Document O9q96brQvwvoMqbgDJbpNd2X

American Journal of Industrial Medicine IS: 283-310 (198?) Critical Review of Cancer Epidemiology in Petroleum Industry Employees, With a Quantitative Meta-Analysis by Cancer Site Otto Wong, ScD, face, and Gerhard K. Raabe, DrPH, ms A critical review of close to 100 published and unpublished but otherwise available epidemiologic reports of petroleum industry employees from the United States, United Kingdom, Canada, Europe, Australia, and Japan was conducted. Analyses by duration of employment and latency are discussed, and summary standardized mortality ratios (SMRs) or meta-SMRs are developed for selected cancer sites. Findings indicate that the industry experienced a significantly lower cancer mortality than the general population for all cancer sites combined, digestive system, stomach, and lung. For the industry as a whole, SMRs similar to the general population were observed for skin, brain, pancreatic, prostaric, and kidney cancers. However, some data indicate that certain small groups within the industry might have elevated prostatic and kidney cancer risk. This review supports the conclusion that some refinery employees, particularly those employed before the 1940s, may have been at increased risk of leukemia. There is some indication that cancer of other lymphatic tissue may also be elevated. Unresolved issues affecting these conclusions are discussed, and specific directions for future research are offered. Key words: occupational cancer, petroleum, gasoline, benzene, meta-analysis, meta-SMR INTRODUCTION A large number of epidemiologic studies of workers in the petroleum industry' have been conducted. Although these studies differ ip design, size, and length of follow-up, they examine the mortality pattern of workers with similar exposures, including refinery process streams chemically similar in composition to commercial gasoline. These workers were exposed to a higher concentration and for a longer duration than the general public. Therefore, these studies represent the best available data on human effects of exposure to gasoline hydrocarbons. Although several epidemiologic reviews have appeared [Raabe, 1983; Savxtz and Moure, 1984; EPA, 1987; Harrington, 1987], none of these reviews examined the complete literature systematically or provided a quantitative meta-analysis. The ENSR Health Sciences, Alameda. California (O.W.). Medical Department, Mobil Corporation, New York (G.K.R.). Address reprint requests to Otto Wong, ENSR Health Sciences, 1320 Harbor Bay Parkway, Alameda, CA 94501. Accented for publication September 30. 1988 EXX-MOR-003360 284 Wong and Raabs purpose of this review is to examine critically this literature and to develop a quantitive assessment (meta-analysis) of the site-specific cancer mortality patterns that emerge from these studies. Close to 100 published articles and unpublished but otherwise available reports from the United States, United Kingdom, Canada, Europe, Australia, and Japan have been reviewed. Particular attention has been directed toward the appropriateness and adequacy of study design, study subject selection, exposure data, and follow-up of study subjects. Since we are dealing with the assessment of carcinogenicity of industrial chemicals, the most appropriate study design is that of a retrospective or historical cohort mortality study. In a cohort study, the population at risk is defined, and the exposures (at least in general terms) are relatively well documented. Further more, mortality is a valid endpoint for most cancer sites. In addition to study design and data quality, statistical methods have been examined, as well as the validity of interpretations and conclusions. Our assessment is based on relevant and appropriate studies. A number of articles were not included in the final assessment because they were found to be inappropriate or inadequate. We will first describe briefly the studies used in this review. Table I provides some of the basic information for the studies reviewed: company, location of facility, length and period of follow-up, minimum length of employment for cohort eligibility, cohort size, person-years, and the number of deaths. A full report with a summary and comments on each study is available upon request. ; In the United States, the first large-scale industrywide cohort mortality study in the petroleum industry was sponsored by the American Petroleum Institute (API) and conducted by Tabershaw/Cooper Associates [TCA, 1974], Based on this study, W'ong [1980] reported an analysis of brain tumor mortality, which was not included in the original TCA report. Subsequently this API refinery study was updated from 1971 to 1980 [Kaplan, 1986], Also under the sponsorship of API was the Memorial Sloan Kettering Cancer Center (MSKCC) registry of mortality and cancer incidence of 19 oil companies. The results of the first three years of data were reported by Schotten- feld et al. [1981]. A large number of studies on refinery workers have been conducted or spon sored by individual oil companies in the United States. Although the majority of the studies were historical cohort mortality studies in design, a small number of studies was actually based on a limited-length of follow-up.`From the methodological point of view, these studies fall into a hybrid category between the traditional cohort (longitudinal) study and a -cross-sectional survey. The individual company studies that fail into the hybrid category include the AMOCO companywide study [Nelson, 1985]; the Exxon study of the Baton Rouge .refinery [Hanis et ai., 1982]; the Exxon study of the Baton Rouge, Baytown, and Bayway refineries and chemical plants [Hanis et al., I985a,b]; and the two Shell studies of the Wood River [McCraw et al., 1985] and Deer Park [Joyner, 1983] refineries. As indicated in Table I, the follow-up was relatively short for these studies with the hybrid design. In most cases, the start date of the follow-up period coincided with the date of the computerization of personnel records. This type of study, although limited from a methodological point of view, was relatively quick and less expensive. The traditional cohort mortality study, with a much longer follow-up, requires the time-consumina task of manually codin a both the demoaraohic and work history EXX-MOR-003361 TABLE !. Description of Cohort SUittie* in the Petroleum industry Autltnr/year Tahcrxhuw Cooper Asxc 11974) Wong |i9Rj Kaplan j I9R6) SchonenfcW cl al. {WKIE Nelson | I9R5 j Wong et a!. {I9B6| llnis cl ol J19R2) Hnnis et al. |I9R5a| Wen et al |19831 Wen cl al. |I9Mc| Morgan and Wong[19K4| Morgan and Wong[l9H5| Enlerline nod Henderson | !9K5| McCraw et al [1985) Joyner | I9R3J Divine et al. 11985) Divine Hnd Barron |IV87] Rushlun and Aldersun 11981a) Rusltlnn and Aldersun | I9R3) Hnnis et at, j 1979) Tbdriault and Goulet [1979] Thdrisult and Proveiu-her 11987) Omstie [1987] Nakamura [I987J Company fiuility Industrywide 17 refineries Industrywide 17 refineries Industrywide 17 refineries 19 companies Refineries ami other facilities AMOCO 10 refineries Chevron Richmond, id Segnntlo Refineries Exxon Baton Rnugc Refinery Exxon Baton Rouge,. Baytown, Bayway Refineries Gitir Port Arthur Refinery Gulf Port Arthur Refinery Mobil Beaumont Refinery Mobil Paulsborn Refinery Mobil Torrance Refinery Shell Shell Texaco Texaco Industrywide 3 companies Imperial Oil Shell Canada Shell Canada Wood River Refinery Deer Park Refinery 13 refineries A other facilities Production + pipeline 8 UK refineries UK distribution centers Refineries and oilier facilities East Montreal Refinery East Mongrenl Refinery Industrywide OH companies in Australia Industrywide 51 Japanese refineries . Pullow-up period Length Minimum Cohort Person T**t :tl years employment sire year* denihs 1962-1971 in 1 + years 20.163 137,153 1 165 1962-1971 1962-1980 1977-1979 1970-1982 1950-1980 in 19 3 13 31 ! + years l 4- years none 6 + months 1 + years 20,163 20.163 55.007 10.763 14 179 137.15,1 297.591 122.607 104.954 266.470 1.165 \.W19 5tl2 *JR3 2.22 1970' 1977 1970-1977 R 8 1 4- month 8,666 52.791 1.199 1 4- months 2I.69H 137.702 1..I9H 1937-1978 1937-1983 1945-1979 1946-1979 1959-1978 42 47 35 34 20 1973-I9R2 i973~l98l 1947-1977 1946-1980 1950-1975 1950-1975 1964-1973 1928-1975 1928-1981 in 9 31 35 26 26 10' 48 54 None None 1 4 years 1 + years 1 4 years 15.095 15.095 6.139 4.263 1.621 372.505 427.394 123.354 93.434 23.154 1269 5.470 1.582 1. i (>4 2511 None None 5 4- years 6 4 months 1 4 years 1 4 years 5 < years 5 f-years 5 4- years 3.976 19.077 11.098 34,781 23.306 15.032 1.205 1.207 358.029 220.414 575.9X2 397.569 82.057 25.224 640 237 4024 I.KHfi 4.406 3,926 1,5M 175 I9BM986 1961-1981 21 65 + years 10.489 40.214 .1 + years 19.475 19.476 133 451 EXX-MOR-003362 286 Wong and Raabe information of the cohort members, usually in the range of several to tens of thousands. Included in this category are the studies of the Chevron Richmond and El Segundd refineries [Wong et al., 1986]; the Gulf Port Arthur refinery [Wen et al., 1983, 1984a]; the Mobil Beaumont [Morgan and Wong, 1984], Paulsboro [Morgan and Wong, 1985], and Torrance refineries [Enterline and Henderson, 1985]; and the Texaco companywide study [Divine et al., 1985; Divine and Barron, 1986, 1987]. These cohorts have aiso generated a number of reports based on specific subcohort analyses. For example, the Gulf Port Arthur data were the bases for two subcohorts: employees assigned to the benzene [Tsai et al., 1983] and to the MEK dewaxing units [Wen et al., 1985]. Based on the cohort data, a few nested case- control studies have also been conducted. An example was the leukemia study on the Shell employees at the Wood River refinery [Austin et al., 1986]. In Britain, an industrywide study of eight refineries was conducted [Rushton and Alderson, 1981a], A subsequent case-control study of leukemia was carried out within this cohort [Rushton and Alderson, 1981b]. A study consisting of employees at the distribution centers of three companies in England was also conducted by the same investigators [Rushton and Alderson, 1983]. There are also a number of cohort mortality studies of refinery workers from Canada, Australia, and Japan. Hanis et al. [1979] reported the mortality experience of the Imperial Oil (Canada) employees at several refineries and other facilities. The Theriault and Goulet [1979] report was based on the mortality experience of the Shell East Montreal refinery, which was subsequently updated by Theriault and Provencher [1987], The Australian.''study was based on the first 3 years of mortality data accumulated in the prospective registry in the petroleum industry sponsored by the Australian Petroleum Institute [Christie, 1987]. Finally, a brief report from Japan described the results of a mortality study of Japanese petroleum workers sponsored by the Petroleum Association of Japan [Nakamura, 1987]. The studies described above have been reviewed in detail. However, because of - the limited space available, detailed comments on specific studies will not be discussed here. Because of the study design, the "defined population at risk, the availability of mortality rates, the defined exposures, and the completeness and length of follow-up, these studies were judged to be adequate methodologically and to be of sufficient quality to form the basis of. our assessment below. A number of studies that we have reviewed were excluded from our assessment because of various serious limitations. Several studies that fell into this category were proportional mortality ratio (PMR) studies based totally on death certificates available to union locals [Thomas et al., 1980; Thomas et al., 1982a; Reeve et al., 1982]. In addition to the well-known methodological deficiencies of PMR studies [Wong and Decoufle, 1982; Wong et al., 1985], these particular studies also suffered from incomplete ascertainment of deaths [Wong and Tabershaw, 1980]. For example, in the Reeve et al. [1982] study, deaths were identified through local union death notifications, obituaries, word of mouth, and lists of surviving families to whom bibles were sent. In the Thomas et al. [1980] study, it was stated that death certificates for 10% of the deaths identified were not retrieved. Furthermore, employment kicrrvr-ioo Tliara O t J 1 n> kl o. +*/- ikaro DK^TD *<* r\ --* ,*4 t-Vio. * .o-rv EXX-MOR-003363 Cancer Epidemiology in Petroleum Industry 287 Two company-sponsored studies were also excluded from our assessment because of methodological or data deficiencies. The study conducted by Baird [1967] on Humble Oil (Exxon) employees was very crude judging by today's standard. The number of deaths reported in that study was likely incomplete, and the estimate of employees at risk was most likely inaccurate. The Thorpe [1974] study of eight Esso European facilities was likewise based on incomplete death ascertainment and inaccurate estimate of population at risk. In addition to studies conducted on employees in the petroleum industry, there are also a number of population-based case-control studies that included petroleum or petroleum-related occupations in the analysis [McLaughlin et al., 1984; Mommsen et ah, 1983; Lin and Kessler, 1981; McLaughlin el al., 1985; Howe et al., 1980; Stemhagen et al., 1983]. The relationship between various occupations and cancers of the. bladder, kidney, pancreas, and liver was investigated in these studies. None of these studies provided adequate information about the nature of exposure or details of employment. Methodological problems (such as control selection, recall of employ ment histories, adjustment of confounding factors in analysis) further complicated the interpretation. With the exception of the concern for kidney cancer raised by the API animal work [MacFarland et al., 1984], none of these cancer sites has been implicated by occupational studies. These studies were excluded from our final assessment. RISK ASSESSMENT BY CANCER SITE In this section, we summarize the epidemiologic data from the cohort mortality studies on petroleum workers by individual cancer site. If a study had been updated, (be updated data were used. Industrywide studies are not included in the data summarization, since part of the data overlap those in individual company studies. Subcohort analyses are not included in the data summarization either, since the data (or the entire study are utilized. In evaluating the data for causal relationship, some of the criteria set forth by Hill [1965] are used. In particular, we emphasize, to the extent possible, the strength of association, consistency of findings, and dose-response relationship. However, our application^ Hill's criteria is limited by the data available from the original studies. An overall index for strength of association is a summary standardized mortality ratio (SMR) or meta-SMR based on all the available appropriate data. A method of calculating the summary SMR or meta-SMR is given below. In terms of dose- response relationship, the issue is more complicated. Most of the studies did not provide any dose-response analysis. Few industrial hygiene data existed in the industry prior to the mid-1970s, whereas relevant exposures occurred much earlier. In some studies, broad exposure categories were used. Some used length of employment as a surrogate for length of exposure. Even for these latter studies, the groupings were so different that no quantitative summarization can be done without access to the original data. Thus, dose-response results can only be used in a qualitative manner in our assessment. The statistical procedure for data summarization (meta-analysis) consists of summing up the observed and the expected deaths for a specific cancer site from al) ffrtrrtwi*. onH rnlrnlafino rhp snmmarv SMR or meta-SMR. (The Hanis et al. 288 Wong and Raabe the data summarization, thus "adjusting" for individual studies while preserving the original adjustment using substrata specific to age, sex, race, calendar year, and country. The summary or meta-SMR is automatically weighted by the size of the individual studies (observed and expected deaths). Chi-squares are calculated to test the significance of the summary SMR. Ninety-five percent confidence intervals are also calculated for the summary SMR (presented in the last table). Similar procedures have been used in summarizing data in several epidemiologic reviews: asbestos and gastrointestinal cancer [Morgan et al., 1985], artificial sweeteners and bladder cancer [Morgan and Wong, 1985], formaldehyde and respiratory cancer [Nelson et al., 1986], chemical dyes and bladder cancer [Matanoski and Elliott, 1981], and radiation and colon cancer [Day, 1985]. In particular, this method was discussed in the interpretation of negative epidemiological evidence for carcinogenicity by the Inter national Agency for Research on Cancer [Wald and Doll, 1985]. Although other statistical procedures for meta-analysis are available, the meta-SMR procedure is used in this review because of its simplicity in computation and interpretation, its preservation of the adjustments in the original studies, and its resemblance to the analyses in the original studies. An additional benefit of meta-analysis is that it comprehends the multiplecomparison problem exhibited in these cohort studies. Most of the cohort studies that we reviewed examined over 40 individual cancer sites each. Statistically significant results for as many as two sites could have occurred by chance alone given a 95% confidence level. By our examination of these multiple sites across these studies, chance excesses or deficits for a particular cancer site in individual studies are less likely to influence the overall interpretation. Since the summary SMR is based on the expected deaths derived from national mortality statistics in the original studies, the issue of appropriateness of applying national rates to specific locales applies to the meta-analysis in this review as well. To be consistent with the data presented in the original papers, we have kept two decimal places in the expected numbers of deaths. Although the summary SMR gives an overall risk index for the industry as a whole, it has the disadvantage of dilution. If the risk is low and affects only a small portion of the employees instead of the entire industry, an overall summary index may not be sensitive. Therefore, data from each study have to be examined for any dose-response relationship and any significant risk for any subcohort. Unfortunately, few studies provided analysis by length of employment on a comprehensive basis, i.e., for ail causes of death. Those reporting comprehensive analysis by length.of employment included the Chevron Richmond and El Segundo refineries [Wong et al., 1986], the Mobil Beaumont refinery [Morgan and Wong, 1984], the Mobil Paulsboro refinery [Morgan and Wong, 1985], and the Mobil Torrance refinery [Enterline and Henderson, 1985]. Length of employment analysis was done for a few selected causes of death in the Gulf Port Arthur study and in the Texaco study. Rushton and Alderson reported length of employment analysis for selected causes of death as well as selected subcohorts in the British refinery and distribution center studies. In addition to length of employment analysis, another analysis that is helpful in establishing an epidemiologic association is an examination of the data by latency or EXX-MOR-003365 Cancer Epidemiology in Petroleum Industry 289 TABLE 11. Summary of Mortality From Cancer of All Sites, in Cohort Studies of Refinery ami Other Petroleum Employees* Author year Facilir\ Observed Expected Statistical Latency Duration death* deaths SMR significance analysis analysis Tahrrshav. Cooper Aasc j 197-4JJ Wnng IIVH0)1 Kaplan 119861" Schotienfeld ei al ;i9Bll"b Nelson 11985] Wong et al. (1986] Hams el al [1982]" Hanis cl al- f 1985a] Wen el al. f 19B2]1 Morgan and Wong 11984] Morgan and Wong [1985] Emeriinc and Henderson [1985] McCrawet al. 11985] Joyner 11983] Divine ei al. (1985] Divine and Barrow 119B7] Rushton and Alderson |19Bla] Rushtor and Alderson (1983) Theriault and Goulet (1979]* Thenaul! and Provencher [1987) Christie [1986] Nakamura (1987) 17 refineries 17 refineries 17 refineries U-S. refineries 10 refineries Richmond. El Segundo Refineries Baton Rouge Refinery Baton Rouge. Baytown. Bayway Refineries Port Arthur Refinery Beaumorn Refiner}' Paulsboro Refinery Torrance Refinery Wood River Refinery Deer Park Refinery 13 refineries & others Production & pipeline 8 UK Refineries UK distribution centers East Montreal Refineries East Montreal Refineries Australia 51 Japanese Refineries Total 365 79.3 127 273 462 249 666 664 346 243 56 161 767 393 1.147 1.002 25 39 52 134 6.405 32" 52 0 HO r 915.59 168.77 330.10 610.70 0.87 0.75 0.83 0.75 r f e f 270.40 0,92 705.00 0.94 681.54 360.27 265.24 69.32 177.00 0.97 0.96 0.91 0.80 0.91 1.019.40 581.10 1.286.40 1.156,70 28.54 48.82 56.90 176.80 0.75 0.6B 0.89 0.87 0.88 0.79 0.91 0.76 f r r r r 7.525 29 0.85, cp<.oooon xd s" xd x" x" . xd xd x xd xd xd xd xd *Note: Blank entries indicate that the observed and expected deaths are not reported by the authors. "Excluded from the calculation of summary' SMR bOnly refinery employees are included. cOnly male hourly employees with 1 or more years of employment are included. dAna)ysis done. ep < -05. fp < .01. and the Texaco studies included some latency analysis by specific causes of death. Finally, several studies fTCA, 1974; Wong et al., 1986; Wen et al., 1984a; Rushton and Alderson, 1981a, 1983] provided some analyses by hire date. Cancer of All Sites Combined (ICD8, 140-2Q9) Table II presents the observed and expected deaths and the corresponding SMRs for cancer of all sites from the eligible cohort mortality studies of petroleum employees from the United States, Canada, Great Britain, Australia, and Japan. The all-cancer SMR ranged from 0.68 (p < .01) for Texaco production and pipeline employees to 0.97 for Gulf Port Arthur male workers. No study found any excess or trend (significant or otherwise) for all cancer. In fact, most of the larger studies crnifirant deficits for all cancer. The deficit was quite consistent for all EXX- 290 Wong and Raabe TABLE III. Summary of Mortality From Digestive Cancer in Cohort Studies of Refinery and Other Petroleum Employees" Author/year Facility Observed Expected Statistical Latency Duration deaths deaths SMR significance analysis analysis Tabershaw Cooper Assc {1974]* Wong |]980)* Kaplan (1986]* Schottenfeld et al. (I9811*-b Nelson {1985] Wong ct at [ 19B6] Hanis et al, [19821" Hanis ei al, (1985a{ Wen ei al. (19R3|C Morgan and Wong [1984) Morgan and Wong [ 1985] Enlerline and Henderson [1985) McCraw et al. |]985] Joyner |1983) Divine et al. [1985] Divine and Barron [1987] Rushton and Alderson [1981a) Rushton and Alderson [1983] Theriuult and Goulet [ 1979]" Theriault and Provencher [1987| Christie [1986) Nakamura 11987) 17 refineries 17 refineries 17 refineries U.S. refineries 10 refineries Richmond, El Segundo Refineries Baton Rouge Baiun Rouge. Baytown. Bayway Refineries Port Arthur Refinery Beaumont Refinciy Paulsboro Refinery Torrance Refinery Wood River Refinery Deer Park Refinery 13 refineries Sc others Production & pipeline 8 UK refineries UK distribution centers East Montreal Refinery East Montreal Refinery Australia 51 Japanese refineries 81 233 31 97 125 73 1% 86 80 10 215 105 12 is (5 95.71 0.84 246.91 41.12 82.30 174.62 0.90 0.75 1.17 0.71 74.90 (1.97 195.30 1 00 e 113.04 80.90 18.30 0.76 0.98 0.54 e 306.80 0.70 168.70 0.62 e e 10.22 18.20 16.70 1.17 0.98 0.90 * x` x- x" x` xd x` X* xd x" XJ X* 7? Total 947 1.174.86 0.81 ipc.oooon 'Note: Blank entries indicate that the observed and expected deaths are not reported by the authors. "Excluded from the calculation of summary SMR. bOnly refinery employees are included. cOnly male hourly employees with 1 or more years of employment are included. dx = Analysis done. ep < .01. Digestive Cancer (ICD8, 150-159) For digestive cancer, four studies showed a' significant deficit (Table HI). Of ail the studies listed in Table HI. only one study, the AMOCO companywide study, showed a slight nonsignificant excess (SMR = 1.17). A closer examination of the AMOCO data indicated that out of the ten refineries investigated, the digestive cancer excess came from only one refinery (Whiting) and among employees hired before 1940. This isolated finding was not consistent with the rest of the data. For the industry as a whole, the digestive cancer meta-SMR was 0.81 (947 deaths observed), significant at the .00001 level. Stomach Cancer (ICD8, 151) Table IV presents stomach cancer SMRs from the eligible cohort mortality studies. No study detected any significant excess. The stomach cancer SMRs ranged * Ct 'J A t --. All ^/s. r\ ni'riAline nrnrUare fn 1 ( nht EXX-MOR-003367 Cancer Epidemiology in Petroleum Industry 291 TABLE TV. Summary of Mortality From Stomach Cancer in Cohort Studies of Refinery and Other. Petroiehrh Employees Author'year Facility- Observed Expected Statistical Latency Duration deaths deaths SMR significance analysis analysis Tabershaw Cooper Assc {1974)* Wong I3 9K0J* Kaplan 11986]" SchottenJeld el al. [1981]"*11 Nelson (1965) Wong el aL [ 1986] Hanis ei al |1982]fc, Hanis el aJ. | 1985a] Wen el at. 1I9R3]1 Morgan and Wong (198-41 Morgan and Wong (1985) Ewerline and Henderson |)98-`i) McCraw et al. [1985) loyner[19B3) Divine et a], (1985) Divine and Bamm [1987] Rushton and Alderson f 1981 a] Rushton and Alderson [1983] TTnfcriaull and Goulet J]979}" Theriault and Provencher (1987] Christie 119B6] Nakamura 11987} Total 17 refineries 17 refineries 17 refineries U.S. refineries 10 refineries Richmond. EJ Scgundo Refineries Baton Rouge Refinery Baton Rouge. Baytown. Bayway Refineries Pon Arthur Refinery Beaumonl Refinery Paulsboro Refinery Torrance Refinery Wood River Refinery Deer Park Refinery 13 refineries & others Production & pipeline 8 UK refineries UK distribution centers Easl Montreal Refinery East Montreal Refinery Australia 51 Japanese refineries 43 6 17 26 n 33 63 2d 14 4 45 13 167 123 7 49 5S5 44,06 5.95 11.90 34.10 0,96 1.00 1.43 0.76 13.70 34.30 0.80 0.96 56.69 26.55 17.56 3.36 i.n 0.90 0.79 1.18 63.90 33.00 160.86 144.47 0.70 0.39 1.04 0.85 e r 4.36 1.60 68.90 659.95 0.71 0.89 c tp<.0Q5) x` xc Xd xd xd xd re" xd xd xd xd xd 'Note: Blank entries indicate that the observed and expected deaths are not reported by the authors, "Excluded from the calculation of summary- SMR. bOnly refinery employees are included. 'Only male hourly employees with 1 or more years of employment are included. dx = Analysis done. ep < .05. " rp < .01. stomach deaths were observed and 4.36 were expected. The East Montreal investi gation [Theriault and Provencher, 1987] was a very small study with only 1,207 employees. Another Canadian study based on Imperial Oil employees [Hanis et al., 1979] detected a significant relative risk of esophageal and stomach cancer among the exposed employees when compared to the nonexposed employees (not shown in Table IV); the excess came from refineries in Ontario and Quebec. Exposure classification in this study was based on partial employment histories (post-1964); therefore, misclassification was likely, and analysis based on job or exposure categories was subject to error. In both Canadian studies, the stomach cancer excess came primarily from Quebec. The possibility of a confounding factor associated with that region should be investigated. In the United States, Nelson [1985] found a nonsignificant stomach cancer excess in AMOCO employees (SMR = 1.43). Similar m iKa fmriino in total digestive cancer, the stomach cancer excess came from the EXX- 292 Wong and Raabe TABLE V. Summary of Mortality From Pancreatic Cancer in Cohort Studies of Refinery and Other Petroleum Employees" Author/year Facility Observed Expected Statistical Latency Duration deaths deaths SMR significance analysis analysis Tabershaw Cooper A..VC 11974|* Wang 119B0|* Kaplan [19861* Schottenfcld ec al, []981p"b Nelson [I9B5] Wong et ai. [1986] Hanis et ai. 119821* Hants al. f I985aj Wen el al. 119S3J" Morgan and Wong [1984] Morgan and Wong [19851 Erucrlinc and Henderson [I985j McCraw cl ah [1985] Joyner (19831 Divine el al. f S 985] Divine and Barron [1987] Rushton and Alderson [198!a| Rushton and Aldereon [19831 Theriault and Goulet |I979| Theriault and Prnvcncher [1987] Christie [19861 Nakamura [1987] 17 refineries 17 refineries 17 refineries U.S. refineries 10 refineries Richmond, El Segundo Refineries Bacon Rouge Refinery Bacon Rouge. Baytown. Bayway Refineries Port Arthur Refinery Beaumont Refinery Paulsboro Refinery Torrance Refinery Wood River Refinery Deer Park Refinery 13 refineries & others Production & pipeline 8 UK refineries UK distribution centers East Montreal Refinery East Montreal Refinery Australia 51 Japanese refineries Total 44 10 18 20 23 42 37 23 15 3 62 27 50 39 336 51.25 8.85 17.00 33.78 0.86 1.12 1.05 0.59 15.10 J9.20 1.52 1.07 37.80 20.14 14.81 3,96 0.99 1.14 1.01 0.75 e "v x" X* 57.70 32.00 51.55 46.89 1.07 0.84 0.97 0.83 x" 354.83 0.95 (p=.3Z> x" x" x" xd xd `Note: Blank entries indicate that the observed and expected deaths are not reported by the authors. u Excluded from the calculation of summary SMR. bOniy refinery employees are included. bOnly male hourly employees with 1 or more years of employment are included. dx = Analysis done. 'p < .05. cancer; the SMR was 0.70 in the refinery employees and 0.39 in the production arid pipeline employees. The Japanese industrywide study also found a significant deficit of stomach cancer. Mortality deficits from stomach cancer were also found in other studies. As Table TV indicates, the Chevron Richmond and El Segundo refineries and the Mobil Paulsboro refinery showed a modest deficit in stomach cancer. Rushton and Alderson [1983] found a lower than expected stomach cancer mortality in distribution workers (SMR = 0.85, p = .05, borderline significance). Summing over all the eligible studies, the stomach cancer meta-SMR was 0.89 (585 observed deaths, p < .005). Thus, taking ail studies into consideration and weighting the finding of each study by its sample size, the petroleum industry as a whole demonstrated a significantly lower stomach cancer mortality. Pancreatic Cancer (ICD8, 157) EXX-MOR-003369 Cancer Epidemiology in Petroleum Industry 293 TABl.E VI. Suonmary or Mortality From Long Cancer in Cohort Studies or Refiner; and Other PetroiMtUB Employees* .. . Auihor-vtar FoeiJily Observed Expected Statistical Latency Duration deaths deaths SMR significance analysis analysis Tahcrshau uwjv.cr A.w. 11974]* Wong Il90f Kaplan 119861' Schonenfcld. at i, [1981J*-1' Nelson [ I9E?) Wong a al II986] Hanis a al. |*822)* Hams et ai jl9K$n] Wen c! al Morgan and "Wong j 1984) Morgan ant Wong 119851 Emerline ur*6 Henderson (1985) McCrew el al, {1985} Joyner |19K3] Divine el ai (1985] Devine and Barton 11987] Rushton arvj Alderson [)98la) Rushton anr Aldcrsctn [1983) Th6riaull andUuoulet f 19797* Theriault and Pmwencher (1987) Christie [1986] Nakamura {%9C7) 17 refineries 17 refineries I" refineries U,S, refineries 10 refineries Richmond. E! Scgundo Refineries Baton Rouge Refinery Baton Rouge, Baytown, Bavu-ay Refinenes Port .Arthur Refinery Beaumont Refinery Paulsboro Refinery Torrance Refinery Wood River Refinery Deer Park Refinery 13 refineries & others Pnxiuclion & pipeline 8 UK refineries UK distribution centers East Montreal Refinery East Montreal Refinery Australia 51 Japanese Refineries Tout! 90 270 48 77 124 78 209 182 J01 61 24 182 109 416 384 3 7 9 18 1.903 112 71 0.79 3 f 7 72 65.07 125.90 185 42 0.85 0.73 0.6) 0.66 85.30 0,91 219.80 0.95 IK3.22 100.11 76.38 25.86 0,99 1.00 0.79 1.00 r c e r f 306.00 179.00 531.70 482.83 8,48 15.78 18.80 19.80 0.59 0.61 0.78 0.80 0.35 0 44 0.48 0.91 2.469 60 0.77 f r r f e e tp-c.0000)1 K Xs X* x` xd x" x" x" x x" x" X4 *Noie: Btank^-onlries indicate thar the observed and expected deaths are not reported by the authors. "Excludes fromn the calculation of summary SMR. bOnly refeerryecmployees are included. "Only rrufe roetirly employees with 1 or more years of employment are included. dx = AmJvsbk done. ep < .05. rp < .01. . significantly low mortality from pancreatic cancer among Chevron employees at the Richmond and E) Segundo refineries [SMR = 0.59, p < .05). Hanis et al. [1982] reported an SMR of 1.52 (p = .05, borderline significance). However, the subse quent Hanis et al. [1985a,b] study found a much reduced SMR of 1.07 (p = .65, not significants). No dose-response relationship was demonstrated by any of the studies: For tthe entire industry, the pancreatic cancer meta-SMR was 0.95 (336 ob served deaths:, p = .32). The data did not indicate an excess of pancreatic cancer in refinery w.orfe ers. Lung Caneerr (1CDB, 162, 163) As indie:ated in Table VI, most of the studies showed a significant lung cancer deficit. Fourcof the five industrywide studies showed a significant deficit. The U.S, industrywide istudy, based on 270 lung cancer deaths, showed an SMR of 0.85 (p < .051. sifrnifc antlv less than the expected [Kaplan, 1986]. The two industrywide 294 Wong and Raflbe studies in Britain also reported similar results; SMR of 0.78 in refinery workers [Rushton and Alderson, 1981a] and SMR of 0.80 in distribution workers [Rushton and Alderson, 1983], both significant at the .01 level. The Australian registry for the petroleum industry also reported a significant lung cancer deficit (SMR = 0.48). The Japanese industrywide study showed a slight nonsignificant deficit. Significant lung cancer deficits were also detected in AMOCO refineries [Nelson, 1985], Chevron Richmond and El Segundo refineries [Wong et al., 1986], Texaco refinery and production workers [Divine et al., 1985; Divine and Barron, 1987], and the Shell East Montreal refinery [Theriault and Goulet, 1979]. The Imperial Oil study by Hanis et al. [1979] concluded that there was a significant lung cancer excess among the exposed employees when compared to the nonexposed employees (not shown in Table VI). As commented above, because of the limited job history data, misclassification of exposure was likely. Furthermore, most of the lung cancer cases in that study were office workers, clerks, boiler makers, and construction workers. In some of these occupations, asbestos exposure was likely. Furthermore, when the data were examined by exposure category, the medium exposure group had a lung cancer relative risk (RR) of 0.39, when compared to the nonexposed. This was not consistent with a causal relationship. The most serious evidence against the positive finding in this study is that its lung cancer result was totally inconsistent with the rest of the industry worldwide. No study has adequate smoking data for a proper analysis incorporating such information. However, in a few studies, some smoking data were available. In the Nelson [1985] study, smoking data were available on 60% of the cohort members, and among those with such information, 69% used tobacco. In the Exxon cohort, smoking information was available on 70% of the employees, among whom 73% smoked cigarettes [Hanis et al., 1985b]. In the Enteriine and Henderson [1985] study, smoking data were collected on 71% of the Mobil Torrance cohort, with 65% among them having ever smoked. Van Peenen et al. [1984] reported that 72% of white male refinery hourly employees studied at Standard Oil of Indiana smoked cigarettes, pipes, or cigars. These Figures from the above companies, although limited, were quite comparable to the percentage of smokers in the general population. Thus, smoking was not likely a serious confounder when comparing petroleum employee cohorts to the general population. : Summing up all the data on lung cancer, the meta-SMR was calculated to be 0.77. This SMR was extremely stable, with 1,903 observed deaths, and was significant at the .00001 level. Based on the totality of the data, there was a significant lung cancer deficit in petroleum workers. Skin Cancer (ICD8, 172, 173) The reported skin cancer mortality in petroleum workers spread over a wide range (Table VII): from a low of 0.24 among the Mobil Paulsboro employees [Morgan and Wong, 1985] to a high of 2.16 in the British industrywide study of refinery employees [Rushton and Alderson, 1981a], In the Rushston and Alderson study, only melanoma was reported, and 11 of the 14 melanoma deaths came from two of the eight refineries. No dose-response analysis was provided in the study, and EXX-MOR-003371 Cancer Epidemiology in Petroleum Industry 295 TABLE Vtl. Summary of Mortality From Skin Cancer in Cohort Studies of Refinery and Other Petroleum Employees* A mbor/ycar Facility Observed Expected Statistical Latency Duration deaths deaths SMR significance analysis analysis Tabcrshau Cooper Asst. | 1974]* Wane 11980(` Kaplun 119R6J' Schonenfeld et a!. |19Sl]a-1' Nelson 11985] Wong ei hi j 1986] Hunis ei aj (19E2)" Hanis el al. 11985a] Wen c( al. |i9E3)r Morgan and Wong ([984| Morgan and Wong [1985] Enreriine and Henderson |J985) McCraw et al. [t985| Joyner 11983 J Divine el al. |198S] Divine and Barron \}9$7) Rusbion and Alderson (1981a] Rushton and Alderson (1983] Theriault and Goulet |I979]" Theriault and Porvencher (1987] Christie II9K6] Nakamura (1987] 17 refineries 17 refineries 11 refineries U.S. refineries 10 refineries Richmond. 1 Segundo Refineries Baton Rouge Refiner)' Baton Rouge. Baytown.. Baywax Refineries Pan .Arthur Refinery Beaumont Refinery Paulsboro Refinery Torrance Refiner) Wood River Refmeiy Deer Pari Refinery 13 refineries & others Production & pipeline 8 UK refineries UK distribution centers East Montreal Refinery East Montreal Refinery Australia 5J Japanese refineries Total 15 3tml 11 10 9 10 7 I KmT 14 7 H(mr 5 4(m)c 93 14 05 3.35 S.70 10.81 1.07 0.89 1.93 0.92 9 10 0.99 9.53 5.03 4.06 1 02 1.05 1.39 0.24 0.98 16.70 9.80 6.48 8.52 0.84 0.72 2.16 0.59 2.48 1.62 89.23 1.04 f tp = .70| X11 x" xd xd xd td x xa xd xd "Note: Blank entries indicate that the observed and expected deaths axe not reported by the authors "Excluded from the calculation of summary-' SMR. hOn!y refinery employees arc included. "Only male hourly employees with 1 or more years of employment are included. dx = Analysis done. e(m) = melanoma only. . rp < .01. Beaumont refinery (SMR = 1.39, 7 observed deaths) and the Australian registry of petroleum workers (SMR = 1.62, 4 observed). In both cases, the numbers of observed deaths were small, and the SMRs were not statistically significant. Further more, analyses by length of employment and latency in the Mobil Beaumont study did not find any association between skin cancer and employment [Morgan and Wong, 1984). On the other hand, skin cancer deficits were reported in the Mobil Paulsboro refiner)' [Morgan and Wong, 1985], the Texaco refinery and production employees [Divine et ah, 1985; Divine and Barron, 1987], and the distribution workers of three companies in Britain [Rushton and Alderson, 1983]. The meta-SMR for skin cancer based on all the eligible studies in the petroleum industrv was 1.04 (93 observed deaths and 89.23 expected), with a p-value of .70. EXX- 25X> Wong and Raabe TABLE VIII. Summary of Mortality From Prostatic Cancer in Cohort Studies of Refinery and Other Petroleum Employees Author/year Facility Observed Expected Statistical Latency Duration deaths deaths SMR significance analysis analysis Tabcrshaw Cooper Assc.. [1974]* 17 refineries Wong [ 19B0J- 17 refineries Kaplan [1986]* 17 refineries Schottenfeld et al, |198l]B`b LLS. refineries Nelson (19851 10 refineries _ Wong cl al. |I9B6] Richmond, El SegundD Refineries Hurtis et ah [1982|* Baton Rouge Refinery Hams el al. [1985a| Baton Rouge. Baytown. Bayway Refineries Wen cl al. [19831' Port Arthur Refinery Morgan and Wong [1984] Beaumont Refinery Morgan and Wong [1985] Paulsboro Refinery Enterline and Henderson (1985] Torrance Refinery McCraw ct al. [19851 Wood River Refinery Joyner [1983| Deer Pari: Rtfinery Divine et ah [1985] 13 refineries & others Divine and Barron [1987] Production & pipeline Rushton ami Alderson [198!a| 8 UK refineries Rushton and Alderson [19B3] UK distribution centers Theriault and Goulet [1979]* East Montreal Refinery Theriault and Provencher }1987j East Montreal Refinery Christie 119861 Australia NaJoimura [1987] 51 Japanese refineries Total 60 5 14 34 19 56 69 36 28 5 62 41 47 53 445 58.81 5.30 J6.20 43.36 1,02 0.94 0.86 0.78 28.60 0.66 73.30 076 59.98' 32.06 20.68 4.53 1.15 1.12 1.35 1,10 73.10 43.70 45.87 48.57 0.85 0.94 1.02 1.09 461.35 0.96 (p= 45) X4 IXC *d xJ x" x" x" xxc xd XJ xd xd Note: Blank entries indicate that the observed and expected deaths are not reported by the authors. "Excluded from the calculation of summary SMR. bOnly refinery employees are included. "Only male hourly employees with 1 or more years of employment axe included. dx = Analysis done. "xx = Analysis done for subcohort only in subsequent report (see text). However, it must be pointed out that the above assessment of skin cancer based on mortality data is limited, since most forms of skin cancer are not fatal. Prostatic Cancer (1CD8, 185) >* - - Table VIII presents the results of prostatic cancer from the cohort studies of petroleum workers. In most cases, the observed mortality was close to the expected. The highest SMR was observed in the Mobil Paulsboro refinery, with a nonsignificant SMR of 1.35 based on 28 deaths. Detailed analysis showed that the white male employees with 20 years of service at the Paulsboro refinery experienced a signifi cantly elevated SMR of 1.60. However, analysis by work location did not reveal any particular high-risk area. Two small studies of Methyl ethyl ketone (MEK)-dewaxing employees at refineries have shown a significantly increased rate of prostadc cancer [Enterline, 1978; Wen et al., 1985]. These included the Shell Deer Park refinery (SMR - 9.52, 2 observed deaths, p < .01) and the Gulf Port Arthur refinery (SMR = 2.48 for the EXX-MOR-003373 Cancer Epidemiology in Petroleum Industry 297 nonwhiles based on 7 deaths, p < .05: and SMR = 11.49 for those with 20+ years of sendee and 20-29 years of latency, p < .05). However, a similar study of MEK-dewaxing employees at the Shell Haven and Stanlow refineries in Britain [Alderson and Rattan, 1980] did not find any deaths from prostatic cancer, when the expected number was 0.46. Based on these three studies, the meta-SMR for MEKdewaxing employees was 1.9? (p < .05). It should be noted that all three MEKdewaxing studies were very small. The meta-SMR for the industry as a whole was 0.96, based on 445 observed deaths (p = .45). Thus, petroleum employees as a group had a similar mortality from prostatic cancer as the general population. However,' data from some refineries suggested that petroleum employees working at MEK-dewaxing units might experi ence a higher prostatic cancer risk. At this point, a firm conclusion on the relationship between prostatic cancer and employment at MEK-dewaxing units cannot be made based on the currently available data. Kidney Cancer (1CDS, 189) Table IX presents the results of kidney cancer from petroleum cohort mortality studies. Interest in this cancer site was aroused after it was reported [McFarland et al., 1984] that male rats experienced an increase in kidney tumors after chronic exposure to wholly vaporized unleaded gasoline. The lowest SMR was observed in Texaco production and pipeline workers, with an SMR of 0.35 (5 observed deaths p < .05). However, production workers have limited contact with downstream gasoline, and, thus, their mortality experience might not be directly relevant. The Chevron refinery study and the three Mobil refinery studies ail reported slight deficits in kidney cancer mortality with SMRs ranging from a low of 0.59 (3 death) to a high of 0.81 (12 deaths). Mortality excess from kidney cancer was observed in both Exxon and Gulf employees. For the three Exxon refineries at Baton Rouge, Baytown, and Bayway , the SMR for kidney was 1.43 (22 deaths, p = .09). The excess came from those hired before 1956, the SMR being 1.47 (22 deaths, p = .07). Unfortunately, no length of employment analysis was done. Wen et al. [1984b] reported some detailed analyses of kidney cancer based on the Gulf Port Arthur study. Overall, the Gulf Port Arthur cohort experienced a nonsignificant SMR of 1.23 based on 22 deaths with a p-value of .33 [Wen et al., 1984b], When the data were analyzed by length of'employment, a weak upward trend was found. For those who worked for 9, 10-20, and 2= 20 years, the SMRs were 0.93, 1.25. and 1.38, respectively. A subsequent nested case-control study detected several areas with nonsignificant increase in kidney cancer. The Japanese industrywide study did not report kidney cancer per se. However, for urinary tract cancer, the Japanese refiner)' workers experienced a borderline signifi cant excess (p = .05), with 7 observed deaths and an SMR of 2.05 [Nakamura, 1987]. As far as downstream gasoline is concerned, the more appropriate group to study is the distribution or marketing employees. In the British distribution employ ees, a nonsignificant excess from kidney cancer of 1.21 (based on 23 deaths, p = .37) was reported [Rushton and Alderson, 1983]. However, detailed analysis (in the original report to the sponsor but not reported in the published article) showed that 298 Wong and Raabe TABLE IX. Summary of Mortality From Kidney Cancer in Cohort Studies of Refinery and Other Petroleum Employees* Author/year Facility Observed Expected Statistical Latency Duration deaths deaths SMR significance analysis analysis Tabershaw Cooper assc. 11974-1* Wong (19801* Kaplan (1986]* Schocienfeid ct al. {19811*4*0 Nelson (1985) Wong et a). (19861 Hartis et al. [1982]a Hunis cl al. [.1985aj Wen el al. (19831c Morgan and Wong [1984] Morgan and Wong [1983] Entcriine and Henderson (1985) McCraw ei al. 119851 Joyner[19831 Divine el al. (1985) Divine and Barron [1987] Rushlon and Alderson [!98!a| Rushion and Alderson [19831 Theriault and Goulei [ 1979|` Theriault and PrDvencher [ 19871 Christie 11986] . Nakamura (1987) 17 refineries 17 refineries 17 refineries U S. refineries 10 refineries Richmond. El Segundo Refineries Baton Rouge Refinery Baton Rouge. Baytown. Bayway Refineries Port Arthur Refinery Beaumont Refinery Pauisboro Refinery Torrance Refinery Wood River Refinery Deer Bark Refinery 13 refineries Sc others Production 4 pipeline 8 UK refineries UK distribution centers East Montreal Refinery East Montreal Refinery Australia 51 Japanese refineries Total 15 9 12 9 19 6 4 i 24 5 22 23 147 22.09 0.68 8.40 14.82 1.07 0.81 5 80 15,40 1.55 1.43 14.87 7 90 6.17 1.69 1.28 0.75 0.64 0.59 25.10 14.10 21.74 19.05 0.96 0.35 1.01 1.21 r 149.24 0.98 i,p = .85) X*1 xxc X* X* x" .X* kXe t" t-1 *Note: Blank entries indicate that the observed and expected deaths are not reported by the authors. "Excluded from the calculation of summary SMR. bOnly refinery employees are included. cOn!y male hourly employees with 1 or more years of employment are included. dx = Analysts done. ' cxx = Analysis done in subsequent report (see text). rp < .05. " deaths, p = .05, borderline significance). A significant excess was detected among men aged 55-64 years (SMR = 1.89, p < .05). For the industry as a whole, the kidney cancer meta-SMR was 0.98, based on 147 deaths. The data were somewhat conflicting. On the one hand, production workers at Texaco experienced a significantly low kidney cancer mortality. On the other hand, drivers among British distribution workers experienced a borderline significant increase. The results from refinery studies ranged from nonsignificant deficits to nonsignificant excesses. Additional data, particularly from those exposed to downstream gasoline, are needed to resolve this issue. Brain and Central Nervous System (CNS) Cancer (1CD8, 191-192) Several studies of petroleum employees showed some nonsignificant excess of r'wc r-^rvi'f.r fTnhlp Yi Thp APf-MlsKPC rp.pistrv showed an SMR of 1.63 EXX-MOR-003375 Cancer Epidemiology in Petroleum Industry 299 TABLE X. Summary of Mortality From Brain Cancer in Cohort Studies of Refinery and Other Petroieunn Employer?* Auihor.'vear Facility Observed Expected Statistical Laiency Duration deaths deaths SMR significance analysis analysis Taberstau Coopwsr Aisc. I! 974)* ]7 refineries Wong 110801* 1 ~ refineries Kaplan 11986]* 17 rermeries Scbonenfeld cl all. |I9B1J*> Nelson [1985] 10 refineries Wong cl al. [198:6! Richmond. E3 Segundo Refineries Hams et al. T198I2J1' Baion Rouge Refinery Ham* et al, j 19815a] Baton Rouee. Baytown, Bavwav Refineries Wen a al. | ]93!t Ron Arthur Refinery Morgan and Wong |I9B4) Beaumom Refinery' Morgan and Wong [1985) Paulsboro Refinery Enterlme and Henderson 11985j Torrance Refiners' McCrew el al. [3985] Wood River Refinery Joyner [1-983) Deer Bart Refinen1 Divine et al. 11985] 13 refineries <5: others Divine and Barron J1987] Production & pipeline Rushton and Alderson [198! a] 8 UK refineries Rushton and Aldcrson [19831 UK distribution centers ibriaul! and Gorulet 119791* East Montreal Refinery !7>6nau)l and Pro-vcncher [1987] East Montreal Refinery hristie [1986J Australia valamuia [198?:! 51 Japanese Refineries 8 22 E 22 5 15 23 9 7 3 31 11 36 39 3 4 2 ) 1 29 24.63 4.90 9.60 17.48 0.71 0,89 1.63 0.83 1.25 4.90 13,00 1.02 1.15 22.85 8.26 6,64 2.09 1.01 1.09 1.05 1 43 28.00 15.90 44.77 36.54 0.77 1.87 2.70 1.11 0.69 0.80 1.07 3.89 2.13 0.74 xd xd xxc XX* x" x" xd X6 xd xd Xd xd X xa otal 210 209.7 1.00 (p = .991 Note: Blank entries indicate that the observed and expected deaths are not reported by the authors. Excluded frown the calculation of summary SMR. 'Only refinery employees are included. Only male bcouriy employees with 1 or more years of employment are included. 'x = Analyses done. '.xx = Analyses done in subsequent report (see text). brain cancer SMR of 1.63 was detected (not significant). In the Mobil Torrance refinery, an SMR of 1.43 based on 3 deaths was reported (p = .50, not significant). In Canada, the Shell Canada East Montreal refiner)' employees had an SMR of 2.13 (p = .12, not significant). All these studies were based on very small numbers (Table X). Bases! on the FMR study of the Gulf and Texaco Port Arthur refineries, Thomas et al. [19800 reported higher than expected brain cancer mortality among the refinery employees. A subsequent PMR study of brain cancer of the Gulf and Texaco Port Arthur and the Mobil Beaumont refineries also reported elevated brain cancer PMRs [Thomas et- al., 3982b]. Other subsequent studies addressing the issue of brain cancer and benign' brain tumors were also reported [Wong, 1980; Wen et al., 1982]. Wong [1980] comcluded that there was no brain cancer excess in the TCA industrywide .< m i----- A ,,,, 0 urac wwifimned bv a later update of the EXX- Although the SMR of 0.97 (30 deaths) for the entire refinery was similar to the expected, those with 20 or more years of employment experienced an SMR of 1.43 (p = .10, not significant), which was more than twice that for those with less than 10 years of service (SMR = 0.59). A subsequent case-control study by Thomas et ai. [1984] on three Texas refineries including the Gulf Port Arthur plant did not find any work category associated with brain tumor mortality. Another approach to evaluate the issue of brain cancer in petroleum workers is to examine the data for any dose-response relationship. A few studies provided analyses by length of employment. No trend was found for brain cancer with respect to length of service in the Chevron Richmond and El Segundo refineries [Wong et al., 1986], the Mobil Beaumont and Paulsboro refineries [Morgan and Wong, 1984, 1985] , and the laboratory workers at the Texaco refineries [Divine and Barron, 1987], Thus, analysis by length of service did not support a causal association between refinery employment and brain cancer. Brain cancer is a highly fatal disease. Little is known about its etiology, and no preventive behavior or medical intervention has demonstrated any significant bene fits. As such, brain cancer in industrial groups is not expected to be influenced by the healthy worker effect, and the corresponding SMR, in the absence of any occupa tional carcinogens, is expected to be close to 1. ' Overall, the petroleum workers experienced an identical brain cancer mortality as the general population (meta-SMR = 1.00, p = .99). Although some studies showed nonsignificant excesses from brain cancer, these studies were small. It has been demonstrated that industrial workers were more likely to have an overdiagriosis of brain tumors as a result of more complete medical evaluation, "diagnostic sensitivity bias" [Greenwald et al., 1981]. It has also been demonstrated that socioeconomic class can also influence diagnostic sensitivity for brain cancer [Levine, 1985], Further inves tigations in this direction on petroleum workers are needed. Lymphatic and Hematopoietic Cancer (ICD8, 200-209) Of the studies listed in Table XI, a few studies showed modest increases in lymphatic and hematopoietic cancer. A significant excess was found in Mobil Beaumont refinery employees (SMR = 1.47, 47 deaths, p < .05). Similar excesses, although not statistically significant, were also found in the Shell Wood River refinery and the participants in the Australian registry. The AMOCO study showed a signif icant deficit (SMR = 0.60, p < .05). Other studies based on a reasonable number of deaths indicated that the mortality was similar to the expected. Several studies with length of service analysis indicated an increased mortality from lymphatic and hematopoietic cancer with increasing length of service. These studies included the Chevron Richmond and El Segundo refineries [Wong et al., 1986] , and the Mobil Beaumont refinery [Morgan and Wong, 1984], On an industry basis, the meta-SMR was 1.03 (based on 460 deaths, p = .58). Individual study results were not consistent. Both significantly high and low SMRs were reported. It would, perhaps, be more informative to examine the specific subgroups within the broad category of lymphatic and hematopoietic cancer. Lymphosarcoma and Recticuiosarcoma (ICD8, 200) Cancer Epidemiology in Petroleum Industry 303 TABLE XL Summary of Mortality From Lymphopoietic Cancer in Cohort Studies or Refinery and Other Petroleum Employees* Authop'ycar Facility Observed Expected Statistical Latency Duration deaths death.- SMR significance analysis analysis Tahenthaw Cooper A,xsc, ] 1 ^7-4)* 1 7 refineries Wong ||980]` Kap.lan 119.86]* 17 refineries 17 refineries Scbottenfeld et al.-|1981)"`l U.S. refineries Nelson 13985) 10 refineries Wong et al. [1986] Richmond. El Segundo Refineries Hanis et al, [1982]* Baton Rouge Refinery Hants- et al, [ 19R5a] Baton Rouge. Baytown. Bavway Rcftnerie*- Wen el at 119S3]* Morgan and Wong 11984) Port Arthur Refinery Beaumont Refinery Morgan and Wong (1985) Paulsboro Refinery Emerline and Henderson |19B5] Torrance Refinery' McCraw el al. [1985] Joyner[1983] Divine et al. f 1985) Wood River Refinery Deer Part Refinery 13 refineries & others Divine and Barron [1987] Production & pipeline Rushton and Alderson 11981a] 8 UK refineries Rushton and Alderson 11983) UK distribution centers Tbdriaull and Goulel f 1979}* East Montreal Refinery Thenau.lt and Provencher |I9S7] East Montreal Refinery Christie (1986) Australia Nakamura ] 1987) 51 Japanese Refineries Total 88 17 64 25 55 62 47 ->* 5 22 106 48 3 3 8 460 82.50 1.07 28.50 60.97 0.60 1.05 23.00 61.R0 1.09 0 89 58.62 31.K9 24.52 6.63 15.20 1.06 1.47 (1.93 0.75 144 96.70 54.60 1.10 0.88 e e 2.36 4.44 4.84 1.27 0.67 1.65 448.71 1 03 (p= .58) y` xa x xd xa xa xa r6 s'1 xa XC xa *Note: Blank entries indicate that the observed and expected deaths are no! reported by the authors. "Excluded from the calculation of summary SMR. bOnly refinery employees are included. cOniy male hourly employees with 1 or more years of employment are included. ax = Analysis done. " 'p < .05. SMR was 1.26. Length of employment indicated that those with 15 years of employment had an SMR of 1.62 (14 deaths, p = .07, not significant). In 1947, the recommended benzene standard [8-hour Time-Weighted Average (TWA)] in the United States was lowered from 100 ppm to 50 ppm, and in 1948, it was further lowered to 35 ppm. To determine the impact of this significant reduction in the benzene standard, analysis by hire date was done in the Chevron study. Among those hired before 1948 at the Richmond refinery, the SMR was 1.87 (13 deaths, p = .05), whereas there was no excess among those hired after 1948. Although this analysis by hire date was somewhat confounded by latency, an adequate latency (some with over 30 years) was achieved by those hired after 1948. Thus, there was some suggestive evidence that historical benzene exposure might be related to an increase in lympho- **---- -3 A iLa 'K.As-vVtal T-ftfinaru onoK/cit VlU EXX-MOR 302 Wong and Raabe TABLE Xn. Summary of Mortality From Lymphosarcoma in Cohort Studies of Refinery and Other Petroleum Employees* ' ------------ 1-------- Observed Expected - Statistical Latency Author/year Facility deaths deaths SMR significance analysis Duration analysis Tabershnw Cooper Assc. 11974]* 17 refineries Wong [1980]* 17 refineries Kaplan [1986]* 17 refineries Schonenfeld et al. 1198IT" U.S. refineries Nelson [1985] 10 refineries Wong et aJ. (1986) Richmond. El Segundo Refineries Hants et aJ. [19821* Baron Rouge Refinery Hams et al. [1985a] Baton Rouge. Baytown. Bayway Refineries Wen et aJ. [I983]c Port Arthur Refinery Morgan and Wong [1984] Beaumont Refinery Morgan and Wong [1985] PauJsboro Refinery Emeriine and Henderson 11985) Tomance Refinery McCraw ei al. [1985] Wood River Refinery Joyner[1983] Deer Park Refinery Divine et al. [1985] 13 refineries & others Divine and Barron [1987] Production & pipeline Rushton and Aldcrson [1981a] 8 UK refineries Rushton and Alderson [1983] UK distribution centers Theriault and Goulet [19791* East Montreal Refinery Theriault and Provencher [1987] East Montrea/ Refinery Christie (1986] Australia Nakamura [1987] 51 Japanese Refinenes Total 16 17 5 13 2 10 3 l 0 20 7 16 7 3 97 17.78 0.90 13.43 1.26 4.20 11.40 1.19 1.14 11.69 6.81 5.36 1.55 3.80 0.17 1.46 0.55 0.64 0.00 22.20 11.80 16.26 8.05 0.90 0.59 0.98 0.87 e 0.40 2.48 111.75 0.S7 (p =. 16) x" x^ xd Xd xd x" xa x" xd ' *Note: Blank entries indicate that the observed and expected deaths are not reported by the authors. "Excluded from the calculation of summary SMR. bOnly refinery employees are included. cOnly male hourly employees with 1 or more years of employment are included. dx = Analysis done. cp < .01. ' cause of death. The lowest SMR was observed in white male employees at the Gulf Port Arthur refinery. 1 observed vs. 10 expected. None of the studies reporting a deficit based on a reasonable number of deaths made any attempt to analyze the data by length of employment. For the industry as a group, the meta-SMR was 0.87 with a p-value of .16. Although the Chevron data were consistent with an increase, the increase was modest and was associated with historical exposure. The weight of the evidence favors a conclusion of a similar or slightly lower mortality risk from this cause for petroleum workers when compared to the general population. Leukemia (ICD8, 204-207) EXX-MOR-003379 Cancer Epidemiology in Petroleum Industry 303 TABLE XIII. Summiir; of Mortality From Leukemia in Cohort Studies of Refinery and Other Petroleum Employees* Author'vr.ar Facility Observed Expected Statistical Latency Duration deaths deaths SMR significance analysis.' analysis Tahershaw Cooper A_j.sc |1974J' Wong 1)980]* Kaplan 11986)* Schottenfeld el al [1981]"-b Nelson (19851 Wong aJ, |1986] Hariis el al 119S2]" Hahis ci a! p9B.Sn) Wen el al |I983]r Morgan and Wong {J 984} Morgan and Wong 11985} Enlerline and Henderson |1985] McCraw el al. f 1985) Joyner J1983] Divine el al. 11985) Divine and Barron |19B7) Rusbton and Aldcrson f398]a) Rushion and Aiderson 11983] Theriault and Goulet (1979}* Theriault and Pravencher f 1987} Christie 11986] Nakamura 11987} [7 refineries 17 refineries 17 refineries l!.S. refineries 10 refineries Richmond. Ei Senundn Refineries Baton Rouge Refinery Baton Rouge. Baytown.. Baywav Refineries Pori Arthur Refinery Beaumont Refinery Paulsboro Refinery Torrance Refinery Wood River Refinery Deer Park Refinery 13 refineries & others Production & pipeline 8 UK refineries UK distribution centers East Montreal Refinery East Montreal Refinery Australia 51 Japanese Refineries Total 10 32 4 4 22 9 20 33 23 10 -* M 6 48 25 30 28 5 9 279 i: 18 0.82 31 68 5.98 9 70 24,92 1.01 0.66 0.4! 0.88 9.30 0.97 25.90 0.77 25.08 13.3! 10.28 2.54 6.60 2.60 40.70 22.70 31.96 26.82 1.32 1.72 0.97 0.78 2.13 2.30 1.18 1.10 0.94 1.04 r r r r 1.93 8.5 253.54 2.59 1.06 1.10 IP = .101 x" XX* Xs X* x" xd x" xxc xa xd xd x4 xx' "'Note: Blank entries indicate thai die observed and expected deaths are not reported by the authors. "Excluded from the calculation of summary SMR. bOnly refinery employees are included. cOnly male hourly employees with 1 or more years of emplovmem arc included. dx = Analysis done. ex.x = Analysis done in subsequent report (see text). `p < .05. ' petroleum industry have also' been done. Table XIII shows the results of the cohort studies. ' ' : `! . Most of the studies reported either a similar or a higher leukemia mortality experience. Only the AMOCO study reported a significant deficit in leukemia. The number of deaths involved was only 4. This result was inconsistent with the rest of the industry. Several studies reported significant leukemia excesses. Although the original Gulf Port Arthur refinery study, [Wen et ah, 1983) found a nonsignificant leukemia excess (SMR = 1.32, Table XIII), an unpublished update of the study through the end of 1983. [Wen et ah, 3984c] showed a significant leukemia excess among employees with 20 or more years of service (SMR = 1.63, p < .01). The excess came mainly from lymphocytic and other unspecified cell types. Furthermore, an upward trend was found by length of service. According to the updated data, of the 43 leukemias at Gulf. 28 had a latency of 30 or more years. 304 Wong and Raabe fi the entire cohort (23 deaths, p < .05). Analysis by length of employment among the white males detected a significant SMR of 2.35 for those with 30 or more years of service. Similar analysis showed a significant SMR of 2.14 in white males with a latency of 20-39 years. Different cell types were observed. Elevated leukemia mortality was found at the Shell refineries at Wood River and Deer Park. At Wood River, 14 leukemias were seen when only 6.6 were expected [McCraw et al., 1985]; the SMR was 2.13 (p < .05). The excess was associated predominantly with acute myelogenous leukemia (SMR = 3.94, p < .01); however, leukemias of other cell types were also observed. A similar significant excess was also found at the Shell Deer Park refinery [Joyner, 1983]. Six leukemia deaths were identified, producing an SMR of 2.30 Ip < .05). Three of the leukemias were acute myelogenous leukemia, and the corresponding SMR was 3.60 (p < .05). However, a subsequent nested case-control study failed to identify any specific locations or jobs responsible for the excess [Austin et al., 1986]. The Australian registry, even though the follow-up was very short, detected an increase in leukemia mortality (5 observed, SMR = 2.59, p < .05). Although the industrywide refinery study [Rushton and Alderson, 1981a] in Britain found a leukemia mortality for the entire cohort similar to that for the general population, a subsequent case-control study [Rushton and Alderson, 1981b] found a more than threefold risk among those with either medium or high exposure when compared to those with low benzene exposure. Although the industry overall experienced only a small excess in' leukemia (meta-SMR = 3.10. p = . 10), there is suggestive evidence that some small groups within the industry may have had significantly elevated leukemia risk. This conclu sion is based on several considerations: the significant excesses observed for the entire refineries in several studies [Morgan and Wong, 1984; McCraw et al., 1985]; increased risks with increasing length of service [Wen et al., 1983; Morgan and Wong, 1984]; and case-control studies within the industry [Rushton and Alderson, 1981b], However, several issues remain unresolved at this time. First, high-risk groups have not been identified, and only a few studies provided job- or area-specific analyses. Most of the leukemia cases in the petroleum industry were not found among those who worked on benzene units directly [Tsai et al., 1983; McCraw et al., 1985]. Large case-control studies* with detailed employment and exposure data are needed to identify high-risk groups. Second, except for the two Shell refineries, most refinery studies reported a variety of ceil types, not limited to acute myelogenous leukemia alone. The issue of whether benzene or other petroleum hydrocarbons can affect only acute myelogenous leukemia or other cell types as well remains unresolved. Third, diagnostic specificity (or lack of) for lymphatic and hematopoietic diseases on death certificates, particularly on older death certificates, may result in inaccuracy of underlying cause of death classification. This misclassi- fication may confound site-specific analyses. Fourth, latency in these leukemia cases was of the order of 20-30 years, different from the short latency reported for benzene-related leukemias in the early literature. By comparison, benzene exposure at refineries is much lower than those reported earlier (e.g.. shoemakers in Turkey) ' - , n i -rv . j....................... a... i,,.-------- ----- -.,1,, t,N EXX-MOR-003381 Cancer Epidemiology in Petroleum Industry 305 TABLE XIV. Summary of Mortality From Other Lymphatic Tissue Cancer in Cohort Studies of Refinery and Other Petroleum Employees* Aurhoryear Facility Observed Expected Statistical Latency Duration deaths deaths. SMR significance analysis analysis [ariersh^vL Cooper () 97^ 1" Vvunc Kaplan (J986J* Schottenfeld et al (3PS)]"*' Nelson [1985] Wonp cl al ] 19R6) Haniselal | i 982.]" Hiinis ci a! U 985 a! Wen ct al 11983V Morgan and Wpng [1984] Morgan and Wong [1985] Enterline and Henderson [1985] McCraw et al. [ 1985] 3oyner|1983] Divine el al. jl9(S5J Divine and Barron [1987] Rushton and Alderson [198! a] Rushton and Alderson (1983] Theriault and Goulet [1979]* Thdriault and Provencher [ 19871 Christie |1986) Nakamura [1987] 1" refinerie> /? icfintirics 17 refineries U.S. refineries 10 refineries Richmond. El Segundo Refineries Baton Rouge Refinery Baton Rouce. Baytown. Buywa\ Refineries Pun Arthur Refinery Beaumont Refinery Pauisboro Refinery Torrance Refinery Wood River Refinery Deer Park Refinery 13 refineries &, others Production & pipeline 8 UK refineries UK distribution centers Easi Montreal Refinery East Montreal Refinery Australia 51 Japanese refineries 30 20 7 15 14 12 8 6 25 12 6 2 22. B2 ! .31 14.12 1.41 7.20 18.60 0.97 0.81 1.5.23 7.62 5.66 1.96 4.80 1.05 1.57 1.4! 1.02 1.26 21.80 13.50 1.15 0.89 3.63 1.65 1.47 1.36 x" xd Xs x" x" x4 x x x" x" Total 122 106.39 1.15 tp = .131 ` Note: Biunk entries indicate that the observed and expected deaths are not reported by the authors. "Excluded from the calculation of summary SKlR. "Only refinery employees are included. "Only male houxly employees with 1 or more years of employment are included. lIx = Analysis done. Cancer of the Other Lymphatic Tissue (ICD8, 202, 203, and 208) Table XIV shows the SMRs for cancer of other 'lymphatic tissue from the eligible cohort studies. For most studies, the SMR was slightly above the norm. These included the API industrywide study [Kaplan, 1986], the Chevron refineries [Wong et al., 1986], the Mobil refineries at Beaumont and Pauisboro [Morgan and Wong. 1984. 1985], the Shell Wood River refinery' [McCraw et al.. 1985], and the British distribution workers [Rushton and Alderson, 1983]. Only two studies showed an SMR less than 1: 0.83 for the three Exxon refineries and 0.89 for the Texaco production workers. As stated earlier, the exposure of production workers might not be similar to the refinery workers. Overall, the meta-SMR was 3.15 (p -- .13). Although the excess is not statistically significant, several points need to be addressed. In regard to the relationship between "cancer of the other lymphatic tissue" and leukemia, recent immunologic studies have shown that there are stem cells that 306 Wong and Raabe TABLE XV. Power Analysis by Cancer Site* Cancer site AH Digestive Stomach Pancreatic Lung Skin Prostsuc Kidney Brain Lymphopoietic Lymphosarcoma Leukemia Other lymphatic tissue Observed deaths 6,405 947 585 336 ! ,903 93 445 147 210 460 97 279 122 Expected deaths 7,525.29 1.174.86 659.95 354.83 2.469.60 89.23 461.35 149.24 209.70 448.71 111.75 253.54 106.39 SMR 0.85 0.81 0.89 0.95 0.77 1.04 0.96 0.98 1.00 1.03 0.87 1.10 1.15 95% Cl LL 0.83 0.76 0.82 0.85 0.74 0.84 0.87 0.83 0.87 0.94 0.71 0.97 0.95 UL 0.87 0.86 0.96 1.05 0.81 1.28 1.05 1. L5 1.15 1.13 1.06 1.23 1.38 p-value <.00001 <.00001 <.005 = .32 <.00001 = .70 = .45 = .85 = .99 = .58 = .16 = .10 = .13 MIN SMR 1.03 1.07 1.10 1.14 1.05 1.28 1.12 1.21 1.18 1.12 1.25 1.16 1.25 *95% Cl = confidence interval (upper and lower limits [LL, UL]); MIN SMR = minimum SMR detectable at .05 level with 80% power. [Minowada et al., 1981]. Further work by Bakhshi et al. [1983] demonstrates that the major clone in chronic myelocytic leukemia involves cells capable of lymphocytic, granulocytic, and erythrocytic differentiation and has led to the conclusion that a transformation event occurs at a very early multipotent stem cell level. Because this ICD group consists of several lymphatic and hematopoietic cancers, little attention has been focused on this category. Because of the diagnostic overlap, the natural course of various lymphatic cancers [Jaffe and Berard, 1978], and the multipotent stem ceil theory [Minowada et al., 1981; Bakhshi et al., 1983; Kersey, 1983], cancer of other lymphatic tissue should be combined with other lymphatic and hematopoietic cancers-for analysis [Wong, 1987; Heath, 1982; Dec- oufle et al., 1983], A thorough examination of the industry data in this direction will help to resolve this issue. - CONCLUSIONS Based on this review of the pertinent epidemiologic studies in the petroleum industry, we conclude that employees in this industry with potential exposure to petroleum products including gasoline-like hydrocarbons are not at increased risk for most cancers examined. The statistical power of the combined data from these epidemiologic studies in detecting a modest risk in most cancers is quite high (Table XV). In fact, the review indicates that these employees experienced a significantly lower mortality than the general population from the following cancers: all sites combined, digestive system, stomach, and lung. For skin cancer and brain cancer, the results from various studies were not always consistent. However, the overall data indicated that the mortality experience for these two cancers of petroleum employees was similar to that in the general population. EXX-MOR-003383 Cancer Epidemiology in Petroleum Industry 307 data indicating that certain small groups within the industry might have an elevated risk. For prostatic cancer, employees at the MEK-dewaxing units may be at higher risk, but the numbers in these studies were very small, and the results were not replicated in a subsequent study, For kidney cancer, one study indicated that gasoline truck drivers were at higher risk. Additional studies of these sites that include exposure information are needed before any relevance to hydrocarbon exposure can be determined. The review supports the conclusion that some subgroups of petroleum-employ ees. particularly those with more than 20 years of service and employed before the 1940s in the petroleum industry, may have been at increased risk of leukemia. This conclusion is based on the fact that several studies of refinery workers showed a significant excess, on the consistency of the data, on the positive trend with increasing length of employment and latency, and on the results of detailed casecontrol studies within the industry. Unfortunately, accurate historical exposure data are lacking. In most refineries, industrial hygiene-monitoring data were not been routinely collected until the mid-1970s, whereas relevant exposures in these cohort studies occurred much earlier. A risk assessment by quantitative exposure data cannot be made based on the currently available data. An industrywide case-control study of leukemia, with a strong focus on historical exposure data, would be helpful in such a risk assessment. Furthermore, there is some indication that mortality from cancer of other lymphatic tissue might also be elevated. Owing to the reasons provided earlier, this broad ICD category should be given more attention in the future. REFERENCES Aksoy M. Erdem S (1978): Follow-up study on the mortality and the development of leukemia in 44 pancviopenic patients with chronic exposure to benzene. Blood 52:285-292. Alderson MR. Ranaii NS (1980): Mortality of workers in an isopropyl alcohol plant and two MEK dewaxing plants. Br J lnd Med 37:85-89. Austin H. Cole P. McCraw DS (1986): A case-control study of leukemia at an oil refinery. J Occup Med 28:1 169-1173. ' Baird VC (1967): Effects of atmospheric contamination on cancer mortality in petroleum refinery employees. J Occup Med 9:415-420. Bakhshi A. Minowada J. Arnold A, Cossman J, Jensen JP, Wheng-Pcng J, Waldmann TA, Korsmeyer SJ (1983): Lymphoid, blast crises of chronic myelogenous leukemia . represent stages .in the development of B-cell precursors, N Enjgl 'J Med,3C)9:,826--B3i.y '. Christie D (1987): Health Watch, The Australian Petroleum Industry Health Surveillance Programme, Sixth Annual Report, 1986. Day NE (1985): Statistical considerations. In Wald NJ, and Doll R (eds.): "Interpretation of Negative Epidemiologic Evidence for Carcinogenicity " Internationa] Agency for Research on Cancer Scientific Publication No. 65, pp 13-27. DeCoufle P. Blattner WA. Blair A (1983i: Mortality among chemical workers exposed to benzene and other agents. Environ Res 30:16-25. Divine BJ, Barron V (1986): Texaco mortality study: II. Patterns of mortality among white males by specific job groups. Am J lnd Med 10:371-381. Divine BJ. Barron V (1987): Texaco mortality study: 111. A cohort study of producing and pipeline workers. Am J lnd Med 11:189-202. Divine BJ, Barron V. Kaplan SD (1985): Texaco mortality study: 1. Mortality among white male refinery'. petrochemical, and research workers. J Occup Med 27:445 --447, Enterline PE (1978): Mortality among workers from Shell Oil Company's lubricants' dewaxing uni! ai Deer Park. Texas. Report submitted to Shell Oil Company. EXX-MOR-003384 m Wong and Raabe i'BS* Environmental Protection Agency (1987): "Evaluation of the Carcinogenicity of Unleaded Gasoline." Washington DC: United States Environmental Protection Agency. Greenwald P, Friedlander BR, Lawrence CE, Hearrie T, Earle K [1981]: Diagnostic sensitivity bias--an epidemiologic explanation for an apparent brain tumor excess. J Occup Med 23:690-694. Hanis NM, Stavraky KM, Fowler JL (1979); Cancer mortality in refinery workers. J Occup Med 21:167-174. Hanis NM, Holmes TM, Shallenberger LG, Jones KE (1982): Epidemiologic study of refinery and chemical plant workers. J Occup Med 24:203-212. Hanis NM, Shallenberger LG, Donaieski DL, Sales EA (1985a): A retrospective mortality study of workers in three major U.S. refineries and chemical plants. Part 1: Comparisons with U.S. population. J Occup Med 27:283-292. Hanis NM, Shallenberger LG, Donaieski DL, Sales EA (1985b): A retrospective mortality study of workers in three major U.S. refineries and chemical plants. Part II; Internal comparisons by geographic site, occupation, and smoking history. J Occup Med 27:361-369. Harrington JM (1987): Health experience of workers in the petroleum manufacturing and distribution industry: A review of the literature. Am J Ind Med 12:475-497. Heath CW (1982): Leukemia. In Schottenfeld D, Fraumeni JF, (eds): "Cancer Epidemiology and Prevention." Philadelphia: WB Saunders Company, pp 728t-738. Hill AB (1965): The environment and health: Association or causation. Proc R Soc Med 58:295-300. Howe GR, Burch ID, Miller AB, Cook GM, Esteve J, Morrison B, Gordon P, Chambers LW, Fodor G, Winsor GM (1980): Tobacco use, occupation, coffee, various nutrients, and bladder cancer. JNCI 64:701-713. Jaffe ES, Berard CW (1978): Lymphoblastic lymphoma, a term rekindled with new precision. Ann Intern Med 89:415, 416. ': Joyner RE (1983): Letter to Dr. J. Donald Millar, Director of National Institute for Occupational Safety and Health, Dated July 28, 1983. Kaplan SD (1986): Update of a mortality studv of workers in petrochemical refineries. J Occup Med 28:514-516. *' Kersey JH (1983): Chronic myelocytic (multipotent-stem-cell) leukemia. N Engl J Med 309:851, 852. Levine RJ (1985): Letter to the editor. Environ Health Perspectives 62:465. 466. Lin RS. Kessler II (198 1): A multifactorial model for pancreatic cancer in njan: Epidemiologic evidence. JAMA 245:147-152. MacFarland HN. Ulrich CE, Holdsworth CE. Kitchen DN, Halliweil WH, Blum SC (1984): A chronic - inhalation study with unleaded gasoline vapor. J Am Coll Toxicol 3:231-248. Matanoski GM, Elliott EA (1981): Bladder cancer epidemiology. Epidemiol Rev 3:203-229. McCraw DS, Joyner RE, Cole P (1985): Excess leukemia in a refinery population. J Occup Med 27:220-222. , McLaughlin JK, Maridel JS, Blot'WJ; Schuthan LM; Mehl ES, Fraumeni JF, Jr (1984)': A .population-' based case-control study of renal cell carcinoma. JNCI 72:275-284. McLaughlin JK, Blot WJ, Mehl ES, Stewart PA, Venable FS. Fraumeni Jr JF (1985): Petroleum-related employment and renal cell cancer. J Occup Med 27:672-674. Minowada J, Koshiba H, Sagawa K, Kubonishi I, Lok MS. Tatsumi E, Han T, Srivastava BI, Qhnuma T (1981): Marker profiles of human leukemia and lymphoma cell lines. J Cancer Res Clin Oncol 101:91-100. Mommsen S, Aaguard J, Sell A (19831: An epidemiological study of bladder cancer in a predominantly rural district. Scand J Urol Nephrol 17:307-312. Morgan RW, Wong O (1984): An epidemiologic analysis of the mortality experience of Mobil Oil Corporation employees at the Beaumont, Texas, refinery. Report submitted to Mobil Oil Corpo ration. Morgan RW. Wong O (1985): Final report. An epidemiological mortality study of Mobil Oil Corporation employees at the Paulsboro, New Jersey, refinery. Report submitted to Mobil Oil Corporation. Morgan RW. Wong O (1985): A review of epidemiological studies on artificaJ sweeteners and bladder cancer. Food Chem Toxicol 23:529-533. EXX-MOR-003385 Cancer Epidemiology' in Petroleum Industry 309 Nelson NA fJ9S5): Mortalitv in Amoco Oil refinerv employees. Phase I. TSCA Section 8 (e) Report BEHQ-05 85-0557. ' ' Nelson N. Levin RJ. Albert RE. Blair AE. Griesemer RA, Landrigan PJ, Stavner LT. Swenberg JA (1986.1: Contribution of formaldehyde to respiratory cancer. Environ Health Perspect 70:23-35. Raabe GK (19831 Kidney cancer epidemiology in petroleum related studies. In Mehlman MA (edl "Rena! Effects of Petroleum Hydrocarbons: Advances in Modem Environmental Toxicology.'" Vol VII. Princeton NJ: Princeton Scientific Publishers. Inc... pp 259-271. Reeve GR. Thomas TL. Kelly VF. Waxweiler RJ, llaya S (1982): A proportionate mortality study of an oil, chemical, and atomic workers local in Texas City, Texas. Ann NY Acad Sci 381:54-61. Rushton L. Alderson MR (1981a): An epidemiological survey of eight oil refineries in Britain. Br J Ind Med 38:225-234. Rushton L. Alderson MR (1981b): A case-control study to investigate the association between exposure to benzene and deaths from leukaemia in oil refinery workers. Br J Cancer 43:77--84. Rushton L. Alderson MR (1983): Epidemiological survey of oil distribution centres in Britain. Br J Ind Med 40:330-339 Savitz DA, Moure R 11 984 ): Cancer risk among oil refinery' workers: A review of epidemiologic studies. J Occup Med 26:662-670. Schottenfeld D, Worshauer ME. Zauber AG, Meikle JG, Hart BR (1981): A prospective study of morbidity and mortality in petroleum industry employees in the United States--a preliminary report. In Peto R. Schneiderman M (eds): "Banbury Report 9: Quantification of Occupational Cancer." Cold Spring Harbor, NY: Cold Spring Harbor Laboratory, pp 247-265. Stemhagen A, Slade 1, Altman R, Bill J (19B3): Occupational risk factors and liver cancer A retrospective case-control srudv of primary liver cancer in New Jersey. Am J Epidemiol 117:443 454. ' ' Tabershaw/Cooper Associates, Inc. (1974): A mortality study of petroleum refinery workers project OH-1. Report submitted to the American Petroleum Institute. Washington, 1X3. Theriault G. Goulet L (1979): A mortality study of oil refinery workers. J Occup Med 21:367-370. Thdriaull G, Provencher S (1987): Mortality study of oil refinery workers: Five-year follow-up. J Occup Med 29:357-360. '' ' Thomas TL, DeC.oufle P. Moure-Eraso R (1980): Mortality among workers employed in petroleum refining and petrochemical plants. J Occup Med 22:97-103. Thomas TL. Waxweiler RJ. Moure-Eraso R, Itaya S, Fraumeni JF. Jr (1982a): Mortality patterns among workers in three Texas oil refineries. J Occup Med 24:135--141. Thomas TL. Waxweiler RJ. Crandall MS, White DW, Moure-Eraso R, Itaya S. Fraumemi J (1982b): Brain cancer amortc OCAW members in three Texas oil refineries. Ann NY Acad Sci 381:120 129. " . Thomas TL, Waxweiler RJ, Crandall MS, White DW, Moure-Eraso R, Fraumeni JF. Jr (1984): Cancer mortality patterns by work category in three Texas Refineries. Am J Ind Med 6:3-16. Thorpe JJ (1974): Epidemiologic survey of leukemia in persons potentially exposed to benzene. J Occup - ' Med 16:375-382. _' Tsai SP, Wen CP, Weiss NS, Wong O, McClellan WA, Gibson RL (1983): Retrospective mortality and medical surveillance studies of workers in benzene areas of refineries. J Occup Med 25:685--692. Van Peenen PFD, Blanchard AG. Wolkonsky PM (1984): Smoking habits of oil refinery employees. AJPH 74:1408, 1409. Wald NJ. Doll R (19851: Interpretation of negative epidemiologic evidence for carcinogenicity. Interna tional Agency for Research on Cancer Scientific Publication No. 65. Lyon Wen CP, Tsai SP. Gibson RL (1982): A report on brain tumors from a retrospective cohort study of refinery workers. Ann NY Acad Sci 381:130-138. Wen CP, Tsai SP, McClellan WA, Gibson RL (1983): Long-term mortality study of oil refinery' workers. I. Mortality of hourly and salaried workers. Am J Epidemiol 118:526-542. Wen CP, Tsai SP, Gibson RL, McClellan WA (1984a): Long-term mortality of oil refinery workers. II. Comparison of the experience of active, terminated and retired workers. J Occup Med 26:118 127. Wen CP, Tsai SP, Moffitt KB, Bondy M, Gibson RL (1984b): Epidemiologic studies of the role of EXX-MOR-003386 310 Wong and Raabe Wen CP, Tsai SP. Moffitt K. Bondy M, Runion H, Gibson RJL (1984c): Leukemia among oil refinery workers and residents in the surrounding communities. (With additional updated tables for leukemia.) Presented at the "Seminar on Cancer Prevention and Control in Texas" in San Antonio, Texas, February, 1984 (sponsored by the Texas Dept, of Health). Wen CP. Tsai SP, Weiss NS. Gibson RJL, Wong O, McClellan WA (1985): Long-term mortality study of oil refinery workers. IV. Exposure to the lubricating-dewaxing process. JNCI 74:11-18. Wong O (1980): An investigation of malignant neoplasms of the brain and other parts of the nervous system among a cohort of petroleum refinery workers. Report submitted to American Petroleum Institute. , Wong O (1987): An industry wide mortality study of chemical workers occupationally exposed to benzene. I General Results. Br J Ind Med 44:365-381. Wong O, Tabershaw IR (1980): Comments on mortality among workers employed in petroleum refining and petrochemical plants, i Occup Med 22:638 -640. Wong O. DeCoufie P (1982): Methodological issues involving the standardized mortality ratio and proportionate mortality ratio in occupational studies. J Occup Med 24:299-304. Wong O, Morgan RW, Kheifets L, Larson Sr (1985): Comparison of SMR, PMR and PCMR in a cohort of union members potentially exposed to diesel exhaust emissions. Br J Ind Med 42:435-448. Wong O, Morgan RW, Bailey WJ, Swencicki RE, Claxton K, Kheifets L (1986): An epidemiological study of petroleum refinery employees. Br J Ind Med 43:6-17. EXX-MOR-003387