatrick TB, Mihm MC, Sober AJ (1987): Risk factors for cutaneous melanoma: a practical method for recognizing predisposed individuals, JAMA 258:3146--3154. Rinsky RA, Smith AB, Hornung R, Filloon TG, Young RJ, Okun AH, Landrigan PJ (1987): Benzene and leukemia: An epidemiologic risk assessment. N Engl J Med 316:1044-1050. Rothman KJ, Boice KJ (1979): Epidemiologic analysis with a programmable calculator. U.S. Dept, of Health, Education, and Welfare, NIH Publication No. 79-1649. Runion HE (1988): Occupational exposures to potentially hazardous agents in the petroleum industry. Occupational Med: State of the Art Reviews, 3:431-444. ''ushton L, AJderson MR (1981): An epidemiologic survey of eight oD refineries in Britain. Br J Ind Med 38:225-234. .ushton L, Alderson MR (1983): Epidemiologic survey of oil distribution centres in Britain. Br J Ind Med 40:330-339 Schnatter AR, Acquavella JF, Thompson FS, Donaleski D, Theriault G (1990): An analysis of death ascertainment and follow-up through Statistics Canada's Mortality Data Base System, Can J Pub! Health 81:60-65. Smith ME, Newcombe HB (1980): Automated follow-up facilities in Canada for monitoring delayed health effects. Am J Public Health 70:261-268. SweTdlow AJ (1983): Epidemiology of eye cancer in adults in England and Wales 1962-77. Am J Epidemiol 118:294-300 Takizawa H, Sato S, Kitajima H, Konishi S. Iwata K, Hayashi Y (1985): Mouse skin melanoma induced in two stage chemical carcinogenesis with 7,12-dimethylbenz(a)anthracene and croton oil Car cinogenesis 6:921-923. Toth B. Tomatis L, Shubik P (1961): Mulupotcotial carcinogenesis with urcihan in the Syrian golden hamster. Cancer Res 21:1536-1541. Vagero D. Ahlbom A, Olin R. Sahlsten S (19B5): Cancer morbidity among workers in the telecommu nications industry.. Br J Ind Med 42:191-195. Ves5elinovitch SD, Mihailovich N, Richter WR (1970): The induction of malignant melanomas in Syrian white hamsters by neonatal exposure to lirethan Cancer Res 30:2543-2547 Wong 0, Raabc G (J989): Critical review of cancer epidemiology in petroleum industry employees, with a quantitative meta-analysis by cancer site.. Am J Ind Med 15:283-310. Mortality in Petroleum Operating Segments 229 APPENDIX; 1CDA Codes Examined for Mesothelioma and Angiosarcoma Deaths Mesothelioma -imary codes ,/th Revision (1958--1968} . Secondary codes 227--Mesothelioma, unspecified 212--Mesothelioma lung, pleura, mediastinum 211--Mesothelioma peritoneum 197--Malignant mesothelioma, unspecified or diaphragm 163--Malignant mesothelioma, lung or pleura 164--Malignant mesothelioma, mediastinum 158--Malignant mesothelioma, peritoneum Same 8th Revision (1969-197B) 228--Mesothelioma, unspecified 212--Mesothelioma, lung 158--Mesothelioma, peritoneum 211--Mesothelioma, peritoneum 163--Mesothelioma, pleura 212--Mesothelioma, pleura, fibrous or mediastinum 215--Mesothelioma, diaphragm 199--Malignant mesothelioma 162--Malignant mesothelioma, lung 171--Malignant mesothelioma, mediastinum 163---Malignant mesothelioma, mediastinum 9th Revision (1979+) 199--Mesothelioma, unspecified 162--Mesothelioma, lung 229--Benign mesothelioma 212--Benign mesothelioma, lung, pleura or mediastinum 215--Benign mesothelioma, diaphragm 211--Benign mesothelioma, peritoneum 171--Mesothelioma, diaphragm 164--Mesothelioma, mediastinum 58 Mesothelioma, peritoneum 53--Mesothelioma, pleura Angiosarcoma Primary codes ?th Revision (1958-1968) 197--Angiosarcoma 156--Angiosarcoma, liver 155--Primary angiosarcoma.. liver 156--Primary angiosarcoma, bile ducts 8th Revision (1969-1978) 171--Angiosarcoma 197--Angiosarcoma, liver 155 --Primary angiosarcoma, liver or intra-hepatic bile ducts 156--Primary angiosarcoma, extrahepsric bile ducts 9th Revision (1979+) 171--Angiosarcoma 155--Angiosarcoma, liver 155--Primary angiosarcoma, liver 156--Primary angiosarcoma, bile ducts Same, except for 199 Same, except for 199 Secondary codes . Same Same Same EXX-MOR-003430