Document VJEy4agJRY2Jp8GYj3xR73NBZ

British Journal of Industrial Medicine 1992;49:345-354 345 An updated mortality study of workers in three major United States refineries and chemical plants Leba G Shallenberger, John F Acquavella, Donna Donaleski Abstract A historical cohort study of refinery and chemical plant workers, originally observed from 1970 to 1977, was updated to the end of 1982. This cohort included 25 321 active, retired, and terminated workers at the Exxon refineries and chemical plants in Baton Rouge, Louisiana; Baytown, Texas; and Baywayl Bayonne, New Jersey. An emphasis was placed on the cancers previously reported to be increased in at least one petrochemical worker population. The population included a high proportion of retirees, providing adequate work experience and latency to evaluate risks associated with employment. Vital status was determined for 98x3% of the population and death certificates were obtained for 99-9% of the decedents. Mortality for "all causes of death" (standardised mortality ratio (SMR) = 89,95% confidence interval (95% CI) 87-91) and for "all cancers" (SMR = 94, 95% CI 89-100) were lower than expected based on state rates. History of cigarette smoking collected from individual medical records and summarised work histories were used for subgroup analyses. There was no indication that cancers of primary interest were increased; however, blue collar workers at one plant had raised rates ofkidney cancer (SMR = 246,95% CI 146390), which may require further evaluation. Across the refinery and petrochemical industry, workers have been found to experience lower mortality from all causes combined and from all cancers combined compared with rates for the general population,"~(and Morgan RW and Wong 0, personal communication). There have, however, Exxon Company USA, Health Services Department, Houston, Texas, USA L G Shallenberger Monsanto Company, Saint Louis, Missouri, USA J F Acquavella Exxon Biochemical Sciences, Inc, East Millstone, New Jersey, USA D Donaleski been inconsistent reports of increased rates for specific cancers including kidney cancer, malignant melanoma, central nervous system (CNS) and brain cancer, and cancers of the haematopoietic and lymphatic systems.9 The International Agency for Research on Cancer (IARC) recently reviewed epidemiological studies of refinery workers10 and classified the refinery environment as 2A, probably carcinogenic to human subjects. This assessment was based primarily on interpretation by IARC of the evidence that leukaemia and malignant melanoma rates are higher than expected among refinery workers. A recent academic review of studies on refinery workers9 reached a different conclusion than IARC, judging the evidence to be insufficient to conclude that the refinery environment was probably carcinogenic to workers. The 1970-7 Exxon cohort study'2 had standardised mortality ratios (SMRs) less than 100 for most causes of death except for non-significant excesses in cancers of the kidney and CNS and brain. At one of the three study locations, the Baton Rouge plant, there was an excess of pancreatic cancer. This result was further evaluated by Louisiana State University investigators studying increased pancreatic cancer rates in Louisiana. Thus far, published findings have suggested associations with lifestyle, including Cajun ancestry, moderate cigarette smoking, and consumption of pork products and rice." Raised SMRs for digestive organ and respiratory site cancers in the New Jersey plants (Bayway and Bayonne), based on United States rates, were further evaluated using state and local county rates to calculate expected numbers. The observed and expected numbers were equivalent using these rates, suggesting that the increased rates in New Jersey as compared with the total United States rates were responsible for these excesses. With the previous background, the present follow up study was planned to update mortality for this cohort. This report is particularly timely in the light of the recent IARC review and the rising public concern about potential occupational hazards. Plant descriptions The three refineries have been operating since the late 346 Shallenberger, Acquavella, Donaleski 1800s (Bayonne) or early 1900s. Early technology such as shell stills or thermal cracking coils were replaced before or during the 1940s and 1950s with modern processes such as pipestills, catalytic cracking, alkylation, and catalytic reforming. Baytown and Baton Rouge produce a wide range of motor oils and other lubricants using deasphalting, dewaxing, and extraction. Baytown, and to a lesser extent Baton Rouge, also produce a wide range of petroleum solvents and other specialty products using fractionation, extraction, and hydrofinishing. During the second world war Baytown was a major producer of high test aviation fuel and toluene for explosives (TNT), and both Baytown and Baton Rouge produced raw materials for synthetic rubber. All three refineries produce a full range of fuel products, using essentially the same type of processes. Because Baytown and Bayway have used higher sulphur and heavier crude oils longer, they began installing desulphurisation processes earlier. Baton Rouge has used coking to produce electrode grade coke since about the 1960s. Baytown did not install coking until the 1980s, and Bayway historically has sold the heavy crude fractions as residual fuel oil. The three petrochemical plants included in this study were similar in terms oftheir mode ofoperation and degree of exposure control but different in size and products. The Baton Rouge chemical plant is the largest of the three and starts with various gas and liquid cracking units and a partial oxidation unit to produce olefins and other intermediates. These are used to manufacture phthalic anhydride, alcohols, isopropanol, neoacids and aromatic compounds. The plant also includes facilities for making synthetic rubber. The Baytown chemical plant, which is smaller than Baton Rouge, emphasised the extraction of aromatic compounds (such as benzene, toluene, and xylene), the production of butyl rubber, polypropylene, cyclohexane, butadiene, isobutylene, and linear paraffins. The Bayway chemical plant near Linden, New Jersey, is the smallest of the three. At the time of the study it produced various motor oil additives, some ketones, and a low molecular weight polymer. Materials and methods COHORT DEFINITION The study population included all regular employees with at least one month service at the Baton Rouge, Louisiana, Baytown, Texas, and/or Bayway/ Bayonne, New Jersey refinery and chemical plant sites during the period 1 January 1970 to 31 December 1982. Also, employees who had retired before 1970 and were still alive at 1 January 1970 were included in the study population. DATA COLLECTION AND HANDLING Company personnel and payroll records provided demographic information and work history. Most of the death certificates for deceased workers were readily available from annuitant benefit records. Missing death certificates were traced through the Social Security Administration (SSA) and the National Death Index (NDI).'2 The underlying cause of death was coded by a trained nosologist according to the eighth revision of the International Classification of Diseases."3 VITAL STATUS FOLLOW UP Vital status was determined by a hierarchical approach beginning with information from company records. The first step in this process was to determine those active employees and annuitants who were alive or deceased according to company records. For those who were deceased, death certificates were obtained, when available, from company files. The remainder of the cohort was submitted to the SSA and the NDI. For potential decedents identified by the NDI, death certificates were requested from the respective states and verified with company information available on each study subject. The SSA search provided an indication of workers as alive (currently paying into SSA or receiving benefits), deceased, or unknown (not currently paying into SSA). Again, death certificates were requested from the respective states and compared with information from company records. Those designated by SSA as alive received the same designation in our study database. Those remaining (n = 429, 1-7%) were considered lost to follow up. DATA QUALITY CONTROL Extensive quality control procedures were employed throughout the course of the study to ensure cohort completeness, data accuracy, and consistency. These included cross checks among different record sources and data audits. Special attention was paid to the quality of the death certificate coding, which was evaluated by sending a sample ofthe death certificates to the National Centre for Health Statistics (NCHS) for review. All death certificates with cancer mentioned, plus an additional randomly selected group of about 1400 certificates, totalling 54% of all deaths, were included in the sample for review. Based on this sample, the upper 95% confidence interval (95% CI) for the error rate for coding the underlying cause of death was less than 1 %. STATISTICAL METHODS Employees who were active before 1 January 1970 contributed person-years starting from 1 February, and new employees hired after 1 January 1970 contributed person-years from their date of employment plus one month, until the end of the An updated mortality study ofworkers in three major United States refineries and chemical plants 347 study period (31 December 1982) or their date of death, whichever came first. Those who retired before 1 January 1970, contributed person-years throughout the study period or until date of death. Employees who terminated during the study period without reaching retirement age were considered at risk until the end ofthe study period unless they were designated as lost to follow up by our searches of company records, SSA, and NDI. In that case, workers who were lost to follow up contributed person years only until their date of termination from company service. Mortalities for workers were compared with those of the general population for the states in which their respective plants were located, by the calculation of SMRs. The SMR is most often expressed as the ratio of observed to expected deaths multiplied by 100. Expected deaths for workers were calculated by multiplying death rates for Louisiana, Texas, and New Jersey, specific for age, sex, race, and calendar period, by the corresponding worker person-years at plants in those specific states.'415 The SMRs were designated as statistically significant if the 95% CIs did not include 100. Rough CI calculations were used when the number of observed deaths exceeded five; Fisher's exact CIs were calculated in other instances.'6 A main occupation was derived for each employee by summarising the individual work histories and determining the occupation in which each employee spent the largest proportion of his or her time. The SMRs were calculated for blue collar workers (the occupational groups with potential for daily exposure to petroleum, petrochemicals, and other related substances) and white collar workers. The following categories were used: blue collar, consisting of process operators, mechanics, and labourers and white collar consisting of managers, professionals, technicians, office workers, service workers, laboratory workers, and field professionals. This categorisation reflects a slight change from that used in the original cohort study.2 In that study, laboratory workers were categorised as blue collar workers. Subsequent discussions with company industrial hygienists suggested that the nature of potential exposures for laboratory workers is different from that of those working on the plant units. Additional analyses were also examined to consider laboratory workers separately. Information on smoking history was abstracted from the plant medical records for the 1970-7 cohort population. Further information was not collected for employees hired during the five year update period. "Smokers" were those who ever smoked and "non-smokers" never smoked. Those with no smoking information in their records were classified as "smoking unknown". Results STUDY POPULATION DESCRIPTION The total study population of 25 321 employees and retirees contributed 232 921 person-years during the 13 year period. During the five year study update period, 3623 new employees and 95 218 person-years were added (table 1). The study population included 19 083 employees who were active for at least one month during the study period and 6238 retirees who were still alive at 1 January 1970. At the end of the observation period 39-6% (n = 10 032) ofthe total cohort were alive and still employed; another 1-7% (n = 422) died while employed. Forty seven per cent (n = 11 973) had retired before or during the study period (fig 1). Employees who had terminated employment before reaching retirement age accounted for 11-4% (n = 2895) of the study population. Only 1-7% (n = 429) of the total population were lost to follow up (table 1). Ninety four per cent of this group were less than 40 years old at last observation. Seventy two per cent had worked less than five years; 92-3% had worked less than 10 years. The total study population comprised 82% white men, 10% black men, 8% white women, and 1% black women. The study population had a bimodal age distribution with peaks in the 30-39 (24%) and 70 or older (28%) age groups at last observation. Roughly 48% of the population had worked for Exxon for 25 years or more. Twenty six per cent were hired before 1940, and 24-5% began employment between 1940 and 1949. Twenty per cent were hired Table 1 Vital status: total study population 1970-82 by plant site Baton Rouge No (%) Baytown No (%) Bayway/Bayonne No (%) Alive Dead Lost to follow up Total workerst Person-years 7709 (77 9) 2000 (20 2) 185 (1 9) 9894 93785 7213 (82-7) 1374 (15-8) 135 (1-5) 8722 80684 4924 (71-8) 1826 (26-6) 110 (1-6) 6860 59257 *One death occurred in a person who worked at two study locations; he was counted once in the total. tIncludes 150 employees who worked at more than one of the three plants. 1970-82 totals No (%) 19693 (77 8) 5199* (20 5) 429 (1-7) 25321 232921 348 Shallenberger, Acquavella, Donaleski Terminated 2895 (11.4%) Died while employed 421 (1 7%) Total alive 19 693 (77 8%) Figure 1 Total study population 1970-82. Total dead 5199 (20 5%) Total lost to follow up 429 (1 7%) during the original eight year observation period and 14% of the cohort were hired during the five year update period. Over the 1970-82 study period, there was a total of 5199 deaths, of which 1995 occurred during the update period. The 5199 deaths were evenly distributed by year from 1970-82. Seventy three per cent of all the deaths were pre-1970 retirees and 18% were persons who retired during the study period. Death certificates were available for all but five (0- 1 %) of the known deaths. These five deaths were included in the SMRs for all causes combined but were not apportioned to specific cause of death categories. Blue collar workers accounted for around two thirds of the workforce at each of the three plants. One per cent of the workers at Baton Rouge and Bayway/Bayonne and 4% at Baytown had missing work histories. Smoking state was determined for 70% of the 1970-7 study population, from medical record data (fig 2). No additional data were collected for those who entered the cohort after 1977. For the total 1970-82 population, among those with known smoking histories, 44% were or had been smokers includ- ing 32% smokers among white collar workers and 50% among blue collar workers. MORTALITY ANALYSES: TOTAL STUDY POPULATION Table 2 presents the SMRs for the total population and three plant populations. The SMRs cannot be directly compared with each other; however, general trends can be noted. The SMRs for all causes of death were significantly low for Baton Rouge (SMR = 86, 95% CI 82-90), Baytown (SMR = 85, 95% CI 81-90), and the total population (SMR = 89, 95% CI 87-91); whereas for Bayway/ Bayonne the SMR was slightly lower than expected (SMR = 97, 95% CI 93-102). The SMRs for most major causes of death were lower than expected. Significant deficits were found in the SMRs for the following: diseases of the respiratory system and for all external causes at all three plant sites; diseases of the digestive system, and for ill defined conditions at Baton Rouge, Baytown, and for the total population. Ischaemic heart disease SMRs were around 100 for all three plant sites and for the total population. The SMRs for all cancers were significantly low for Baton Rouge (SMR = 91, 95% CI 83-100), Bay- An updated mortality study of workers in three major United States refineries and chemical plants 349 Unknown 1 Non-smokers Smokers a significant deficit for the SMR at Bayway/Bayonne (SMR = 55,95% CI 27-98). Mortality from kidney cancer was significantly high at Baton Rouge (SMR = 192, 95% CI 113-303), but was similar to expected values at the other two plant sites. Mortality from brain cancer was increased at Baytown (SMR = 156, 95% CI 80-272), and Baton Rouge (SMR = 121, 95% CI 55-231) and lower than expected at Bayway/Bayonne (SMR = 67, 95% CI 18-171). For oesophageal cancer, the SMRs were moderately increased at all three plant sites. The number of deaths from all haematopoietic and lymphatic system cancers was slightly increased at Baton Rouge (SMR = 125,95% CI 93-165), and similar to the expected number at the other two plant sites. Total Baton Rouge Bayton (21 743) (8695) (7250) Bayway (5798) Figure 2 Refinery/chemical plant cohort smoking data from medical records by site 1970-7. Number in parentheses- total site population. town (SMR = 87, 95% CI 78-97), and for the total population (SMR = 94, 95% CI 89-100). For Bayway/Bayonne, the observed number of cancer deaths was consistent with the expected number (SMR = 103, 95% CI 93-114). Significant deficits were found in the SMRs for lung cancer at Baton Rouge, Baytown, and for the total population. The SMR for pancreatic cancer was increased at Baton Rouge (SMR = 139,95% CI 98-191), but there was WORK HISTORY AND SMOKING ANALYSIS Tables 3 and 4 show the SMRs for white collar and blue collar workers categorised as smokers, nonsmokers, and unknown smokers. In these and subsequent tables, the analyses focused on cancer sites that showed increases in previous petrochemical studies.9 Most of the SMRs were less than 100. For blue and white collar workers, the SMRs were higher for smokers compared with non-smokers for most causes of death. No significant excesses occurred in the white collar group (table 4). White collar nonsmokers had significant deficits for all causes (SMR = 48, 95% CI 39-59), ischaemic heart disease (SMR = 59, 95% CI 43-80), and all cancers (SMR = 47, 95% CI 30-71). Blue collar nonsmokers also had significant deficits for all causes (SMR = 51, 95% CI 45-57), ischaemic heart disease (SMR = 70, 95% CI 59-82), and all cancers (SMR = 46, 95% CI 35-59). Blue collar workers with smoking unknown had a significantly raised SMR for ischaemi_ heart disease. Blue collar smok- Table 2 SMRs for selected causes of death for the three plant populations and total study population at risk, 1970-82 Baton Rouge Baytown Bayway/Bayonne Total population Cause of death (ICD-8) Obs SMR Obs SMR Obs SMR Obs SMR All causes of death (001-998) 2000 86* 1374 85* 1826 97 Ischaemic heart disease (410-413) 859 100 545 102 856 104 Disease of respiratory system (460-519) 102 66* 93 77* 90 72* Allcancers(140-209) 448 91* 308 87* 402 103 Oesophagus (150) 12 117 9 125 11 126 Stomach (151) 15 75 12 93 26 125 Large intestine (153) 39 104 23 82 51 107 Pancreas (157) 38 139 17 86 11 55* Lung (162) 152 85* 98 76* 116 103 Skin-melanoma (172) 5 80 5 74 4 74 Kidney (189) 18 192* 8 97 9 114 Brain/CNS (191-192) 9 121 12 156 4 67 All haematopoietic and lymphatic (200-209) 50 125 32 103 30 99 Leukaemia(204-207) 23 129 15 114 9 68 Lympho/reticulosarcoma (200) 9 156 6 117 6 111 Other lymph (202-203. 208-209) 17 122 9 84 14 141 5199t 2260 285 1158 32 53 113 66 366 14 35 25 111 46 21 40 89* 102 71* 94* 122 99 100 98 87* 76 137 119 110 104 129 116 *Statistically significant based on 95% CIs. tOne death occurred in a person who worked at two study locations; he was counted once in the total popuiation caiculations. 350 Shallenberger, Acquavella, Donaleski Table 3 SMRs for selected causes of deathst for white collar workers by smoking statet, 1970-82 Total Smokers Non-smokers Smoking unknown Obs SMR Obs SMR Obs SMR Obs SMR All causes Ischaemic heart disease All cancers Pancreas Kidney Skin-melanoma CNS/brain All haematopoietic and lymphatic Leukaemia Lympho/reticulosarcoma Other lymph Benign neoplasms 937 376 217 10 5 2 7 28 15 4 7 4 79* 505 85* 200 84* 125 71 6 89 3 46 2 131 5 127 13 159 5 119 3 95 4 152 1 90* 98 94 42 95 23 84 1 101 0 93 0 177 1 120 4 111 3 160 1 107 0 75 0 48* 334 59* 134 47* 69 38 3 02 00 82 1 95 11 174 7 138 0 03 03 81* 86* 89 72 132 0 78 158 221 0 138 375 *Statistically significant based on 95% CIS. tSelected causes of death in this table are consistent with those emphasised in other refinery and petrochemical worker studies. tNot collected for 1978-82 new employees. Table 4 SMRs for selected causes of deatht for blue collar workers by smoking state, 1970-82 Total Smokers Non-smokers Smoking unknown Obs SMR Obs SMR Obs SMR Obs SMR All causes 4121 91* 2101 91* 302 Ischaemic heart disease 1783 102 910 105 147 All cancers 918 95 544 104 61 Pancreas 56 107 29 102 10 Kidney 29 149 21 193* 2 Skin-melanoma 12 87 7 96 2 CNS/brain 18 117 10 107 6 All haematopoietic and lymphatic 82 106 47 113 5 Leukaemia 30 88 19 108 2 Lympho/reticulosarcoma 17 138 8 120 2 Other lymph 32 120 19 129 1 Benign neoplasms 11 117 6 114 1 51* 1718 104 70* 768 116* 46* 313 102 137 17 102 70 6 104 98 3 69 232 2 57 46 30 119 43 9 77 114 7 177 26 12 147 78 4 140 *Statistically significant based on 95% CIs. tSelected causes of death in this table are consistent with those emphasised in other refinery and petrochemical worker studies. tNot collected for 1978-82 new employees. Table 5 SMRsfor selected cancer deathsfor Baton Rouge workers by job category and smoking statef, 1970-82 Cancer site Blue collar White collar Total Smoker Non-smoker Unknown Total Smoker Non-smoker Unknown Obs SMR Obs SMR Obs SMR Obs SMR Obs SMR Obs SMR Obs SMR Obs SMR Kidney 18 246* 14 257* 2 Pancreas 32 146 22 140 5 All lymphatic and 41 130 33 149* 3 haematopoietic Leukaemia 18 128 15 154 2 Lympho/reticulosarcoma 7 157 6 188 0 Otherlymph 12 116 10 132 1 169 2 283 0 0 0 0 0 144 5 185 6 113 4 116 0 60 5 120 7 86 5 99 1 89 1 47 3 84 1 46 1 0 1 183 2 158 2 249 0 60 1 88 0 0 0 0 0 00 0 0 2 277 58 1 75 136 1 155 00 0 00 0 *Statistically significant based on 95% CIs. tNot collected for 1978-82 new employees. ers had a significantly raised SMR for kidney cancer (SMR = 193,95% CI 119-295), due in large part to findings from the Baton Rouge plant. PLANT-SPECIFIC MORTALITY ANALYSES Baton Rouge Table 5 shows selected SMRs for men at the Baton Rouge refinery and chemical plant stratified by potential exposure group and smoking. All the deaths from kidney cancer were in blue collar workers, resulting in significantly increased SMRs for all blue collar workers and for known smokers. The SMRs for non-smokers and those with un- An updated mortality study of workers in three major United States refineries and chemical plants 351 Table 6 SMRs for kidney cancer, pancreatic cancer, and leukaemia deaths in Baton Rouge blue collar men, 1970-82 cohort By year employed Before 1940 to 1950 or 1940 1949 later Kidney cancer Pancreatic cancer Leukaemia Obs SMR Obs SMR Obs SMR 6 180 20 180* 4 50 10 260* 12 120 13 210* 2 1000* 0 0 1 130 By age first employed < 20 20 to 29 30 years years years 16 200 150 1 14 80 130 1 10 140 140 11 380* 17 180 7 110 By latency < 20 20 to 29 years years 05 0 420* 00 00 10 170 0 30 years 13 220* 21 170* 17 140 By duration employed < 20 20 to 30 years years years 28 250 330* 27 80 100 28 100 180 8 200 23 190* 8 90 known smoking histories were also raised based on two observed deaths in each subgroup. The SMRs for pancreatic cancer were raised for all blue collar subgroups and for white collar smokers and un- knowns. The SMR for all haematopoietic and lymphatic cancers was significantly raised for blue collar smokers, and not raised in any of the white collar subgroups. Likewise, there were more deaths from leukaemia and lymphoma than expected in blue collar smokers. Table 6 further evaluates the kidney and pancreatic cancer and leukaemia SMRs in the Baton Rouge blue collar men. The SMRs for kidney cancer were increased for all employment periods and significantly increased for those hired during or after 1940. Similarly, all age at hire groupings had raised SMRs with a significant increase for those hired at 30 or more years of age. Grouping the subjects by latency showed SMRs that were significantly increased after 20 years from initial employment. Analyses by duration of employment showed increased mortality for all durations, with a significant increase for those employed for 20 to 29 years. For pancreatic cancers, the SMR was significantly increased in men who were hired before 1940 (SMR = 180, 95% CI 110-278), and slightly increased for those hired in the 1940s (SMR = 120, 95% CI 62-210). There were more deaths than expected for those hired at age 20 or more, particularly among those more than 30 years old at hire Table 7 Frequency of leukaemia cell types by job category for Baton Rouge male employees 1970-82 Type of leukaemia (ICD-8) Baton Rouge Blue White collar collar Unknown Totals Lymphatic (204-0,204 1,204 9) 81 1 Acute myeloid (205-0) 41 1 Chronic myeloid (205 1) 21 - Myeloid unspecified (205 9) - - - Acute monocytic (206 0) - - Other and unspecified (207-0,207-1,207 9) 4-- Totals 18 3 2 10 6 3 - 4 23 (SMR = 180, 95% CI 105-288). All of the deaths for blue collar workers were in subjects with at least 30 years latency. Most of the pancreatic deaths also occurred in workers with at least 30 years of employment, which resulted in a significantly increased SMR (SMR = 190, 95% CI 120-285). For the deaths from leukaemia, there was a significant increase only in men hired during the 1940s (SMR = 210,95% CI 112-359). There was no trend by age at hire. Most of the deaths from leukaemia occurred in men with at least 30 years latency (n = 17) and 20 or more years duration of employment (n = 16), but the corresponding SMRs were not significant. Table 7 lists the ICD-8 classification for the 23 leukaemia deaths in Baton Rouge. Six of the 23 leukaemia deaths were classified as acute myelogenous leukaemia (AML). Focusing on the subgroup of blue collar workers who were hired in the 1940s, two of the 13 leukaemia deaths were AMLs. Baytown Attention at the Baytown refinery and chemical plant focused on the one cancer found to be somewhat increased for the total plant populationnamely, brain/CNS cancer. Also, the category of benign neoplasms was reviewed as many of these tumours involved the CNS. For Baytown men nonsignificant increases in brain/CNS cancer and benign neoplasm SMRs occurred in blue and white collar smokers and non-smokers, but these were based on small numbers. Mortality from benign neoplasms was increased in white collar more than in blue collar workers. Due to the small number of brain cancers in Baytown blue collar men (n = 10), the SMRs by latency and duration are somewhat unstable. The SMRs were raised for all three employment year groups, but somewhat more so for those employed during 1940-9 (SMR = 210, 95% CI 77-457). Seven of the 10 cases were 20-29 years old when hired and three were 30 or more years old. All the latency groups had increased SMRs, but most of the cases had 30 or more years of latency (n = 6). Five cases worked for 20-29 years, and four had a duration of 30 or more years. Bayway/Bayonne The SMR for ischaemic heart 352 Shallenberger, Acquavella, Donaleski disease was significantly high in Bayway/Bayonne at each ofthe three plant locations. Eighteen ofthe 22 blue collar workers reflecting a 9% (95% CI 107-117) cases were in the subpopulation of operators, increase over state rates, and was significantly low in mechanics, and labourers, giving a directly adjusted white collar workers. SMRs were not significantly mortality of 1 6 per 10 000 person-years compared increased for any of the smoking subgroups, but with the United States rate of 1 - L2 The increase was more than half of the deaths had an unknown not statistically significant. With the additional five smoking history. Further examination of SMRs for years ofdata, the excess appears to have become more blue collar workers showed that most of the deaths prominent at Baton Rouge, exclusively among blue were in men who were hired before 1940 and that the collar workers, and to have diminished at Bayway/ SMR was significantly increased for men who began Bayonne and Baytown. The only other study with an employment at 30 or more years ofage (SMR = 114, excess ofkidney cancer was a United Kingdom study 95% CI 104-125). An overwhelming majority had 30 of distribution workers, which showed a slight kid- or more years latency (n = 805), whereas those with ney cancer excess concentrated among lorry 20-29 years latency had a significantly low SMR drivers.'8 (SMR = 67, 95% CI 44-97). The SMRs were Mortality from kidney cancer by smoking state for slightly higher in those who had worked for less than blue collar workers showed increases in smokers, 30 years. non-smokers, and those with unknown smoking Laboratory workers In one analysis, SMRs for histories, but was most pronounced among known laboratory workers were considered separately from smokers. Fourteen of the 18 deaths were known white collar workers. There were very few deaths in smokers and two had smoking unknown. As none of this group (n = 49) and no significant mortality the deaths from kidney cancer were in the white excesses. collar workers, it was not possible to separate the effects of smoking from the effects of potential work- Discussion place exposure. Evaluation of death rates for our study population Because smoking is a risk factor for kidney cancer showed these to be lower than expected based on in the general population, proper control for smoking state mortality for all causes of death, all cancers, and would involve calculating SMRs for Exxon smokers respiratory diseases. These results are consistent and non-smokers v death rates from kidney cancer with findings from other historical cohort studies of for the respective general population groups. As refinery and petrochemical workers. ' The inclusion smoking specific rates are not available for the general of a large group of retirees and older employees in population, however, comparisons with rates for the this cohort differs somewhat from other petro- general population include a mixture of smokers and chemical populations which have tended to include a non-smokers in the comparison group. Therefore, high percentage of younger active and terminated SMRs for Exxon smokers are overestimated and workers. Accordingly, a strength of this study is that SMRs for non-smokers are underestimated for any it allows for the detailed examination of mortality disease that is related to smoking. Even though the among long term workers with extremely long SMRs in blue collar non-smokers and unknown latency periods. One limitation is that data were not smokers are based on extremely small numbers, the available for retired workers who died before 1 fact that the SMRs are increased in all three sub- January 1970. Their inclusion may have contributed groups argues against smoking as an explanation for additional information, but we do not believe that the the raised SMR for kidney cancer. The smoking lack of this information compromises the validity of analyses are further limited by the crudeness of the the data that were collected and analysed. smoking data, specifically the lack of information It is difficult to interpret differences in mortality regarding the amount smoked, the type of cigarettes, among the three refinery petrochemical complexes or changes in smoking habits over time. based on what we know about the histories of the The raised SMRs for kidney cancer in men hired plants. The significant increase in kidney cancer in during the 1940s and in men 30 or more years old at the Baton Rouge population is of interest. Concern the time of hire, suggests that there could have been about kidney cancer among refinery populations unique service or other occupational influences dur- stemmed from the results ofa toxicological study that ing the war years. We have no information to evaluate exposed rats to wholly vapourised unleaded this further. petroleum. The study found a dose related incidence The detailed individual work histories for the of kidney tumours among male rats, but not among Baton Rouge kidney cancer cases were reviewed by female rats.'7 Epidemiological studies of kidney local refinery and chemical plant management. It was cancer among refinery workers have been largely difficult to determine exact work locations for many of negative. One exception was the Exxon 1970-7 the cases, especially the mechanics who often worked cohort study of refinery and chemical plant workers. in many areas of the plants. It was also impossible to In this study, rates for kidney cancer were increased determine what their specific occupational exposures An updated mortality study of workers in three major United States refineries and chemical plants 353 may have been, preventing firm conclusions concern- Pancreatic cancer was increased more so in blue ing occupational exposures and the kidney cancer collar workers in Baton Rouge than in white collar findings. The American Petroleum Institute (API) workers. Increases in pancreatic cancer have not been recently completed a case-control study of kidney seen in other studies on petrochemical workers.'8 cancers in five major petroleum companies and This finding may reflect non-occupational risk included many of Exxon's cases. An exposure matrix factors." was utilised to evaluate the individual work histories Cancers of the CNS/brain in Baytown occurred and exposures. The results of this multi- slightly more in blue collar workers than in white company study indicate that all the comparisons collar workers, but none of the SMRs were statis- between kidney cancer and refinery exposures tically significant. Some of the deaths classified as showed either no association or weak, non-significant benign neoplasms based on nosological rules may associations (Poole C, Satterfield MH, Levin L, have actually been due to malignant brain cancers; Rothman KJ, Dreyer NA, personal communication). however, this would also be true for the deaths in the It can be seen in tables 3 and 4 that the SMR for Texas general population used for comparison. leukaemia is raised for all white collar workers and Several studies of refinery and petrochemical work- less than expected for blue collar workers. The non- ers, including the 1970-7 Exxon cohort study, have significant excess of leukaemia (23 observed, 18 shown increased SMRs for cancer of the CNS/brain. expected) in Baton Rouge, with most ofthe deaths in Most ofthe differences in these studies were small and the blue collar/smoker subgroup, is the reverse not statistically significant.'9 In the original Exxon situation and suggests that a different set of circum- study, the overall SMR was 115; but in an internal stances may have been present in the Baton Rouge comparison of potentially exposed (blue collar) with population. Of the 23 total leukaemia deaths, 26% non-exposed (white collar) workers, the non-exposed (n = 6) were classified as AMLs. In the Surveillance CNS cancer mortality was higher. In one study of Epidemiology End Results (SEER) data for all age refinery workers, there was a 26% excess of CNS groups for the period 1973-7,'9 31% of 2077 deaths cancer but no consistent trend with duration of from leukaemia were due to AMLs. Applying that employment.5 In a study of Canadian refinery work- proportion to the expected number of deaths from ers, the SMR was 390 and statistically significant, but leukaemia gave 5-6 AMLs expected v six observed, it was based on only three cancer deaths. No common and provides no indication of an AML excess in this exposure could be identified.8 Delzell et al concluded population. This indirect method for estimating the that there is not sufficient evidence for an increased expected number of AMLs was used because state risk of CNS/brain cancer in refinery workers.9 death rates for AMLs were included with other types Researchers at the University of Texas and of leukaemia. This method should somewhat Louisiana State University have been conducting a underestimate the expected number of AMLs as hospital based case-control study of brain/CNS SEER data included all ages and AMLs are more cancers in the Gulf Coast area and Louisiana. Results common in older age groups as in our study popula- from that study are expected in 1992. tion. The significantly increased SMR for ischaemic Whereas it is clear that high level exposure to heart disease in blue collar workers in the New Jersey benzene contributes to a significant excess risk of plants was examined further. Almost 400 of these leukaemia,20 specifically AML, a relation has not been deaths were in white men who had lived to the age of confirmed at the low benzene exposure typical of 85 or longer. The expected numbers of deaths for the refinery environments. Most previous petrochemical age intervals 85-89, 90-94, and 95 or older may be studies did not find evidence of increased rates for underestimated as they are calculated using the death these cancers.9 Excesses for leukaemia were reported rates for those aged 85 or older, which is heavily in two studies on refinery workers92' (and Morgan RV weighted by the younger persons in that interval. and Wong 0, personal communication), but follow Therefore, the increased SMR may be due, in part, to up analyses using available information on an underestimated expected number of deaths in the occupational exposure have been unable to identify oldest age groups. specific causal agents. Further analyses of the leuk- All the analyses based on main occupation omitted aemia mortality by cell type in one of these studies workers who had missing work histories. This showed an excess of lymphocytic leukaemia among included 1% of the Baton Rouge and Bayway/ deaths occurring before 1970 and an excess of AML Bayonne populations and 4% of the Baytown in employees hired in 1940 or later and dying population. A further mortality analysis was per- between 1970 and 1984. It was suggested, but not formed for this small group. Most ofthe cause specific confirmed, that this time restriction ofthe myelocytic SMRs were increased, based on very small numbers. leukaemia may have been related to increased use of The two SMRs which were significantly high were benzene in the 1940s and later. Direct correlations for leukaemia in Baton Rouge, based on two deaths with known exposures were not made.22 (SMR = 1333, 95% CI 161-4816), and for all causes 354 Shallenberger, Acquavella, Donaleski in Baytown, based on 88 deaths (SMR = 133, 95% geographic site, occupation, and smoking history, J Occup CI 106-163). We have no additional information to evaluate this group further. Med 1985;27:361-9. 3 Wen CP, Tsai SP, McClellan WA, Gibson RL. Long-term mortality study of oil refinery workers, I. Mortality of hourly and salaried workers. Am J Epidemiol 1983;1 18:526-42. Conclusions 4 Wen CP, Tsai SP, Gibson RL, McClellan WA. Long-term mortality study of oil refinery workers, II. Comparison of the Overall, the results of this study show lower than expected mortality for workers at all three plants and experience of active, terminated, and retired workers. J Occup Med 1984;26:118-27. 5 Wong 0, Morgan RW, Bailey W, et al. An epidemiologic study no indication of significant mortality excesses for of petroleum refinery employees. Br J Ind Med 1986;43:6-17. specific causes ofdeath across plants. The last findings contrast with the recent IARC review and are 6 Divine BJ, Barron V. Texaco mortality study: II. Patterns of mortality among white males by specific job groups, Am J Ind Med 1986;10:371-81. consistent with the previously referenced review of refinery worker studies.9 Among the cause specific 7 Nelson NA, Van Peenen PFD, Blanchard AG. Mortality in a recent oil refinery cohort. J Occup Med 1987;29:610-7. 8 Theriault G, Provensher S. Mortality study of oil refinery mortality findings at individual plants, the excess of kidney cancer in blue collar workers at Baton Rouge workers: five year follow-up. J Occup Med 1987;29:357-61. 9 Delzell E, Austin H, Cole P. Epidemiologic studies of the petroleum industry. Occupational Medicine: State of the Art would require a more detailed evaluation to assess Reviews, 1988;3:455-74. more specifically the potential role of occupational exposures and smoking. A follow up case-control 10 World Health Organisation, International Agency for Research on Cancer. Monographs on the evaluation ofcarcinogenic risks to humans: occupational exposures in petroleum refining; crude oil study of kidney cancer at Baton Rouge is planned. The study is expected to go beyond the API case- and major petroleum fuels. Lyon: WHO, IARC, 1989. 11 Falk RT, Pickle LW, Fontham ET, Correa P, Fraumeni JF. Life-style risk factors for pancreatic cancer in Louisiana: a control analysis and will use additional sources for exposure assessment and a toxicology review to case-control study. Am J Epidemiol 1988;128:324-6. 12 Acquavella JF, Donaleski D, Hanis NM. An analysis of mor- tality follow-up through the national death index for a cohort identify potential aetiological agents. The develop- of refinery and petrochemical workers. Am J Ind Med 1986; ment of New Jersey state death rates for the groups aged 85 or older could be studied further, for 9:181-7. 13 Manual ofthe International Classification ofDiseases, Injuries and Causes of Death, 8th ed. Geneva: World Health Organisation, comparison with the large group of very old retirees in the Bayway/Bayonne population. 1967. 14 Marsh G. State mortality rates. University of Pittsburgh, Mor- tality and Population Data System. Pittsburgh: University of Pittsburgh (Sci) 1987. We acknowledge the help of our many coworkers at Exxon Biomedical Sciences Inc and the Exxon 15 Monson RR. Analysis of relative survival and proportional mortality. Comput Biomed Res 1974;7:325-32. 16 Rothman KJ, Boice JD. Epidemiologic analyses with a program- Company, USA Headquarters Medical Staff. Eugene Sales helped us with vital status follow up and mable calculator. Bethesda: National Institute of Health, 1979:(Publ No 79-1649). 17 MacFarland HN et al, eds. Chronic gasoline toxicology. Proceed- nosology, Gail Jorgensen provided database management and analytic support. Micki Vodarsik man- ings ofthe symposium. The toxicology ofpetroleum hydrocarbons. American Petroleum Institute 1982:78-86. 18 Wong 0, Raabe GK. Critical review of cancer epidemiology in aged the data coding, data entry, and summarised the petroleum industry employees, with a quantitative meta- work histories. Jerry Lynch, CIH, and Bruce Simpson, CIH, provided industrial hygiene support. We analysis by cancer site. Am J Ind Med 1989;15:283-310. 19 US Department of Health and Human Services, Surveillance, epidemiology, and end results: incidence and mortality data, are grateful to Jeffrey Wacaser and Paul Theall for typing our many drafts. 1973-77, National Cancer Institute Monograph 57, Bethesda: National Institute of Health, 1981:78-9. 20 Rinsky RA, Smith AB, Hornung R, et al. Benzene and leukemia: an epidemiologic risk assessment. N Engl J Med 1987; 1 Hanis NM, Shallenberger LG, Donaleski DL, Sales EA. A retrospective mortality study of workers in three major US refineries and chemical plants, part I: comparisons with US population, J Occup Med 1985;27:283-92. 2 Hanis NM, Shallenberger LG, Donaleski DL, Sales EA. A retrospective mortality study of workers in three major US 316:1044-50. 21 McCraw DS, Joyner RE, Cole P. Excess leukemia in a refinery population. J Occup Med 1985;27:220-2. 22 Wongsrichanalai C, Delzell E, Cole P. Mortality from leukemia and other diseases among workers at a petroleum refinery. J Occup Med 1989;31:106-1 1. refineries and chemical plants, part II: internal comparisons by Accepted 2 September 1991