Document npQ5VYMmYYKNMXvjk2pMn1EOm

American Journal of Industrial Medicine 10:371-381 (1986) r I \ Texaco Mortality Study: II. Patterns of Mortality Among White Males by Specific Job Groups . Barbara J. Divine, PhD, and Virginia Barron, MS While an earlier report on the Texaco Mortality Study cohort showed no statistically significant elevations for any cause of death for the white males, it did not preclude any excess risk of mortality within subgroups of workers. In this study, an employee's completejob history was used to determinehis work categories, and patterns of mortality were examined for the more commonjob categories. All of thejob categoriesexamined showed deficits for mortality overall, and the patterns seen were similar to those for the entire cohort. Significant elevations were seen in pancreas cancer mortality for office and managerial people and in leukemia mortality for pipefitters and boilermakers. Other elevations of particular interest based on five or more deaths were brain cancer for laboratory workers and benign neoplasms in pipefitters and boilermakers. These associations were examined by latency and years worked, and no consistently positive associations were seen. It was not possible to take into account calendar time of exposure in this type of analysis, nor could any specificchemicalsor levels of exposure be associated with the job categories where the standardized mortality ratios were elevated. Key words: environmental carcinogens, longitudinal studies, mortality, neoplasms, occupational diseases - -- INTRODUCTION The Texaco Mortality Study (TMS) is a cohort mortality study of refinery, petrochemical, and research workers who were employed at least five years and who worked between 1947 and 1977. The first paper on the TMS cohort discussed the overall patterns of mortality among the white males [Divine et al, 19851, The major finding of the study was a significant deficit for all the major causes of death. No ! statistically significant elevations were seen for any specific cause of death. However, small elevations were seen for pancreatic cancer, brain cancer, leukemia, other lymphatic cancer, and benign neoplasms. The question remained whether an increased . risk of mortality in a subgroup of workers was obscured in the analysis of the cohort as a whole. This paper examines the mortality patterns for white men employed in specific jobs in order to determine if any workplace locations or tasks are associated with an increased risk of mortality. The above-mentioned causes with elevated Texaco Inc.. Houston, Texas Dr. Barron's current address is EDS Corp.. 7171 Forest Lane, One Forest Plaza, Suite 210. Dallas. TX 75230. Address reprint requests to Dr. Barbara Divine, Texaco Inc.. P.O. Box 1404, Houston, TX 77098. Accepted for publication April 2, 1986. 0 1986 Alan K. L i s , Inc. 372 Divine and Barron standardized mortality ratios (SMRs) are examined, in particular, in order to determine if they are associated with any particular job group. In this study, a worker's job history at participating locations was used to group him into one or more subcohorts for analysis. Although numerous mortality studies of petroleum workers have been carried out, none have examined the complete work histories of the entire cohort. Work histories have been used mainly in case-control studies within a cohort [ H a i s et ai, 19821 or to identify subcohorts exposed to a specific chemical or process unit [Tsai et ai, 1983; Wen et al, 19851. METHODS The cohort consisted of all persons employed five years or more at a Texaco refinery, petrochemical plant, or research laboratory and who worked at least one day between January 1, 1947, and December 31, 1977. Follow-up through the study end date was done using company records and information from the Social Security Administration. Death certificates were requested for all known deaths. Further details are in the previous paper on this cohort [Divine et al, 19851. In addition to the demographic data, available work history data were also used. Each job held by an employee consisted of a unit (eg, crude still, maintenance), a task (eg, operator, pipefitter), and the dates that position was held. The task and unit codes were based on the American Petroleum Institute (APIJ Job Code Classification System of Petroleum Refineries and Selected PetrochemicalOperations [1979]. Job history coding accounted for all periods of time between a subject's initial hire date and final termination date, or the common closing date of December 31, 1977. Periods of not working for one month or longer were coded as such. Because much of the cohort experienced numerous job changes over relatively short periods of time, the following method was developed for selecting jobs to be coded on each work history. An employee's initial job was coded, Job changes which occurred at least 6 months subsequent to the last coded job were also coded. Job changes which occurred within the 6 months were ignored. This method continued until the employee transferred, terminated, died, or until the study end date was reached. When two or more jobs were held concurrently, they were all coded with the time apportioned equally. If either of two conditions existed, a job change was also coded: transfer to another facility, or employment in a task or unit of special interest. For a quality control check of the job history coding, a systematic sample of employee work histories was selected for duplicate coding by the original coder and for recoding by a randomly selected alternate coder. Error rates were less than 10%. Specific jobs (units, tasks, and combinations thereof) to be analyzed were selected in consultation with company industrial hygienists. In addition, certain tasks and units with common exposures or responsibilities were grouped together for analysis. The job subcohorts were defined as those ever employed, those employed one year or more, and those employed 5 years or more in the job.The results from four major job groups (office, maintenance, operations, and laboratory) are presented. Although many other job groups were analyzed. only results for the job category of pipefitters and boilermakers are presented because two of the SMRs were significantly elevated. These causes of deaths, plus two others in two of the major job groups, are also analyzed in more detail. Mortality Patterns by Specific Job Groups 373 c Person-years were counted after the date at which a person had worked the requisite len,@ of time in the job, met the eligibility requirements for the study, or the study beginning date-whichever came last. For analysis by duration of employment, the years employed in the specific job were used. Expected numbers of deaths based on the U.S. white male population of similar age and calendar time were calculated using the Monson program [Monson, 19741. This program was also used to generate the observed and expected numbers for the analysis of latency and duration. RESULTS . There were 18,798 white men who were eligible for inclusion in the cohort and for whom work histories were located. Those who had only demographic data available were excluded from this analysis (n = 279). The mean year of birth for this latter group was 1907 (about 13 years less than for the former group), the mean year first employed was 1937 (eight years less), and the mean year last employed was 1960 (10 years less). Thirty-six percent of the cohort with a work history was less than 30 years old at entry into the study, 35.4% was aged 30-39, 18.3% was 40-49, and 10.3% was 50 and over. The average age at entry was 36. The cohort was followed for an average of 19 years. The average age at entry and the average follow-up was similar for each of the job category groups. There were 70,163 jobs recorded for the total cohort. The avera e len time a job was . The job with the greatest nu7m+e-r-of person-yeaOrfs was pipefitter; the job held most often was yard laborer, although the average length of time spent in this job was short (2 years). Table I shows the number of persons employed in the major job subgroups, the average number of person-years spent in each, and the total number of person-years of exposure. Tables 11-VI show the observed number of deaths, the SMRs, and the 95% confidence limits associated with each for the major job categories, plus the job group of pipefitters and boilermakers. It should be noted that for all the analysis groups, the all causes SMRs and their confidence limits are less than 100. All of the job groups show deficits for the major causes of death, such as all cancer, lung cancer, cardiovascular disease, and external causes of death. Most of these and many of the other causes of death are significantly low. Most groups show slight elevations for those causes which were elevated for all white males, including pancreatic cancer, leukemia, brain cancer, other lymphatic cancer, and benign neo- plasms [Divine et al, 19851. TABLE I. Job Categories and Number of Persons and Person-Years of Exposlire Grouped job category subcohort No. of persons No. of person-years of exposure Office personnel Opcrarions Maintenance Laborarory 6.969 12,102 13.569 2.698 117,714 225,928 260.903 5 1.676 Average years hcld 12.79 11.53 8.90 13.12 374 Divine and Barron TABLE II. SMRs for Selected Causes in White Males Employed as Omce and Managerial Workers, by Length of Time Employed > 1 Year > 5 Years Cause of death (ICD 7th)* Observed Observed deaths SMR 95% C1 deaths SMR 95% C1 All causes (000-998) All infective and parasitic diseases (001-138) All cancer (140-205) Cancer of digestive system (150-159) Cancer of stomach (151) . Cancer of large intestine (153) Cancer of pancreas (157) Cancer of lung (162,163) Cancer of prostate (177) Cancer of kidney (180) Cancer of bladder (181) Cancer of skin (190) Cancer of brain (193) Lymphosarcoma (200) Hodgkin's disease (201) Leukemia (204) Cancer of other lymphatic tissue (202,203, 205) Benign neoplasms (210-239) Endocrine, metabolic diseases (240-289) Vascular lesions of CNS (330-334) Diseases of circulatory system (400-468) Arteriosclerotic heart disease (420) Respiratory diseases (470-527) Emphysema (527) Diseases of digestive system (530-587) Cirrhosis of liver (581) Diseases of G-U system (590-639) External causes of death (800-998) Accidents (800-962) 1,445 75 (75-79) 1,134 72 (68-76) 7 26 (10-53) 275 74 (66-83) 5 23 (7-54) 210 69 (60-79) 92 81 (65-99) 13 55 (29-93) 72 77 (60-97) 9 46 (21-88) 27 77 (51-112) 32 151 (103-213) 59 53 (41-69) 20 69 (42-107) 9 100 (46-190) 11 85 (42-152) 5 88 (28-206) 6 65 (24-142) 6 78 (28-169) 1 27 (0-150) 18 124 (74-197) 22 76 (48-116) 25 143 (93-212) 47 51 (38-68) 18 75 (44-119) 7 95 (38-195) 9 84 (38-160) 3 65 (13-191) 4 54 (15-138) 4 64 (17-163) 1 34 (0-188) 11 93 (46-167) 8 97 (67-136) 5 109 (35-255) 6 92 (34-201) 3 81 (16-236) 22 61 (38-92) 18 60 (36-95) 113 72 (60-87) 92 72 (58-88) -. 727 79 (73-85) 570 75 (69-82) 592 80 (73-86) 55 48 (36-63) 15 46 (26-76) 466 76 (70-84) 46 49 (36-65) 12 44 (23-77) 40 46 (33-62) 19 47 (29-74) 17 57 (33-92) 80 62 (49-78) -54. 62 (46-81) 33 47 (32-65) 15 46 (26-77) 13 54 (29-93) _ _58 57 (43-74) 47 62 (45-84) *ICD 8th Rcvision codes for 1968-1977 deaths were converted to ICD 7th Revision. Those employed in the office and professional job categories for one or more years show a statistically significant increased rate of cancer of the pancreas (SMR = 151). However, the SMR drops to 143 and is no longer significant for the group employed five years or more. Those employed as operators show elevated SMRs for the same causes as for the total cohort, although none are significant. The only causes showing higher SMRs for the five-years-or-more group than for the one-year-or-more group are cancer of the pancreas (SMR = 103,119) and cancer of the prostate (SMR = 90,107). Those employed as maintenance workers also show nonsignificant elevations for the above causes of death plus skin cancer. Those causes where the elevated SMR Mortality Patterns by Specific Job Groups 375 C TABLE 111. SMRs for Selected Causes in White Males Employed as Operators by Length of Time Employed > 1 Year > 5 Years Cause of death (ICD 7th)* Observed Observed deaths SMR 95% CI deaths SMR 95% CI All causes (OOO-998) 2,274 77 (73-80) 1.771 77 (73-80) All infective and parasitic diseases (001-138) 17 38 (22-61) 15 44 (24-72) All cancer (140-205) 425 75 (68-82) 1,328 74 (66-83) Cancer of digestive system (150-159) 119 70 (58-83) 93 69 (56-85) Cancer of stomach (151) 30 84 (57-120) 23 82 (52-123) Cancer of large intestine (153) Cancer of pancreas (157) 28 53 (35-77) 33 103 (71-144) 20 49 (30-75) 30 119 (80-170) Cancer of lung (162.163) 108 63 (52-77) 83 63 (50-78) Cancer of prostate (177) 37 90 (63-124) 35 107 (74-149) Cancer of kidney (180) 11 80 (39-141) 9 83 (38-158) Cancer of bladder (I8 I ) 10 53 (25-97) 9 60 (27-114) Cancer of skin (190) 3 33 (7-95) 2 29 (3-103) Cancer of brain (193) 19 123 (74-192) 13 112 (59-191) Lymphosarcoma (200) 13 106 (56-181) 8 85 (36-167) Hodgkin's disease (201) 7 106 (43-219) 4 82 (22-210) Leukemia (204) 31 137 (93-194) 21 120 (74-184) Cancer of other lymphatic tissue (202.203.205) 16 132 (75-214) 12 127 (66-223) Benign neoplasms (210-239) 9 121 (55-229) 6 105 (38-229) Endocrine, metabolic diseases (240-289) 50 90 (67-119) 40 93 (66-126) Vascular lesions of CNS (330-334) 195 86 (74-98) 147 82 (69-96) Diseases of circulatory system -. (400-468) 1,120 80 (76-85) 878 81 (75-86) Arteriosclerotic heart disease (420) 932 83 (78-89) 730 83 (77-89) Respiratory diseases (470-527) 99 58 (47-71) 77 58 (46-72) Emphysema (527) 36 74 (52-103) 27 71 (47-103) Diseases of digestive system (530-587) 86 62 (49-76) 72 67 (53-85) Cirrhosis of livcr (581) 41 62 (45-84) 33 66 (45-93) Diseases of G-Usystem (590-639) 29 64 (43-92) 21 60 (37-91) External causes of death (800-998) 112 48 (40-58) 87 51 (41-63) . Accidcnts (XOO-96-2)~ I 75 47 i37-59) 58 50 (38-65) *ICD 8th Revision codes for 1968-1977 deaths were converted to ICD 7th Revision. increases for the one-year-or-more to the five-years-or-more group are: cancer of the pancreas (SMR = 117,140), cancer of the kidney (SMR = 96,129). cancer of the skin (SMR = 103,130), cancer of the brain (SMR = 103,127), Hodgkin's disease (SMR = 86.138). cancer of other lymphatic tissue (SMR = 142,161), and benign neoplasms (SMR = 140,162). The elevations found for those employed for one year or more as laboratory workers include cancers of the brain, pancreas, and skin, and benign neoplasms. The SMR of 210 for cancer of the brain increases to 221 for those employed five years or more. The SMR for benign neoplasms increases from 209 to 261. None of these elevations, however, is statistically significant. I 376 Divine and Barron TABLE IV. SMRs for Selected Causes in White Males Employed as Maintenance Workers by Length of Time Employed > 1 Year >5 Years Cause of Death (ICD 7th)* Observed Observed deaths SMR 95% CI deaths SMR 95% CI All causes (000-998) All infective and parasitic diseases (001- 138) All cancer (140-205) Cancer of digestive system (150-159) Cancer of stomach (151) Cancer of large intestine (153) Cancer of pancreas (157) Cancer of lung (162,163) Cancer of prostate (177) Cancer of kidney (180) Cancer of bladder ( 181) Cancer of skin (190) Cancer of brain (193) Lymphosarcoma (200) Hodgkin's disease (201) Leukemia (204) Cancer of other lymphatic tissue (202,203,205) Benign neoplasms (210-239) Endocrine, metabolic diseases (240-289) Vascular lesions of CNS (330-331) Diseases of circulatory system Arteriosclerotic heart disease (420) Respiratory diseases (470-527) Emphysema (527) Diseases of digestive system (530-587) Cirrhosis of liver (581) Diseases of G-U systcrn (590-639) External causes of dcath 1800-998) Accidents (800-962) 2,358 12 473 132 29 36 39 124 36 14 11 10 17 6 6 27 18 11 48 192 1,184 ' 969 82 27 80 34 25 1 I3 80 77 (74-80) 26 (1346) 80 (73-88) 75 (62-88) 79 (53-114) 66 (46-92) 117 (83-160) 69 (58-83) 87 (61-120) 96 (52-161) 57 (28-101) 103 (49-189) 103 (60-166) 46 (17-101) 86 (31-186) 115 (76-167) 142 (84-225) 140 (70-251) 83 (62-111) 83 (72-96) 82 (78-87) 83 (78-89) 47 (37-58) 54 (36-79) 55 (43-68) 48 (33-68) 54 (35-80) 46 (38-55) 48 (38-59) 1,511 2 326 93 I9 25 30 81 25 12 7 8 13 2 6 14 13 8 33 133 735 596 50 14 53 16 17 68 48 77 (73-81) 7 (1-25) 86 (77-96) 82 (66-100) EO (48-125) 71 (46-105) 140 (94-200) 72 (57-89) 91 (59-134) 129 (67-226) 55 (22-114) 130 (56-257) 127 (68-218) 24 (3-88) 138 (50-299) 93 (51-156) 161 (86-275) 162 (70-319) 89 (61-125) 88 (73-104) 79 (74-85) 80 (74-87) 44 (33-58) 43 (24-73) 57 (43-75) 37 (21-59) 57 (33-91) 45 (35-57) 46 (34-61) *ICD 8th Revision codes for 1968-1977 deaths were converted to ICD 7th Revision. From the analyses of numerous task and unit subcohorts, one group was singled out for further analysis. The group of pipefitters and boilermakers employed one year or more showed elevations in cancers of the pancreas and kidney, lymphosarcoma, leukemia, other lymphatic cancers and benign neoplasms. A statistically significant SMR of 212 for leukemia in those employed one year or more increased to 285 for those five years or more. This group also showed increasing SMRs with increasing ycars of employnient for cancer of the kidney (SMR = 143,193), lymphosarcoma (SMR = 117,133), and benign neoplasms (SMR = 193.369). The latter SMR is also statisticslly signiiicant. Mortality Patterns by Specific Job Groups 377 Y TABLE V. SblRs for Selected Causes in White Males Employed as Laboratory Workers by Length of Time Employed Cause (ICD 7th)* All causes (000-998) All infective and parasitic diseases (001-138) All cancer (140-205) Cancer of digestive system (150-159) Cancer of stomach (151) Cancer of large intestine (153) Cancer of pancreas (157) Cancer of lung (162.163) Cancer of prostate (177) Cancer of kidney (180) Cancer of bladder (181) Cancer of skin (190) Cancer of brain (193) Lymphosarcoma (200) Hodgkin's disease (201) Leukemia (201) Cancer of other lymphatic tissue (202,203,205) Benign neoplasms (210-239) Endocrine, metabolic disease (240-289) Vascular lesions of CNS (300-334) Diseases of circulatory system (400-468) Arteriosclerotic heart disease (420) Respiratory diseases (470-527) Emphysema (527) Diseases of digestive system ( 5 3 0 487) Cirrhosis of liver (581) Diseases of G-U system (590-639) External causes of death (800-998) Accidents (800-962) > lYear >5 Years Observed Observed deaths SMR 95% CI deaths . SMR 95% CI .328 64 (57-71) 255 62 (55-70) 3 40 (8-116) 64 63 (48-80) 3 51 (10-149) 52 64 (48-84) 18 63 (37-99) 3 54 (11-158) 7 78 (31-162) 6 105 (38-229) 18 55 (32-86) 1 18 (0-99) 1 38 (0-212) 1 34 (0-190) 2 109 (12-394) 7 210 (84-432) 1 42 (1-236) 1 72 (1-403) 3 74 (15-218) 11 49 (24-87) 2 46 (5-166) 6 85 (31-185) 3 66 (13-193) 17 64 (37-103) 1 23 (0-130) 1 48 (1-265) 1 44 (1-244) 1 67 (1-375) 6 221 (81431) 1 53 (1-293) 1 90 (1498) 3 94 (19-274) 1 43 (1-242) 3 209 (42-610) 0 0 (0-199) 3 261 (52-761) 6 63 (23-137) 5 66 (21-155) 26 79 (52-116) 19 75 (45-117) -152 65 (55-76) 119 64 (53-77) ~. 127 66 (55-79) 100 66 (53-80) 13 46 (24-79) 9 41 (19-77) 7 86 (34-177) 5 78 (25-183) 8 30 (13-58) 4 28 (7-71) 1 14 (0-80) 32 64 (44-90) 1..9 5.6. (\-34-88)I 3 14 (3-40) 2 17 (2-62) 0 0 (0-68) 28 69 (46-1001 18 66 (39-105) *ICD 8th Revision codes for 1968-1977 deaths were converted to LCD 7th Revision. e The significantly elevated SMRs were examined further, looking at latency from the starting date in the job and the number of years worked in the job. None of the job categories showed an increasing risk of disease with increasing length of employment and latency. Table VU shows the SMRs for pancreatic cancer for the office and managerial workers. The SMRs decrease with both latency and number of years workcd, although all but two of the 35 deaths occurred after 20 or more years of latency. Table VI11 shows similar SMRs for brain cancer in laboratory workers. The SMR is highest for those who worked less than 10 years in the job and is lowest for those who worked as laboratory workers more than 20 years. All seven cases occurred after 20 or more years latency. 378 Divine a n d Barron TABLE VI. SMRs for Selected Causes in White Males Employed as Pipefitters and Boilermakers by Length of Time Employed > 1 Year >5 Years Cause of death (ICD 7th)* Observed Observed deaths SMR 95% CI deaths SMR 95% CI All causes (000-998) 852 80 (75-86) 458 82 (75-90) All infective and parasitic diseases (001-138) All cancer (140-205) Cancer of digestive system 4 25 (7-64) 170 84 (72-98) 48 77 (57-103) 1 12 (0-68) 91 85 (69-105) 24 74 (47-110) (150-159) Cancer of stomach (151) - Cancer of large intestine (153) Cancer of pancreas (157) Cancer of lung (162,163) Cancer of prostate (177) Cancer of kidney (180) Cancer of bladder (181) Cancer of skin (190) Cancer of brain (193) Lymphosarcoma (200) Hodgkin's disease (201) Leukemia (204) 13 99 (53-170) 13 68 (36-117) 12 104 (54-182) 35 59 (41-82) 13 82 (44-141) 7 143 (57-295) 1 14 (0-79) 2 64 (7-230) 5 98 (31-228) 5 117 (38-274) 1 46 (1-256) 17 212 (123-339) 6 88 (32-191) 5 50 (16-117) 6 99 (36-215) 19 60 (36-94) 8 99 (43-195) 5 193 (62-449) 1 27 (0-151) 1 60 (1-336) 2 73 (8-265) 3 133 (27-387) 1 88 (1-488) 12 285 (147498) Cancer of other lymphatic tissue (202,203,205) Benign neoplasms (210-239) 7 164 (66-338) 5 193 (62-451) 1 44 (1-246) 5 369 (119-861) Endocrine, metabolic disease (240-289) Vascular lesions of CNS (330-334) 17 85 (50-136) 65 76 (58-97) 7 67 (27-138) 36 82 (57-113) Diseases of circulatory system (400-468) 443 88 (80-96) 238 90 (79-102) Arteriosclerotic heart disease (420) 366 90 (81-100) 196 92 (80-106) Respiratory diseases (470-527) 32 51 (35-73) 16 49 (28-80) Emphysema (527) 9 51 (23-97) 5 54 (17-125) Diseases of digestive system (530-587) Cirrhosis of liver (581) 23 48 (30-72) 9 41 (19-78) 10 39 (19-72) 2 17 (2-62) Diseases of G-Usystem (590-639) 8 48 (21-95) 3 35 (7-103) Ex~ernalcauses of death (800-998) 45 60 (44-80) 26 67 (44-98) Accidents (800-962) 30 58 (39-83) 18 68 (40-107) *ICD 8th Revision codes for 1968-1977 deaths were converted to ICD 7th Revision. TABLE VU. ShlRs for Pancreas Cancer in Offce and Managerial Workers by Latency and Years Years worked in job 0-9 10-19 20 + Total <20 Years Latency 2 20 Years Total Obs Exp SMR 95% CI Obs Exp SMR 95%CI Ohs Exp SMR 9 5 % C I - -2 0.79 253 (28-914) 15 8.24 182 (102-300) 17 9.03 188 (110-301) 0.62 - 9 5.14 175 (70-313) 9 5.77 156 (71-296) -- - - 9 9.26 97 (44-185) 9 9.26 97 (44-185) 2 1.42 141 (16-510) 33 22.64 146 (100-205) 35 24.06 145 (101-202) I Mortality Patterns by Specific Job Grdups 379 *. TABLE VIII. SMRs for Brain Cancer in Laboratory Workers by Latency and Years Worked Years worked (in job) 0-9 10-19 20 + Total - <20 Years Latency >,20 Years Total Obs Exp SMR 95% CI Obs Exp SMR 95% CI Obs Exp SMR 95% CI 0 0.49 0 0 0.44 0 -0 - - 5 1.23 408 (131-952) 5 1.72 292 (94-680) - - -0 0.54 0 - 0 0.97 0 - 2 1.21 166 (19-599) 2 1.21 166 (19-599) 0 0.92 0 - - -- 7 2.97 236 (95486) 7 3.89 180 (72-371) -- --- TABLE IX.SMRs for Leukemia in Pipefitters and Boilermakers by Latency and Years Worked Years worked in job <20 Years Latency 320 Years Total Obs Exp SMR 95% CI Obs Exp SMR 95% CI Obs Exp SMR 95% CI 0-9 10-19 20 + Total ._ 1 1.03 97 (1-542) 2 0.36 552 (62-1.995) --- - 3- i.39 216 (43-631) 11 8.04 137 (68-245) 12 9.07 132 (68-231) -.4 1.00 400 (108-1,024) 6 1.36 441 (161-959) 3 1.75 171 (34-500) 3 1.75 171 (34-500) 18 i3zo' 167 (99-264) 2 i 3 2 . 1 7 172 (107-264) - TABLE X. SMRs for Benign Neoplasms in Pipefitters and Boilermakers by Latency and Years Years worked injob 0-9 10-19 20+ Total <20 Years Latency >,20 Years Obs Exp SMR 95% C1 Obs Exp SMR 95% CI 0 0.49 0 - 0 2.46 0 - 1 0.16 610 (8-3,393) 3 0.30 1,OOO (2012,922) - - - - 1 0.54 185 (2-1,028) 1 0.65 154 (2-856) 4 3.30 121 Total Obs Exp SMR 95% CI - 0 2.95 0 - 4-.0.46 862 (232-2.202) 1 0.54 185 (2-1,028) 5 3.95 127 (41-295) Table IX shows the SMRs for leukemia in pipefitters and boilermakers. Most of the 21 deaths occurred in those who worked less than ten years in this job and those with 20 or more years of latency, with no other category containing five or more deaths. Examining years employed in the job, the SMR increases for those employed 10-19 years in the job and decreases for those employed more than 20 years. No consistent pattern is seen. Table X shows the SMRs for benign neoplasms for those who worked as pipefitters and boilermakers. Four of the five deaths cluster in the group who worked in this job for 10-19 years, giving very high SMRs which increase with increasing latency. One death occurred in those who worked 20 or more years after 20 years of latency. However, the number of deaths in each cell is three or less. Again, the pattern is not consistent. DISCUSSION Earlier mortality studies of refinery workers generally have looked at all plant personnel or at hourly workers (Hanis et al, 1982; Hanis et al, 1979; Rushton and .. 380 Divine and Barron Alderson, 1981; Theriault and Goulet, 1979; Gaffey et al, 1974; Thomas et al, 1980). Occasionally the overall cohort is subdivided, using factors such as length of employ- ment or time period first employed. Some studies have used the work history to define cohort subgroups exposed to a specific chemical or process [Tsai et al, 1983; Wen et al, 19851. One study looked at exposed vs nonexposed workers manis et al, 19791, and another used the last job title to divide the cohort into three exposure intensity groups [Gaffey et al, 19743. None of these studies used the employees' complete work histories to determine multiple job groups for analysis. This study was an attempt to determine if mortality rates for any specific cause of death were elevated for employees who worked in specific job categories. Those causes of death which showed elevated SMRs for the total cohort were also studied in detail to determine if they clustered in any job categories. Complete work histories at TMS participating locations were used to categorize the workers into subgroups based on jobs held and length of time held. All job categories showed significant deficits for the all causes of death category, and large deficits were seen for almost all the major causes of death for each subgroup. Essentially, the mortality patterns in the job subcohorts were very similar to those for the entire cohort of white males [Divine et al, 19851. Most of the causes of death with elevated SMRs are based on small numbers of deaths. When the cohort is subdivided into job categories, the number of deaths becomes even smaller. Although some SMRs are elevated when examined by latency and years worked, most are based on fewer than five deaths. No patterns of increasing SMRs with increasing latency and number of years worked are seen for any of the job groups. The SMR is elevated for the job subcohorts of office and managerial workers and pancreatic cancer; laboratory workers and brain cancer; and pipefitters and boilermakers and leukemia and benign neoplasms; however, the results of more specific analyses do not support a causal association. Any trends seen should be interpreted with caution, since SMRs are indirectly adjusted ratios, and the underlying age distributions are likely to be dissimilar. Furthermore, because of the small number of deaths, the SMRs are very unstable and the confidence limits are very wide. Although many of the SMRs seem to decrease with increasing duration of employment, the confidence limits overlap to such an extent that the opposite effect cannot be ruled out. A major problem in carrying out these analyses was deciding what criteria to use in assigning persons to the specificjob groups. Typically, a person was first hired into one of the entry-level positions such as yard laborer. In the first several months to years, he had numerous job changes. Also, at many of the plants, an employee may have had as many as 12 concurrentjobs listed on the work record, with no means of determining how much time was spent in each job. This list of concurrentjobs also changed frequently. Usually, after a few years of numerous job changes, the employee settled into a specific craft or became an operator on a specific process unit. These patterns of numerous job changes make it difficult to decide which job(s) should be considered when assigning persons to job cohorts. If each job ever held is used, certain entry-level positions would include most of the cohort, and persons with many job changes would be counted many times. If the longest job held is used, a person is only counted once but the earliest and possibly the jobs with the greatest potential for exposure would be ignored. For cach job, three cohorts were defined: ever employed. employed one year or more, and employed five years or more. Essentially, the patterns were similar for --- .- e Mortality Patterns by Specific Job Groups 381 P all groups. However, the numbers exposed became small and the results unstable for many of the job groups as length of employment increased. These results should be interpreted with caution. Most of the elevated SMRs are based on very few cases. In addition, because of the small number of deaths for the causes of interest in the job subcohorts, calendar t h e of exposure was not taken into consideration. Numerous comparisons were done, and one out of 20 results could be statistically significant by chance alone. An attempt was made to group tasks and units with similar exposures and responsibilities; however, almost all jobs in a refinery involve exposures to a wide variety of straight chain and aromatic hydrocarbons. Only limited information is available on the chemicals to which these persons were exposed for the periods before the 1970s. Employees with a large number of job changes have the potential for exposure to almost every chemical in the workplace. Also, most of the decedents in this study were hired before 1940, and some as early as 1903. Some of the jobs, process units, refining techniques, and exposures have changed substantially throughout this century. It is, therefore, impossible to classify a job category by levels of exposure to specific chemicals. These factors would tend to bias against finding a positive association. In summary, an examination of work history by grouped job categories shows deficits for the leading causes of death. Although some elevated SMRs are seen, the small numbers and inconsistent trends preclude drawing conclusions about whether these elevations may be work-related. Further investigations of this cohort will be carried out, including case-control studies of the causes of death of interest and updates of the cohort over time. REFERENCES -American Petroleum Institute (1979): Job Code Classification System Part I Petroleum Refineries and Selected Petrochemical Operations. Washington. D.C. Divine BJ, Barron V, Kaplan SD (1985): The Texaco Mortality Study. I. Mortality among refinery. petrochemical, and research workers. J OCCUPMed 27:445-447. Gaffey W, Weaver NK, Tabershaw IPA (1974): Mortality Study of Petroleum Refinery Workers. Washington, D.C.: The American Petroleum Institute. Hanis NM, Holmes TM, Shallenberger LG, et a1 (1982): Epidemiologic study of refinery and chemical plant workers. J Occup Med 24:203-212. Hank NM. Stavrake KM. Fowler JL (1979): Cancer mortality in oil refinery workers. J Occup Med 2 1 :167- 174. Monson RR (1974): Analysis of relative survival arid proportional mortality. Comput Bionied Res b 7~325-332. Rushton L, Alderson MR (1981): An epidemiological survey of cight oil refineries in Britain. Br J Ind Mcd 38~225-234. Theriault G, Goulet I(1979): A mortality study of oil refinery workers. J Occup Mcd 21:367-370. Thomas TL, Decouflk P. Moure-Ernso R (1980): Mortality aniong workers cniployed in petroleum refining and petrochcmical plants. J Occup Med 22~97-103. Tsai SP, Wen CP. Weiss NS. et al (1983): Retrospective mortality and medical surveillance studies of workers in benzene areas of retineries. J Occup Med 25:685-692. Wen CP, Tsai SP, Weiss NS, et al (1985): Long-term mortality study of oil rcfinery workers. IV. Exposure to the lubricating-dewaxing process. JNCI 74: II - 18. .- I