Document 5L7r4bDQRpBq8p84Gk8O5RZv0

[""`""I z Ex? Y TEXACO MORTALITY STUDY TEXACO INC. MAY, 1983 d I' i CONTENTS 1. Mortality Among White Male Refinery, Petrochemical, and Research Workers. . . . . . . . . . . . . . .. I I I . Patterns of Mortality Among Nonwhite Males and Among Females.. .................... . . I 5 Ill. Analysis of Mortality by Job and Case-Control Studies of Specific Causes of Death.. . . . . . . . . . . . . . .23 4 Appendix: Job Task and Process Unit Descriptions. . . . . . . . . . . . .41 I I. MORTALITY AMONG WHITE M A L E REFINERY, PETROCHEMICAL, AND RESEARCH WORKERS Barbara J. Divine, Ph.D. Virginia Barron, M.S. Texaco Inc. P.O. Box 37327 Houston, Texas 77237 Samuel D. Kaplan, M.D. SRI International 333 Ravenswood Avenue Menlo Park, California 94025 ABSTRACT The Texaco Mortality Study is a cohor: mortality siudy of all persons who were employed at least five years in a refining, petrochemical. or research facility and who worked sometime between January 1, 1947, and December 31, 1977. Basic demographic data and complete work history information were collected and coded for all eligible employes. This is a report on the study methodology and on the overall mortality results for white males. There were 19,077 white males in the cohort with 14,609 alive, 4,024 dead, and 444 with vital statuseunknown. There were 5.559 expected deaths giving an all causes standardized mortality ratio (SMR) of 72. Large statistically significant deficits were seen for all the major causes of death including all malignant neoplasms ( S M R = 73). cancer of the lung (SMR = 58), stroke (SMR = 75), arteriosclerotic heart disease (SMR = 77): respiratory disease (SMR = 50),and accidents (SMR = 51).There were six causes of death where the SMR was greater than 100: however, none of these increases was statistically significant! and each was based on only a few deaths. The elevated SMRs were seen for cancer of the pancreas (SMR = 104), cancer of the brain (SMR = 108), Hodgkin's Disease (SMR = 106), leukemia (SMR = 113), cancer of oiher lymphatic tissue (SMR = 111). and benign neoplasms (SMR = 144). INTR0DUCTI0N Numerous recent studies have examined patterns of mortality in petroleum workers. While most have shown overall mortality rates to be substantially less than those for the general population, deaths from a number of causes have been elevated. These causes include cancer of the lung,', ', cancer of the '.skin,', cancer of the nasal cavity and sinuses,'. cancer of the digestive 3. 4. cancer of the brain,2,4.6cancer of the pancreas,'.' cancer of the kidney,7 lymphoma,6,8and It can be noted, however, that the same malignancies have not been elevated in all the studies, and many of the studies suffered from methodological drawbacks. This report details a cohort mortality study of workers who were employed five years or more at Texaco refineries, research laboratories, or chemical plants and who worked between January 1, 1947, and December 31, 1977. The major objective of the study was to determine the patterns of mortality in these workers and, in particular, to determine if any of the elevated causes of death found previously also occurred in the Texaco population. The study was designed to provide detailed information on a large number of long-term Texaco employes in the above mentioned types of facilities, in order to avoid most of the limitations of previous refinery worker studies. Definition of Cohort METHODS -The study cohort consists of all employes of Texaco Inc. who worked at selected refinery, petrochemical, and research facilities at least one day between January 1, 1947 and December 31,1977, who were employed at these facilities for t otal of at least five yGarsoo,wh-a ye8 by the facility on either their last day of employment or the common closing date (December 31,1977). The participating facilities were all those in operation when the study was designed and are listed in Table 1. Two groups of employes were excluded from the cohort:. Employes of the Administrative, Engineering, and Medical Services Departments who could not be identified from the personnel rosters (see below). Employes of the former Jefferson Chemical Company (see Table 1) locations who terminated before 1955 and whose work erc- 1 .+ Type of Facilitv -Refining TABLE 1 PARTICIPATING FACILIT1ES Location Amarillo, Texas Anacortes, Washington (Puget Sound) Casper, Wyoming Convent, Louisiana Eagle Point, New Jersey El Paso, Texas Houston, Texas (Headquarters 0f f ice) Lawrenceville, Illinois Lockport, Illinois Los Angeles, California Port Arthur, Texas (includes Port Arthur Terminal) Packaging Chemical - -Research Bayonne, New Jersey Port Arthur, Texas Austin, Texas' Conroe, Texas* Eagle Point, New Jersey Port Arthur, Texas Port Neches, Texas' -. Beacon, New York Montebello, California Port Arthur, Texas ' Formerly Jefferson Chemical Company Date Founded 1920 1958 1923 1967 1949 1929 1963 1931 1911 1928 1903 1906 1910 1909 1903 1949 1961 1961 1957 1948 1931 1946 1954 2 All employes who mer the eligibility requirements were identified from personnel system separation rosters listings all employes who separated from the company between 1947 and 1978, and from personnel listing of employes active in 1978. Numerous other cohort mortality studies have required a minimal length of employment much shorter than the five year requirement used in this study (e.Q.one year, six months, one day). The argument against a short employment reauirement is that the resulting cohort may include an undue proportion of people who spent t k largest part of their working lifetime at other companies in jobs with unknown exposures. On the other hand, the argument for a short employment requirement is that it is unknown whether a short exposure, especially early in life. may have adverse health effects. In order to test whether the five-year requirement affected the pattern of results (sarticularly whether it made some of the cancer SMRs too low), persons who worked one to five years at the Port Arthur facilities (excluding the Package Division) were also studied. Comparison of mor?ality ?atterns for t h e two groups were made. Data Collection and Coding Data collected on each employe included name, social security number, race, sex. date of birth, and a complete work histcry, i f available. Vital status information for former employes was obtained from Texaco where available or the Social Security Administration (SSA). Copies of death certificates were obtained either from Texaco at the time of data collection, or from the states where the deaths occurred.* Each death certificate was coded for the underlying cause of death by a nosologist. Deaths occurring through 1967 were coded according to the 7th Revision of the International Classification of Diseases and deaths after 1967 according to the 8th Revision. Cohort Verification In order to verify the accuracy and completeness of cohort ascertainment, data were collected at two locations without reference to the personnel system lists of eligible persons. The names found were compared with the roster of eligible workers. The vast majority of persons missing from the personnel system lists were employed by ineligible departments (e.g. accounting, medical) or had transferred to a nonparticipating location before the study end date and, therefore, were appropriately excluded from the study. A number of persons were on the personnel rosters but had been missed in the process of screening these rosters for eligible employes. These persons were added to the study cohort. Although a small number of eligible persons are likely to be missing for other participating locations as well, this should not bias the patterns observed since a person's inclusion in the study should not have been related to his mortality experience. A very small number of persons (less than 1.5%) appeared to be eligible according to their work histories but were not on the personnel roster. No reason for this could be determined. RESULTS The Studv PoDulation There were 21,314 persons eligible for inclusion in the study cohort. Table 2 shows the number of subjects according to sex, race, and vital status as of the study end date, December 31, 1977. Those persons with race and sex unknown were assumed to be white males. (There were 28 persons with sex unknown and 344 with race unknown.) White males made up 89.5% of the total cohort and 94.1O/O of the deaths. All analyses were done on the subcohort of white males. Further reports will examine mortality in nonwhites and females. Table 3 shows the breakdown of the cohort by plant and vital status as well as the person years of observation at the larger locations. Of the 19,077 total white male cohort, 14,609 (76.6%) are still alive, 4,024 (21.1%) are deceased, and 444 (2.3%) have vital status unknown. Approximately 75% of the cohort is concentrated in six of the 20 locations with over 43% at the Port Arthur facilities (Port Arthur Refinery, Chemical Plant, Terminal, and Package Division). Table 4 shows the cohort employment status at the end of the study. Almost half were still active employes. and over a quarter were retired employes. We would liKe t 3 ackiowledge the cooperaticn of t i e Lariaus stare health departments, includinG the state of Neb Jersey. in pro vidlnc copies o: C s a i h cert,ficates for decease2 T ~ P D S ' S c' tne cohort 3 Table 5 shows the distribution of the cohort and of the deaths by year of birth; the births appear to be evenly distributed throughout for the total cohort while almost half of the decedents were born before 1900 and over three-quarters before 1910. Table 6 shows the distribution of the cohort and the deaths by year first employed. Well over half of the cohort was employed before 1950 and about one third of the cohort was employed before 1940. Almost three-quarters of the decedents, however, were first employed before 1940. Most of the cohort (56.3 G ; ~ ) were first employed before age 25 and very few (9.7%) after age 35. Table 7 shows thedistribution of the cohort and the deaths by length of employment. Over one third of the cohort and one half of the deaths were employed 30 years or more. These findings substantiate the claim that employment in the petroleum industry tends to be long-term and stable. Table 8 shows the breakdown of the decedents by year of death and Table 9 by age at death. Most of the cohort died in the latter half of the study period, and most were age 55 to 79 at the time of death. White Males Nonwhite Males White Females Nonwhite Females Total TABLE 2 COHORT VITAL STATUS BY RACE AND Sf3 Vital Status Alive 14,609 868 771 56 16,304 Dead 4,024 176 72 5 4,277 Unknown 444 16 262 3 725 Total 19,077 1,060 1,105 64 21,306 4 TABLE 3 WHiTE MALES B Y PLANT AND VITAL STATUS Type of-Faciiityl Plant Location Refining: Amarillo Plant Casper Plant Eagle Point Plant El Paso Plant Houston Office Lawrenceviiie Plant Lockport PIant Los Angeles Plant Louisiana Plant Port Arthur Plant Port Neches Plant Puget Sound Plant West Tulsa Plant Packaging: Bayonne Terminal Petrochemicals: Austin Research Lab Conroe Chemical Plant Neches Chemical Plant Research: Beacon Research Lab Montebello Research Lab Port Arthur Research Lab Unknown Total Vital Status Alive D_e_ad Unknown 239 292 859 238 99 713 992 953 241 6,194 337 331 475 53 59 94 23 33 305 473 266 6 1,906 202 27 180 2 22 9 19 6 77 13 64 1 142 6 2 12 266 102 0 199 10 4 105 3 3 697 43 11 978 54 327 0 14.609 189 4 37 3 4,024 30 15 3 3 444 T-otal 294 383 972 280 138 1,095 1,478 1,283 248 8.242 545 360 667 368 213 110 751 1,197 83 367 3 19,077 Person Y-ears 6.727 9.685 15.962 6,150 - 24,065 34 -923 24,068 - 170.834 17,392 4.871 14.985 9,217 2,875 1,073 11!177 24,846 1,496 6.564 - 365.510 5 i TABLE 4 WHITE MALES BY EMPLOYMENT STATUS AS OF DECEMBER 31,1977 Active Retired Resigned Died Dismissed Disability Retirement Left Texaco, Reason Unknown Total Number 8.998 5.281 1,962 1,405 277 994 160 19,077 Percent 47.2 27.7 10.3 7.4 1.5 5.2 0.8 100.0 Work history not found Before 1890 1890 - 1899 1900 - 1909 1910 - 1919 1920 - 1929 1930 - 1939 1940 B After Unknown Total TABLE 5 WHITE MALES BY YEAR OF BIRTH Total Cohort NO. (%) 607 (3.2) 1,954 (10.2) 3,119 (16.3) 3,596 (18.8) 3,984 (20.9) 2,547 (13.4) 3.268 (17.1) 2 19,077 Deaths NO. (%) 514 (12.8) 1,352 (33.6) 1,202 (29.9) 640 (15.9) 316 (7.9) 4,024 6 Before 1940 1940 - 1949 1950 - 1959 1960 - 1969 1970 8, After Total TABLE 6 WHITE MALES BY YEAR FIRST EMPLOYED Total Cohort No.(%) 6,411 (33.6) 4,923 (25.8) 3.406 (17.9) 2.692 (14.1) 1,645 (8.6) 19.077 Deaths NO.(%) 3,000 (74.6) 829 (20.6) 195 (4.8) 4,024 5 - 9 Yrs 10 - 19 Yrs 20 - 29 Yrs 30+ Yrs Total TABLE 7 WHITE MALES BY LENGTH OF EMPLOYMENT Total Cohort NO.(O/o I 4,201 (22.0) 3,002 (15.7) 5,081 (26.6) 6,793 (35.6) 19.077 Deaths NO.(O/o) 205 (5.1) 641 (15.9) 1,156 (28.7) 2.022 (50.2) 4.024 1947-1949 1950-1959 1960-1969 1970-1977 Unknown Total TABLE 8 WHITE MALE DEATHS BY YEAR OF DEATH Number 108 726 1,507 1,679 4 4>G24 7 / Percent 2.7 18.0 37.5 41.7 0.1 100.0 .IUnder 40 40-49 50-59 60-69 70-79 80 8. Over Unknown Toral TABLE 9 WHITE MALE DEATHS BY AGE AT DEATH Number 105 403 858 1,240 1,065 349 4 4,024 Percent 2.6 10.0 21.3 30.8 26.5 8.7 0.1 100.0 The 444 persons for whom vital status was unknown were assumed to be alive in the analysis of mortality Because the percentage is so small (2.3%) and because they are distributed throughout the cohort. it is not expected that this assumption will affect the results. Death certificates were obtained for alt except 152, or 3.8% of the 4,024 decedent white males. Patterns of Mortality Analysis of mortality was performed using Monson's computer program to compare the observed and expected number of deaths using the U S . white male population as the comparison group. For the 4,024 deaths observed 5.558.9 were expected giving a standardized mortality ratio (SMR) of 72.' Table 10 shows the observed and expected number of deaths for the major causes of death plus the SMRs and their lower and upper 95% confidence limits. The most striking result is the large deficit seen for almost every major cause of death category including all malignant neoplasms (SMR = 73), cancer of the digestive organs (SMR = 67), cancer of the lung (SMR = 58), metabolic diseases (SMR = 75), diabetes ( S M R = 77), diseases of the nervous system (SMR = 76), vascular le- sions of the central nervous system (CNS) (SMR = 75), arteriosclerotic heart disease (SMR = 77), all respiratory diseases (SMR = 50), all diseases of the digestive system (SMR = 54), and all external causes of death (SMR = 52). For none of these categories does the 95% confidence interval include 100. There are a few causes of death for which the SMR is greater than 100; however, for none of these does the lower confidence limit exceed 100. The highest SMR (144) is for benign neoplasms with 20 deaths observed and 13.9 expected. Of these. 12 were brain tumors; 9.3 were expected, giving an SMR of 130. The SMR for cancer of the brain and CNS was slightly elevated at 108. Other causes of death for which the SMR was greater than 100 are cancer of the pancreas (SMR = 104), Hodgkin's Disease (SMR = 106), leukemia (SMR = 113), and cancer of other lymphopoietic tissue (SMR = 111). * It should be noted that i f those u i t h vl!al s!atrls unknown are assumed lost on the date last employed, the overall SMR IS 75 Even under the extremely unlikely assurnptlon that they are all deceased mldway between the date last employed and the study end date, the overall SMR I S 83 8 * Cause TABLE 10 SMRs FOR SELECTED CAUSES OF DEATH WHITE MALES, 1947-1977 Observed Expected SMR All Causes of Death All Infective & Parasitic Diseases All Malignant Neoplasms Cancer of the Digestive Organs Cancer of the Stomach Cancer of the Large Intestine Cancer of the Pancreas Cancer of the Lung Cancer of the Prostate Cancer of the Kidney Cancer of the Bladder Cancer of the Skin Cancer of the Brain Lymphosarcoma & Reticulosarcoma Hodgkin's Disease Leukemia Cancer of Other Lymphatic Tissue Benign Neoplasms Endocrine, Metabolic Diseases All Diseases of the Nervous System All Vascular Lesions of CNS All Diseases of the Circulatory System Arteriosclerotic Heart Disease All Respiratory Diseases Emphysema All Diseases of the Digestive System Cirrhosis of Liver All Diseases of G-U System All External Causes of Death All Accidents Unknown 4,024 22 767 215 45 64 62 182 62 24 19 14 31 20 13 48 25 20 78 360 322 1,975 1,623 159 52 141 56 43 227 152 152 5,558.9 82.4 1,056.4 318.6 66.4 98.2 59.8 315.1 77.9 25.9 35.5 17.3 28.7 22.9 12.3 42.3 22.6 13.9 103.8 473.8 431.7 2,609.4 2,101.1 319.1 89.7 260.2 122.9 85.1 434.8 297.0 72 27 73 67 68 65 104 58 80 93 54 81 108 87 106 113 111 144 75 76 75 76 77 50 58 54 46 51 52 51 -LCL 70 17 68 59 49 50 80 50 61 59 32 44 73 53 56 84 72 88 59 68 67 72 74 42 43 46 34 37 46 43 -UCL 75 40 78 77 91 83 133 67 102 138 84 136 154 135 181 150, 164 222 94 84 83 79 81 58 76 64 59 68 59 60 Similar analyses of the patterns of mortality were carried out for each of the plants in the study excluding those with fewer than 20 deaths and those opened since 1958. A person was counted at each plant where he was ever employed and. therefore, could be included more than once. Essentially, the patterns of mortality seen at each facility were similar to those for the overall cohort. Deficits were seen for almost all of the major causes of death at every location. The Port Arthur complex (Refinery, Terminal, Package Division, and Chemical Plant) includes almost half of the t d a l cohort and almost half of the deaths. Table 12 shows the SMRs for selected causes of death for all white males employed five years or more, and who were ever employed at the Port Arthur complex. The mortality patterns are the same as those for the entire cohort. The SMR for benign neoplasms is higher for the Port Arthur facilities than for the total cohort and is statistically significant. It should be noted that while the distribution of lymphatic and hematopoietic cancers differs between the total cohort and Port Arthur, the overall SMRs are the same. The differences may be due to diagnostic preferences. Table 12 shows SMRs for selected causes of death for the expanded cohort of white males employed one year or more at the Port Arthur complex (excluding the Package Division). There were 1,614 additional persons in the cohort and 72 additional deaths. A comparison with Table 11 shows the same pattern of mortality with many of the SMRs slightly lower for the latter group. The lower SMRs can be seen particularly for the major causes of death. For example, in looking at the one to five year group alone, the all causes of death SMR is only 64 compared with 73 for the five years or more group, that for malignant neoplasms is only 44 compared with 76, and that for arteriosclerotic heart disease is 65 compared with 75. Lower SMRs are seen for the one to five year group for every cause of death category with more than four deaths. It is, therefore, concluded that restricting the study to those employed at least five years or more has not obscured any patterns of death. 10 TABLE 11 SMRs FOR SELECTED CAUSES OF DEATH WHITE MALES, PORT ARTHUR COMPLEX EMPLOYED FIVE YEARS OR MORE Cause" Observed Expected SMR LCL UCL All Causes of Death 1.958 2,667.7 73 70 77 All Infective & Parasitic Diseases 8 40.8 20 8 39 All Malignant Neoplasms 393 509.2 77 70 85 Cancer of the Digestive Orcjans 120 154.4 78 64 93 Cancer of the Stomach 27 32.3 84 55 122 Cancer of the Large Intestine 36 47.5 76 53 105 Cancer of the Pancreas 33 28.9 114 79 160 Cancer of the Lung 88 151.6 58 47 72 Cancer of the Prostate 37 37.9 98 69 135 Cancer of the Kidney 12 12.4 97 50 169 Cancer of the Bladder 11 17.3 64 32 114 Cancer of the Skin 7 8.1 86 35 178 Cancer of the Brain 15 13.5 111 62 183 Lymphosarcoma & Reticulosarcoma 12 10.9 110 57 192 Hodgkin's Disease 6 5.9 103 37 223 Leukemia 19 20.3 93 56 146 Cancer of Other Lymphatic Tissue 14 10.8 129 71 217 Benign Neoplasms 13 6.7 195 104 333 Endocrine, Metabolic Diseases 42 49.9 84 61 114 All Diseases of the Nervous System 193 227.7 85 73 98 All Vascular Lesions of CNS 172 207.5 83 71 96 All Diseases of the Circulatory System 919 1,253.7 73 69 78 I Arteriosclerotic Heart Disease 754 1,008.6 75 70 80 All Respiratory Diseases 77 154.1 50 39 62 Emphysema 26 44.0 59 39 87 All Diseases of the Digestive System 69 123.5 56 43 71 Cirrhosis of Liver 23 57.4 40 25 60 All Diseases of G-U System 20 41.1 49 30 75 All External Causes of Death 116 204.3 57 47 68 All Accidents 74 140.2 53 41 66 Unknown 59 11 TABLE 12 SMRs FOR SELECTED CAUSES OF DEATH WHITE MALES, PORT ARTHUR COMPLEX EMPLOYED ONE YEAR OR MORE Causg Observed All Causes of Death All Infective & Parasitic Diseases All Malignant Neoplasms Cancer of the Digestive Organs Cancer of the Stomach Cancer of the Large Intestine Cancer of the Pancreas Cancer of the Lung Cancer of the Prostate Cancer of the Kidney Cancer of the Bladder Cancer of the Skin Cancer of the Brain Lymphosarcoma & Reticulosarcoma Hodgkin's Disease Leukemia Cancer of Other Lymphatic Tissue Benign Neoplasms Endocrine, Metabolic Diseases All Diseases of the Nervous System All Vascular Lesions of CNS All Diseases of the Circulatory System Arteriosclerotic Heart Disease All Respiratory Diseases Emphysema All Diseases of the Digestive System Cirrhosis of Liver All Diseases of G-U System All External Causes of Death All Accidents Unknown 2,030 9 400 120 27 36 33 91 37 12 11 7 16 13 6 14 13 42 197 176 939 769 78 26 71 25 21 138 08 73 Expected 2,780.7 43.8 525.2 158.0 33.0 48.7 29.6 155.8 38.2 12.8 17.6 8.7 14.5 11.5 6.7 11.2 7.1 51.6 233.2 211.3 1,283.8 1,031.8 158.1 44.6 129.2 60.5 43.0 245.0 170.6 SMR 73 21 76 76 82 74 112 58 97 94 63 81 110 113 90 125 183 81 84 83 73 75 49 58 55 41 49 56 52 LCL 70 9 69 63 54 52 77 47 68 48 31 32 63 60 33 69 97 59 73 71 69 69 39 38 43 27 30 47 41 UCL 76 39 84 91 119 102 157 72 133 164 112 166 179 194 196 210 313 110 97 97 78 80 62 85 69 61 75 67 64 12 DISCUSSION The most outstanding result of this study is the large deficit of deaths seen for all the leading causes of death. Although an overall SMR less than 100 is expected because of the 'healthy worker effect', the all causes SMR of 72 is lower than is L;sually seen in studies of this size ana length of follow-up. txcept for that in the refinery worker study by Gaffey et al.s which had only 10 years of follow-up, th '.than those reported in all other mortality studies of petroleum w o r k e r ~ . ~',.lo xaco employes are snouia not extend rovided to I exaco employes, persons who remain with Texaco have access to a comfortable standard of living and quality medical care. It was anticipated that excluding those employed less than five years might have lowered some SMRs or obscured some excess cause of death, but this is not the case. The SMRs are consistently lower for those employed one year or more. and the pattern of deaths is the same as for those employed five years or more. There are a few causes of death for which the SMR is greater than 100. These are cancer of the pan- creas. E3n)eotap-slm1s. Mortality from Lauses nas aiso be'en"eT8a'ted in otner stilaies OT petroieum workers. Cancer of the pancreas and cancer of the brain in particular have been associated with refinery e r n p l ~ y m e n t . ' ~H~ow. ~ever, in this study the SMR's for each of these two causes are increased less than lo%, much less than the increases seen in other studies. The largest SMR observed is 144, each is based on an excess of seven or fewer deaths, and none is statistically significant. These excesses seem small and insignificant when compared with the number and size of the deficits. It should be noted that many of the causes of death with large deficits are causes for which mortality rates have declined dramatically over the past 50 years and/or are diseases where medical technology and preventive medicine have had an impact. This can be seen particularly in the SMR of 27 for parasitic and infective diseases. This is a category of diseases where a good standard of living and access to medical care would be expected to have a substantial effect in reducing mortality. This same effect would be expected to a lesser extent and is seen for causes such as cancer of the large intestine, stroke, and heart disease. In contrast, those causes of deaths mentioned above as having elevated SMRs are causes for which no means of prevention are known and for which survival is generally very short after diagnosis. Therefore, the 'healthy worker effect' and access to medical care would not be expected to have any impact on these causes of death. It is, therefore, expected that mortality from these causes would not be significantly different from the general U.S. population. Plant specific mortality patterns were also examined and in general resembled those for the total cohort. However, the S M R s which were elevated for the total cohort were not consistently elevated at all locations. Except for the largest refineries which have process units not found at the smaller refineries, all the refineries have similar process units and, therefore, similar on-the-job exposures. If mortality is, in fact, associated with employment, the same patterns should be seen at all refineries large enough to provide a stable estimate of the SMR. The fact that the elevated SMRs were not seen at all refineries argues against an association with refinery employment. Other studies of petroleum workers have shown excesses for cancers of the digestive cancer of the stomach,* cancer of the ' I and cancer of the skin.',' Elevations for these causes were not seen in the Texaco cohort. In fact. for cancer of the digestive organs, cancer of the stomach, and cancer of the lung, a statistically significant deficit is seen. Recent toxicology studies in animals have associated exposure to gasoline with an increased risk of kidney cancer', and at least one other refinery worker cohort study has shown an elevated SMR for kidney cancer.' No indication of an excess risk of kidney cancer was seen for the TMS cohort. The lower SMR for cancer of the lung (SMR = 58) is particularly interesting and is probably due to the nonsmoking rules in effect throughout the plants. The possibility remains that by looking at mortality in the cohort as a whole. excess risks of death associated with e r n p i o m s p eG r n U' at a sPeJITIC C m T I n Ob-Bcured. i-urtner studies of th% ccnart will inciuas an -I> o i niorraiity a\i speciiic jobs ana case-controi stuares or those causes -0 d t w' in riT r slner srudies ot rerinel * Persora! cm?Cnli~i:a;iO-~ the American Peirote-n: IPsiitute. REFERENCES 1. Blot WJ, Brinton LA. Fraumeni JF et at.: Cancer mortality in U.S. counties with petroleum industries. Science 19851-53, 1977. c 2. Thomas TL, Decoufle P, Moure-Eraso R: Mortality among workers employed in petroleum refining and petrochemical plants. JOM 22:97-103, 1980. 3. Hanis NM, Stavraky KM, Fowler JL: Cancer mortality in oil refinery workers. JOM 21:167-174, 1979. 4. Rushton L, Alderson MR: An epidemiological survey of eight oil refineries in Britain. Br J Ind Med 38:225-234, 1981. 5. Theriault G, Goulet I: A mortality study of oil refinery workers. JOM 21:367-370, 1979. 6. Thomas TL, Waxweiler RJ, Moure-Eraso R, et at.: Mortality patterns among workers in three Texas oil refineries. JOM 24.135-141, 1982. 7. Hanis NM, Holmes TM1 Shallenberger LG et at.: Epidemiologic study of refinery and chemical plant workers. JOM 24:203-212, 1982. 8. Gaffey W, Weaver NK, Tabershaw IPA: Mortality Study of Petroleum Refinery Workers. Washington, D.C.: The American Petroleum Institute, 1974. 9. Monson RR: Analysis of relative survival and proportional mortality. Comput Biomed Res 7:325-332,1972. 10. Wen CP, Tsai SP, Weiss NS, et at.: A population- based cohort study of brain tumor mortality among oil refinery workers with a discussion of methodological issues of SMR and PMR, in Banbury Report 9: Quantification of Occupational Cancer, Pet0 R, Schneiderman M, ed. Cold Spring Harbor Laboratory, 1981, pp 413-432. 11. Gottlieb MS: Lung cancer and the petroleum industry in Louisiana. JOM 22:384-388, 1980. 14 11. PATTERNS OF MORTALITY AMONG NONWHITE MALES AND AMONG FEMALES Barbara J. Divine, Ph.D. Virginia Barron, M.S. Texaco Inc. P.O. Box 37327 Houston, Texas 77237 ABSTRACT An earlier report on the iexaco Mortality Study discussed overall patterns of mortality in white male refinery. petrochemical, and research workers who were employed at least five years at a participating location and who worked sometime between January 1, 1947,and December 31,1977.The standardized mortality ratio (SMR) for all causes of death was 72, and large deficits were seen for all the major causes of death. For a few causes of death the SMR was greater than 100; however, none was statistically significant, and all were based on a few deaths. This is a similar analysis of mortality for the nonwhite males and the females in the cohzrt. The all causes SMR for nonwhite males is 69,and again, large deficits are seen for all the major causes of death. Elevated SMRs are seen for a few causes of death, but each is based on three or fewer deaths, and none is statistically significant, The all causes SMR was also low for females; however, the percentage of females lost to follow-up is extremely high (24%), and the results should be interpreted with caution. INTRODUCTION The previous paper on the Texaco Mortality Study (TMS) was concerned with patterns of mortality among the white males in the study cohort.' This study showed a large deficit for all the major causes of death. There were a few causes of death for which the standardized mortality ratio (SMR) was greater than 100. However, for none of these causes of death was the increase statistically significant, and both the increased risk and the increased number of deaths are very small. This paper is a similar description of the patterns of mortality among the nonwhite males and the females in the TMS cohort. METHODS The cohort eligibility requirements and the data collection methods are the same as described in the paper on white males. RESULTS There were 1,061 nonwhite males, 1,105white females, and 64 nonwhite females eligible for inclusion in the TMS cohort. Table 1 shows the vital status breakdown of the cohort as of the study end date of December 31,1977.f o r nonwhite males, the percentage with vital status unknown is only 1.6;however, for white females 23.7% have vital status unknown. Sixteen percent of the nonwhite males and 6.5% of the white females are deceased. Because of the small number of nonwhite females, they are combined with the white females in the remainder of the paper. Table 2 shows each group by year of birth as well as the breakdown for the deaths. For nonwhite males, over one-third were born after 1940.For females, over one-half of the group was born in the 1920s and 1930s. The distribution by year of birth for white males was more evenly distributed throughout.' In contrast, about one-half of the deaths in each group were born before 1910.Table 3 shows the distribution of each group and the deaths by year first employed. In contrast to the white males where over one-third were employed before 1940,'more than 90% of the total group was first employed in the 1940s and later. While the percentages are not as large, a substantial proportion of the deaths were also first employed after 1940;over threequarters of the white male deaths were first employed before 1940.l More than 70% of the nonwhites were either still employed at the end of the study or left in the last three years. In comparison, white females left regularly throughout the study time period, and only 28.1O/O were actively employed at the closing date of the study. Few in any group were retired, and almost half of the white females had resigned. In contrast, almost half of the white males were still employed at the study end date and over one quarter had retired.' Table 4 shows the distribution of the groups by length of employment. For each total group, the largest percentage is in the five to nine years of employment category, and for the deaths, the largest percentage is in the 10 to 19 years of employment category. This is in contrast to the white males where over 6Ooh of the total and over 7540 of the deaths were employed 20 years or more. More than half of each group was employed less than 20 years whereas for white males over half were employed 30 years or more.' 15 Table 5 shows the breakdown of deaths by year of death. Approximately half of each group died between 1970-1977. Table 6 shows the distribution of the deaths by age at death. For nonwhite males and females. the distributions are similar; however, higher percentages died at earlier ages and lower percentages in the older age groups than the white males.` Patterns of Mortality Persons for whom vital status was unknown were assumed to be alive in the analysis of mortality. For nonwhites, D e percentage is so small that ?he assumption is unlikely to have affected the results. For white females. the percent lost to follow-up is so large that the results should be interpreted with caution. Analysis of mortality was performed using the Monson program' to compare the observed and ex- pected number of deaths. For nonwhite males, the U.S. nonwhite male population was used as the com- parison group, and for females the U.S. white female population was used. Table 7 shows the observed and expected number of deaths for the major causes of death, the SMRs, and their upper and lower confidence limits (95%) for nonwhite males. SMRs are not calculated when the number of observed deaths is less than five. There were 253.4 expected deaths and 176 observed, giving an all causes SMR of 69. All of the major causes of death show substantial deficits. There are a few causes of death for which the observed number of deaths is greater than the expected number (cancer of the pancreas, cancer of the brain, cancer of the skin, cancer of other lymphatic tissue, benign neoplasms, and emphysema). However, each is based on three or fewer deaths, and the confidence limits are very wide and include 100 in every case. Table 8 is a similar table for females. There were 104.8 expected deaths and 77 observed for an all causes SMR of 73. Again, most of the major causes of death show large deficits. However, in contrast to both male cohorts, cancer of the lung shows an elevated SMR of 174 although the confidence limits include 100. In several other instances, the observed number of deaths is greater than expected (cancer of the bladder, cancer of the brain, leukemia, cancer of other lymphatic tissue, and all diseases of the genito-urinary system); however, again all are based on three or fewer deaths, and the confidence limits are very wide and include 100. DISCUSSION As in the study of mortality among white males, the most striking result of this study is the large deficit of deaths seen for all the leading causes of death. The patterns of mortality seen for nonwhite males are very similar to those reported earlier for white males.' Although the actual number of deaths is very small for each cause, more than expected numbers of deaths for the nonwhite cohort are observed for the same causes of death as for white males, namely cancer of the pancreas, cancer of the brain, cancer of other lymphatic tissue, and benign neoplasms. Subsequent case-control studies of these causes of death in white males will attempt to determine if these elevations are associated with job-specific employment in the petroleum industry. While the results for females must be interpreted with caution because of the high percentage with vital status unknown, it is interesting that the size of the deficit in mortality is similar to that for males. It is also interesting that while a deficit is seen for cancer of the lung in males, an excess is seen for females. However, since the largest portion of the females would have worked in office areas where there are no restrictions on smoking, this is not totally surprising. Because the females traditionally worked in jobs with little or no chemical exposures, it seems highly unlikely that the excess is associated with employment in the petroleum industry. Nonwhite males and females show a number of differences in the breakdown by demographic characteristics when compared with the white males. In particular, the nonwhite males and females tend to be younger and to have worked for shorter time periods. For both these groups, this is due to the fact that substantial numbers were not hired until the 1960s and later. In addition, for females this reflects their tendency to work for a few years and then resisn because of marriage or pregnancy, which also may explain the high percentage lost-to-follow-up. Because both of these groups are young and have more recent employment, it will be many more years before the number of deaths is large enough to provide more definitive answers about the relationship of employment in the petroleum industry and mortaliry in nonwhites and females. 16 REFERENCES 1. Divine BJ. Barron V, and Kaplan SD: The Texaco Mortality Study. I. Patterns of mor- tality among white male refinery, petrochemical, and research workers. (Submitted for pub1ication.j 2. Monson RR: Analysis of relative survival and proportional mortality. Comput Biomed Res 7:325-332,1974. .a. 17 TABLE 1 VITAL STATUS FOR NONWHITE MALES, FEMALES AS OF DECEMBER 31,1977 Nonwhite Males White Females Nonwhite Females Alive 868 77 1 56 Vital Status Dead Unknown 176 17 72 262 53 Total 1,061 1,105 64 Before 1900 1900-1909 1910-1919 1920-1929 1930-1939 1940 & After Total TABLE 2 NONWHITE MALES, FEMALES BY YEAR OF BIRTH Total Cohort Deaths Nonwhite Males Number(o/o) 56 (5.3) 98 (9.2) 201 (18.9) 243 (22.9) 104 (9.8) 359 (33.8) -_____ 1.XI1 Females Num ber(%) 37 (3.2) 91 (7.8) 198 (16.9) 402 (34.4) 238 (20.4) 203 (17.4) 1.169 Nonwhite Males Number(O h ) 45 (25.6) 43 (24.4) 50 (28.4) 38 (21.6) 176 Females Number(O h ) 20 (26.0) 17 (22.1) 20 (26.0) 20 (26.0) 77 i8 Before 1940 1940-1949 1950-1959 1960-1969 1970 & After Total TABLE 3 NONWHITE MALES, FEMALES BY YEAR FIRST EMPLOYED Total Cohort Deaths Nonwhite Males Number{%) 17 (1.6) 429 (40.4) 169 (15.9) 216 (20.4) 230 (21.7) 1,061 Females Number(Yo) 97 (8.3) 528 (45.2) 274 (23.4) 175 (15.0) 97 (8.3) 1,169 Nonwhite Males Number(%) 11 (6.3) 141 (80.1) 24 (13.6) 176 Females Number(%) 25 (32.5) 43 (55.8) 9 (11.7) 77 5- 9 Years 10-19 Years 20-29 Years 30+ Years Total TABLE 4 NONWHITE MALES, FEMALES BY LENGTH OF EMPLOYMENT Total Cohort Nonwhite Males Number("/o) 441 (41.6) 171 (16.1) 258 (24.3) 191 (18.0) 1,061 Females Numbet(%) 599 (51.2) 250 (21.4) 187 (16.0) 133 (11.4) 1,169 Deaths Nonwhite Males Number(%) 23 (13.1) 77 (43.8) 59 (33.5) 17 (9.7) 176 Females Number(%) 18 (23.4) 27 (35.1) 22 (28.6) 10 (13.0) 77 19 1947-1949 1950-1959 1960-1969 1970-1977 Unknown Total TABLE 5 NONWHITE MALES, FEMALES DEATHS BY YEAR OF DEATH NWM Number(%) 1 (0.6) 27 (15.3) 57 (32.4) 91 (51.7) 176 Females Number(%) 0 11 (14.3) 27 (35.1) 38 (49.4) 1 (1.3) 77 Under 40 40-49 50-59 60-69 70-79 80 & Over Unknown Total TABLE 6 NONWHITE MALES, FEMALES DEATHS BY AGE AT DEATH NWM Number(O/O) 14 (8.0) 29 (16.5) 47 (26.7) 47 (26.7) 35 (19.9) 4 (2.3) 176 Females Number(%) 8 (10.4) 15 (19.5) 16 (20.8) 20 (26.0) 12 (15.6) 5 (6.5) 1 (1.3) 77 20 TABLE 7 SMRs FOR SELECTED CAUSES OF DEATH Cause NONWHITE MALES, 1947-1977 Observed Expected All Causes of Death All Infective & Parasitic 176 253.4 Diseases All Malignant Neoplasms Cancer of the 2 8.4 30 43.5 Digestive Organs Cancer of the Stomach Cancer of the 11 14.4 3 4.0 Large Intestine Cancer of the Pancreas Cancer of the Lung Cancer of the Prostate Cancer of the Kidney Cancer of the Bladder Cancer of the Skin Cancer of the Brain Lymphosarcoma & 1 2.7 3 2.4 9 12.7 0 4.2 0 0.7 0 1.o 1 0.3 2 0.6 Reticulosarcoma Hodgkin's Disease Leukemia Cancer of Other Lymphatic Tissue Benign Neoplasms 0 0.6 0 0.4 D 11 - 2 1.1 2 0.6 - Endocrine, Metabolic Diseases 5 5.4 92 All Diseases of the Nervous System 21 27.9 75 All Vascular Lesions of CNS 17 25.6 67 All Diseases of the Circulatory System Arteriosclerotic 66 89.7 74 Heart Disease All Respiratory Diseases Emphysema All Diseases of the Digestive System Cirrhosis of Liver All Diseases of G-U System 48 56.9 84 6 13.8 43 3 1.6 - -1 12.6 0 6.5 - 6 7.2 84 All External Causes of Death 27 31.9 85 All Accidents Unknown 15 18.3 82 4 -LCL 60 3 47 38 15 0 25 32 0 0 0 5 38 0 0 n 21 36 30 47 39 57 62 16 37 0 0 31 56 46 -U-CL 81 86 99 137 221 207 362 135 87 537 382 2,229 1,216 666 997 332 677 1,148 215 115 106 94 112 94 539 44 56 182 123 135 \ 21 TABLE 8 SMRs FOR SELECTED CAUSES OF DEATH ALL FEMALES, 1947-1977 Cause ~ Observed Expected All Causes of Death All Infective & Parasitic Diseases All Malignant Neoplasms Cancer of the Digestive Organs Cancer of the Stomach Cancer of the Large Intestine Cancer of the Pancreas Cancer of the Lung Cancer of the Breast Cancer of the Kidney Cancer of the Bladder Cancer of the Skin Cancer of the Brain Lymphosarcoma & Reticulosarcoma Hodgkin's Disease Leukemia Cancer of Other Lymphatic Tissue Benign Neoplasms Endocrine, Metabolic Diseases All Diseases of the Nervous System All Vascular Lesions of CNS All Diseases of the Circulatory System Arteriosclerotic Heart Disease All Respiratory Diseases Emphysema All, Diseases of the . Digestive System Cirrhosis of Liver All Diseases of G-U System All External Causes of Death All Accidents Unknown 77 104.9 1 1.5 24 31.5 1 7.3 1 1.o 0 3.1 0 1.2 5 2.9 4 7.7 0 0.4 1 0.3 0 0.5 1 0.9 0 0.6 0 0.4 2 1.1 1 0.6 0 0.6 0 3.5 9 10.8 9 9.5 22 33.7 16 23.7 3 3.9 0 0.6 3 5.5 0 2.9 3 1.9 7 8.1 4 5.1 4 - 83 95 65 67 - - - 86 - -LCL 58 1 49 0 1 0 0 56 14 0 4 0 1 0 0 21 2 0 0 38 43 41 39 15 0 11 0 32 34 21 UCL 92 375 114 77 547 117 305 406 134 825 1,801 741 599 577 990 676 925 600 105 158 180 99 110 223 666 159 127 460 177 202 22 Ill. ANALYSIS OF MORTALITY BY JOB AND CASE-CONTROL STUDIES OF SPECIFIC CAUSES OF DEATH Barbara J. Divine, Ph.D. Virginia Barron, M.S. Texaco Inc. P.O. Box 37327 Houston, Texas 77237 INTRODUCTION The introductory paper on the Texaco Mortaiity Study (TMS) cohort was concerned with overall patterns of mortality among white male employes. Thi! major result of the study was a large deficit for all the major causes of death. There were a fev4 causes of death where the standardized mortality ratio (SMR) was greater than 100. However. the increases were small. and none were statistically significant. The question still remained whether an increased risk of mortality in a small sub-group of workers was obscured in the analysis of the cohort as a whole. In addition. a further examination of those causes of death with elevated SMRs was desced in order to determine if there were work areas where the decedents had clustered and which were not common to the whole cohort. This paper is an analvsis of the mortality patterns for white males employed in specific jobs in order to determine if any workplace characteristics are associated with an increased risk of mortality. Two different types of studies are used: the first is an analysis of the overall patterns of death for sub-groups employed in specific jobs. and the second is a series of case-control studies of those causes with elevated SMR's. In addition, two other causes of death which have Seen suggested as being associated with employment in the refining industry are included in the case-control studies. METHODS The cohort eligibility requirements and the data collection methods were those described in the previous papers. In addition to the demographic data. available work history data were also used. All persons without work history records had to be excluded from this analysis. Each job held by an employe consisted of a unit (e.g. crude still, maintenance), a task (e.g. operator, pipefitter), and the dates that position was held. Each combination of task and unit title was coded into one of approximately 1.200 exposure categories, based on the American Petroleum Institute (API)Job Code Classification System of Petroleum Refineries and Seiected Petrochemical Operations (1979). Job history coding accounted for all periods of time between the cohort member's initial hire date and final termination date, or the common closing date. Periods of not working for one month or longer were coded. Because of numerous job changes over relatively short periods of time, a method was developed for selecting jobs to be coded on each work history. An employe's initial job was coded. Subsequent jobs to be coded were identified by scanning the work history for the next job that began at least six months later and had a different task or unit code. This process continued until the employe transferred, terminated. or died, or until the study end date was reached. When one or more jobs were held concurrently, they were all coded. If any of three conditions existed, however, a job change was coded, regardless of this 'six-month sample' method. These conditions were: Transfer to another facility. Employment in a task or unit defined in advance. Periods of nonwork due to personal, military, or medical leaves of absence of one month or longer. To obtain estimates of intra-coder reliability and inter-coder validity for the job history coding, a systemaiic sample of employe work histories was selected for duplicate coding by the original coder and for recoding by a randomly selected alternate coder. As noted below, no significant differences between the original coding and the duplicate coding were found. The 28 process units and 40 job tasks were condensed to 18 process groups. based on similar process feed stocks, raw materials, or produced products. and 12 job task groups, according to similar craft and maintenance activities (see 4ppendix I for a Iistinl; of each grouping).These groupings were developed after consultation with industrial hygier,e and plant personnel. 23 For the case-control portion of the study, the six case groups consist of all white males with a work history record whose underlying cause of death was: 1) cancer of the pancreas (n = 62), 2) neoplasm o f the brain benign or unspecified (n = 12) and malignant (n = 31); 3) leukemia (n = 48); 4) other cancers of the lympnatic and hematopoietic tissues including Hodgkin's Disease (n = 13), lymphosarcoma and reticulosarcoma (n = 20) and unspecified (n = 24): 5) cancer of the stomach (n = 44); and 6) cancer of the I , kidney (n = 24). One lymphatic cancer case and one stomach cancer case were excluded because their work histories could not be located Each case group was analyzed separately using the same control group I for all analees. The control group was ranaornly selected from the cohort of noncase white males who had a work history record and who had died after the age of 30. The demographic characteristics of the control group were compared with those for the cohort of all white male decedents, and no significant differences were observed. 11 'I :I 1 . Demographic Characteristics: type of facility, year of hire, year of death, age at death; and 'iI I i RESULTS I I Coding Reliability and Validity I It was found that 97.1% of the duplicate coding matched on length of employment and 96.8% of the ran- - domly selected alternate coder records agreed with the original. In only one of the 644 records that were recoded by the same coder was a different decision made as to eligibility for the total cohort. There was no disagreement on eligibility in the 636 records that were recoded by a different coder. Each job was matched on the date started, and agreement on unit code, job code, and plant code was checked. It was found that 8.5% of the job codes on the reliability check and 12.6% on the validity check did not have a corresponding job starting on that date. Because of the complex coding rules regarding multiple jobs, this rate of disagreement is generally satisfactory. The error rates for the plant code are approximately i.79; for the reliability check and 0.8'3~for the validity check. 24 Patterns of Mortality There were 18.798 white males who were eligible fo: inclusion in the cohort and for whom work histories were located. Two hundred eiGhty-six u!hite pales had only demographic data available and, therefore. had to be excluded from this analysis. There were 70.163 task-unii combinations recorded for the total cohort. The average length of time a job was held was 6.2 years giving 434.668 person years of employment. The job with the greatest number of persondiears was pi2efitter in maintenance. The job held the most number of times was yard laborer in maifltenance. although the average length of time spent in the job was short (two years). Table 1 shows the number of white males employed five years or more in specific jobs, the observed and expected numbers of deaths. and the SMRs associated with each. It should be noted that all of the overall SMRs are jess than 100. and most of these deficits are statistically significant. Table 2 shows the number of observed and expected deaths and the SMRs for the four leading causes of death. all rnalignant neoplasms. vascular lesions of the central nervous system (CNS), arteriosclerotic heart disease. and accidents. SMRs are not calculated when there are fewer than five deaths. For malignant neoplasms, there are two job categories where the SMR is greater than 100, cleaners in maintenance and yard laborers in maintenance. For accidents. the SMR is elevated for one job category, boilermakers in maintenance. For vascular lesions of the CNS. there are four job categories where the SMR is greater than loo! the delayed coking unit. lube oil refining, equipment operators, and electricians. For heart disease, there are five categories where the SMR is greater than 100. lube oil refining, the solvent dewaxing unit, the asphalt plant. the delayed coking unit, and the treating unit. In every instance, however, the increased number of deaths is less than eight, and none are statistically significant. Therefore, it is most likely that these elevated SMRs are due to chance and are not due to the workplace environment. Those causes of death which are elevated for the total cohort are also elevated for some of the jobspecific groups. However, it is felt that the case-control analyses provide more information in terms of whether the association is meaningful. Further discussions will. therefore, be limited to the results of the case-control studies. It should be noted in interpreting the results of the case-control studies that several of the following criteria should hold for an association to be considered further; i.e., that it is compatible with a causal association. The results should be statistically significant, i.e. both confidence limits are above one; the results should be consistent for both the ever worked and the longest job held categories; the length of employment in that job category should be longer for the cases than for the controls; and it should make sense from a biological point of view. Furthermore, because of the laws of probability, when using the 95% confidence limits, one out of every twenty tests may be expected to show a statistically significant result due to chance alone. Because of the large number of tests done as part of these studies, many of the elevated rate ratios (RR) are likely to be due to chance. Pancreas Cancer Table 3 shows the rate ratios for the pancreas cancer cases and controls. The cases who ever worked in the asphalt plants have a RR of 2.9. The RR is 4.4 i f this position was the longest job held. However, the adjusted average job duration for the cases is two years shorter than the controls for those ever employed in asphalt plants, and 9 years shorter for the cases whose longest job was on this process unit. The cases who ever worked in packaging and shipping ( P S )departments were there for an average of five years longer, but the RR is 1.1. For those cases and controls whose longest employment was in P&S. the RR is 2.7, but the standardized job duration is roughly the same. None of the specific job tasks have rate ratios statistically different from 1.0 except for the manager category with RR = 1.a. but the average job duration is shorter for the pancreas cases. Any exposure relating to disease might have occurred in jobs heid previous to this job. but is unlikely since no consistent patterns for other jobs ever held are seen. Regardless of whether an employe was ever employed or held his longest job as manager in the maintenance department. the R R is elevated, 2.2 and 2.9. respectively. ivith roughly equivalent durations of employment betiween ?he pancreas cases and the controls. These managers pro- bably worked in maintenance activities for m@stof their employment: however. the RR for all jobs in maintenance is only 1.2. The RR for controlman as the longest job held in the P&S departments is 3.2, and the cases worked an average of four years longer. No other rate ratios are significant at the 5% level, and no job duration patterns are observed. It appears most likely that any elevaiions in pancreas cancer are due to some unknown lifestyle factors and not to the workplace. Other factors which have been associated with pancreas cancer include coffee drinking, alcohol use. and smoking. Brain Neoplasms- Malignant, Benign, and Unspecified Some importatt demographic differences between the cases and the controls are seen due to the short survival time and the younger ages which are generally affected by these types of tumors. The cases are more likely to have begun employment after the 1940s (51O/O vs 23%), and to have died before age 65 (80% vs 46%). Because of the early age at onset and the short survival time for brain tumors, most persons who were employed very early. i.e. before 1940, and developed brain tumors probably did not survive long enough to be included in the study. Although the time employed may have overlapped with many others in the cohort, anyone dying or terminated before January 1, 1947, was not eligible for the study. This cohort will be followed further in order to determine if this difference in time first employed diminishes in later years. Table 4 shows the rate ratios associated with the various job categories. For those ever having been employed in the quality control laboratories, or for whom this was the longest job held, the RR is elevated, 3.0 and 4.1, respectively. The adjusted average job duration is five years longer for the cases than the controls ever employed in these laboratories, but the same for the longest job held. If employment in the research laboratories is also included. the RR remains elevated at 3.4 and 3.9 for ever worked and longest job held, respectively. However, the difference in the average duration is much less. While no specific job task in the laboratory shows significant excess risk, the RRs for all laboratory job tasks are 2.9 for the ever employed category and 2.7 for the longest job held. The adjusted job duration is slightly less for the cases than the controls. The RR for ever being employed in the paraffin, refrigeration, and ammonia plant process is 3.4, with a two-year average job duration for the cases and over six years for the controls. None of the cases held this job as their longest job. When employment at the filter plant, paraffin, lube oil extraction, and solvent dewaxing units are combined into lube oil refining processes, the RR is 3.6 for ever working on these units and 4.1 for the longest employment. The adjusted average job duration is shorter for the cases than the controls. the difference being approximately a year for those ever employed in these process units to over seven years for the longest employment. Therefore, none of these results are likely to be meaningful. The RR for longest job in the filter plant is 7.3. However, only two cases and three controls were so employed on this unit with the cases working an average of 16 years longer than the controls. The confidence limits are extremely wide, and this result is most likely due to chance. While the welder job task has a RR of 2.6 for ever employed and 3.7 for longest job held, no clear job duration trend exists. The RR for ever being employed in general maintenance tasks is 1.1 with an average duration of five years for the cases and three years for the controls: however, if these tasks were the longest job held, the RR is 3.2. Here, the cases have a shorter average job duration than the controls. The only association with brain tumors which meets the criteria outlined above is the association with employment in the quality control laboratories and to a lesser extent with employment in both quality control and research laboratories. If there is a meaningful association, however, there is still no indication of what specific chemicals might be involved. Furthermore, most of these decedents were first employed 3 0 or more years ago. Laboratory work practices have changed dramatically over the past 20 years through automat!on of test procedures and advances in the design of laboratory ventilation. Future studies of this group v.,ill be carried out to determine i f the association disappears over time as anticipated. Leukemia Tabls 5 sriaws the rate ratios for the leukemia cases and controls. The only process units which show RRs signifi:s.I!ly different from 7.0 are utilities and lube oil refining. Light ends processes have a RR of 26 2.3 for ever employed. but the 95Oh confidence intervals are 0.8 to 6.1, and no job duration trend is indicated. The RR for the asphalt plant is 2.6 and is on the borderline of being statistically significant. However, the controls were employed six years longer than the cases, and for no cases was this the longest job held. All other process units have RRs near or below 1.0. The controlmen in utilities show a RR of 2.8 for ever worked and 5.1 for longest job held, but the average job duration for the longest job held is seven years shorter for the cases than the controls and roughly equivalenrfor ever employed. The asphalt plant has a RR of 2.6, but the cases worked only half the time of the controls on this unit. No cases worked their longest job in the asphalt plant. Ever working in lube oil refining processes shows a RR of 3.2. but for only one case was the longest employment on this unit. Again. the cases worked a much shorter average time on this unit than the controls. The only job task with a significantly elevated FIR is longest job as a pipefitter in maintenance and yard activities (RR = 2.8), but the standardized average employment duration is nearly the same. Thus, it is unlikely that employment in any of the jobs is associated with leukemia. Although other stildies have associated benzene exposure with an excess risk of leukemia, it is striking that no excess risk is found in those job categories where there may be potential exposure to benzene (product handling, catalytic reforming units, petrochemicals, laboratories, and maintenance). Lymphatic Cancer, Excluding Leukemia Significant differences are observed for several demographic characteristics. Even with small numbers. a higher proportion of the cases thar; the controls were last employed in petrochemical facilities (5% vs 1O/O). The cases were more likely to have begun employment after 1940 (39% vs 23%) and they died, on the average, five years younger. By adjusting the average duration, using the controls as a standard, some of these factors have been taken into account. Table 6 shows the rate ratios for the lymphatic cancer cases and controls. Most of the process units show no association except for those ever employed in the fluid catalytic cracking unit (FCCU). While the R R is 3.1, the adjusted duration for the cases is equivalent to the controls, showing no length of employment effect. For those whose longest employment was on the FCCU, the R R is 1.8 (0.6-5.5) with the adjusted duration for the cases being four years longer than the controls. Controlmen who ever worked on the FCCU have a RR of 3.0, and controlmen whose longest job was in the Receipt. Storage, and Movement department have a RR of 2.5, but the average job durations are roughly equivalent. When looking at the longest job as controlman on the FCCU or ever employed as controlman in RS&M, the R R is not statistically significant, but the cases worked an average of four years longer than the controls. Again, these associations are not consistent with a causal association. Stomach Cancer Even though the SMR for stomach cancer in the white male cohort is significantly low, some workplace associations are statistically significant (see Table 7). The stomach cancer cases are more likely than the controls to have ever worked in maintenance and yard activities, specifically yard laborer and cleaner job tasks, but the average duration in those jobs for cases and controls is nearly the same. When considering the longest employment in these units and tasks, the RR is not statistically different from one, but the number of cases is small. The lube oil refining process units have a RR of 5.1 for those ever employed, but the average job dura- tion is only half as long for the cases as the controls. The RR for longest employment on these units is 1.3 (0.2-10.5),based on only one case. Longest employment in the gas plant has a RR of 5.1, but this is based on two cases and four controls. The other job task ever held with an elevated RR is metal working. but the cases again have a shorter job duration than the controls. Only two cases were employed in this task the longest. and the RR is 1.0 (0.2-4.3).Examination of the structural steel task group shows a RR of 1.6 (0.9-2.9) for those ever employed in these jobs an3 1.9 for the longest job held. The average job durations between the cases and controls are not substantially different. 27 Because of the lack of any apparent timetrend, i.e. jobs where the RR i s high and the cases employed longer than the controls. these associations are not likely to be important. Kidnev Cancer All of the RRs calculated for process units and job tasks are below or near 1.0 (see Table 8) with the exception of ever being employed as manager in the maintenance department (2.5) with an adjusted average duration of nearly 15 years for the cases and 10 years for the controls. The rate ratio is 5.4 for this category when exaflning the longest job held. but the average years worked are roughly equivalent. Kidney cancer cases whose longest job was yard laborer in maintenance have a RR of 5.1, but worked an average of six years less than the comparisons. Although toxicology studies in animals have associated an excess risk of kidney cancer with exposure to gasoline and other middle distillates, such an association was not seen here, i.e. no associations were found with those jcb categories where there may be potential exposures to gasoline and middle distillates (product handling, catalytic reforming, thermal cracking, light ends, catalytic cracking, treating, and Iabo rator ies). 28 TABLE 1 NUMBER EMPLOYED FIVE YEARS OR MORE IN SPECIFIC JOBS, DEATHS FROM ALL CAUSES AND STANDARDIZED MORTALITY RATIOS, 1947-1977 Total Cohort Managers, Office Personnel All Operating Personnel Product HandIing Crude Stills Catalytic Cracking Lube Oil Refining Solvent Dewaxing Treating Light Ends Asphalt Plant Delayed Coking Research, Quality Control Lab Quality Control Lab Laboratory Staff All Maintenance Personnel Cleaners Electricians Insulators Yard Laborers Equipment Operators Pipef i tters Boilermakers Structural Steel Number 18,798 4.923 7,745 1.209 1,109 1,040 271 106 402 480 314 285 Observed Expected Deaths Deaths 3,984 1,146 1,875 294 431 252 78 26 131 104 121 102 5,307.1 1,672.0 2,454.6 388.0 562.7 321.3 96.9 40.5 165.9 150.6 136.9 107.5 SMR 75 69 76 76 77* 78* 80 64 79 * 69 * 88 95 2.200 930 1,883 8,075 190 816 255 1,074 521 1,325 341 2.931 297 161 255 1,917 71 121 57 299 157 320 128 705 478.5 240.6 418.7 2,477.6 94.3 169.7 88.2 400.6 203.5 414.7 151.O 920.1 62* 67 61 77 * 75 71 65 75 77 77 85 77 p ? 0.05 29 TABLE 2 OBSERVED AND EXPECTED DEATHS AND STANDARDIZED MORTALITY RATIOS FOR SELECTED CAUSES, 1947-1977 WHITE MALES EMPLOYED FIVE YEARS OR MORE IN SPECIFIC JOBS Malignant Neoplasms Vascular Lesions of CNS Obsvd. Expctd. Obsvd. Expctd. JOB Deaths Deaths SMR Deaths Deaths SMR Total Cohort Managers, Office Staff Operating Personnel Product Handling Crude Stills Catalytic Cracking Lube Oil Refining Solvent Dewaxing Treating Light Ends Asphalt Plant Delayed Coking Research, Quality Control Lab Quality Control Lab Laboratory Staff Maintenance Personnel Cleaners Electricians Insulators Yard Laborers . . Equipment Operators Pipefitters BoiIermakers Structural Steel p 50.05 762 1,015.7 211 321.7 348 471.O 57 74.8 64 107.3 44 63.5 18 19.1 4 8.0 20 32.4 16 30.3 16 26.2 16 20.6 75 66 74 76 60* 69 * 9-4 62 53* 61' 78 58 94.5 61 39 47.7 82 52 82.8 63 404 474.2 85' 21 18.0 117 25 33.7 74 11 16.9 65 80 75.8 105 31 38.6 80 70 79.7 88 21 28.5 74 155 176.7 88 321 403.0 92 133.5 152 187.9 20 29.3 37 47.8 19 23.1 7 6.8 2 2.8 11 12.7 5 10.4 7 10.9 14 8.6 22 29.6 14 15.5 19 25.8 169 190.1 6 8.1 11 10.9 6 6.8 29 32.9 19 16.9 23 31.8 10 13.0 51 70.2 80 69 81 68 77 82 103 - 86 48 64 163 74 91 74 89 74 101 89 88 113 72 77 73 30 TABLE 2-continued OBSERVED AND EXPECTED DEATHS AND STANDARDIZED MORTALITY RATIOS FOR SELECTED CAUSES, 1947-1977 WHITE MALES EMPLOYED FIVE YEARS OR MORE IN SPECIFIC JOBS JOB Total Cohort Managers, Office Staff Operating Personnel Product HandIing Crude Stills Catalytic Cracking Lube Oil Refining Solvent Dewaxing Treating Light Ends Asphalt Plant Delayed Coking Research, Quality Control Lab Quality Control Lab Laboratory Staff Maintenance Personnel Cleaners Electricians InsuI ator,s Yard Laborers Equipment Operators Pipef itters Boilermakers Structural Steel ' p 20.05 Heart Disease Obsvd. Expctd. Deaths Deaths SMR Accidents Obsvd. Expctd. Deaths Deaths 1!608 474 768 131 196 104 37 17 71 38 54 45 2,001.7 642.3 930.4 147.3 218.7 122.8 36.8 15.4 63.8 57.9 52.3 41.2 80 74 83 89 90 85 101 110 111 66' 103 109 151 288.7 43 77.1 67 128.4 10 20.6 13 22.2 15 16.7 1 5.1 0 2.2 5 7.7 6 7.7 5 6.4 4 5.0 117 177.0 58 90.1 100 155.0 66 64 65 21 32.7 9 15.0 18 28.5 772 936.2 82* 63 132.2 21 36.5 58 2 3.8 52 63.3 82 6 10.9 17 33.5 51 2 4.6 97 152.3 64 * 4 19.3 67 77.7 86 3 9.5 133 157.5 84 11 21.7 56 57.9 97 7 6.4 308 348.7 88* 17 48.6 SMR 52. 56* 52 49 * 58 90 - 65 78 78 64 60 .63* 48 5-5 - 51 110 35 31 Ux U cC d V 5 U ? u) w V tmo -0 U m J *I -0 m Imd 1 h W Io m 3V Io 4 0 U cC 0 Id cO w -0 0 W m OD E mW m V &I 0 w Io td Id nQ) 5 LI 1m r Id d u4 .I m Id W h e I+ * -x Nm h 0 M U m ,V r, W cJ Y L L oc 8 a > i 0 wE 0 fll I. wa>l SI x, h m UJ r. N d N, I I m N > U > tr E n I: P L E 0 E i 7. : U X aJ e, 4C P n ?I l4 V Lo al TI mm a La,I Lo 1 .m Pm u) lm4 h m a V u) 4 0 Uc U k w ac U P) m Lo a, m LI W u) d $4 m P J U Lo l4 d W ..I rn m b c H U .K mm 2 M &I m u 0 anJ 0 M * I IY I1 3Y 1 > I aI 1 3 > 1 d4 34 : J 4 4 r 4 13 4 4 I I 3 I I I I I I I I I AJ IX IU dC a 01 n d l4 U m a! 1 -0 m re -J al U 1 I m a m -0 m , lm4 Cx I' al m 1 c um r m 4 0 U >C 1 CJ l4 >0 u > I -J 0 md U a! m m al Y 1 mm l4 0 u m 3d l4 n0 2 U %m 3 l4 r( W.., m bl m Ix J C w .I .I x 0 M U 01 i U m 4U UW , CI -i; zC M 8 % . 3 f , 4 2 7 . d 3 3 3 7+ -9 f n U n f L u x U C d a D 4 u U m aW rn m Ti U rn 3 v a m wrn PI h PI m V m 4 w0 U c U w 0 W Tei U PI m m PI v1 m b 0 u4 II) d u PI Da e &I mw d w .L m u m x e I4 U* x h 0 M a L1 m U a W 2 Li oc I LmD cI c < r C c C C Cr r C r r3 r4 i P J1 4, Y1 I 4I 1I 1 , ! r \ 1 1 J 4 f4 LJ I I t J U X 0) U C cd U al 2 U 1 u) al rn 10 4 Q W U u) .n W m & m h aJ m fU u) 4 0 Y f bl 0 ly U C CmD u i al u) um) um) bl 0 ly ID 44 b 9) P 5 U 4 u) 3 bl .wd u) bl m 4 axl 1 C H * 2' to m U m u W W 2 I a L LG, 50 -x I m w 4 1m Wa m r( *Lo 4 4z 4 4 mc e?. wc Az 10 40 Lo 5 ew?. Uz 5 5 z 2 M ./ ' N 0 b P- O -h \o Ca I-j rf N.m. 90 NN .* JI .I -1 r-4 h m N d NI .I 0 W a 4 Nm 3 AC Y x U C d 'FI 2 U 10 W '0 m m -0 Y m 5 'r) W m bml x 0m1 m V m ri 0 Y C U $4 0 u a C u a m m 01 m bl 0 u-4 ra d l-8 sJ2 G U m l-8 .rl IC( ..I m l-8 m x 3C c U x 0 M al U m 3 ;W " L 00 c DISCUSSION Texaco Mortality Study covers employment at all Texaco locations of refining, petrochemical. and research facilities in operation during 1947 to 1977. For those white males with work histories available, two different types of analyses, cohort studies by job and case-control studies of specific causes of death, were used to determine if any workplace characteristics were more common among those persons dying from certain causes of death than in the rest of the cohort. In intertreting the work history variables, an understanding of worker movement within the workplace is necessary. Typically, most new employes begin in general yard and maintenance jobs, working there for only a short period of the time, and then moving on t o controlman on a specific process unit or specific craft jobs for the remainder of their active employment. Although an attempt was made to group the job categories by potential exposures and responsibilities, in many instances limited information is available on the potential chemical exposures before the 1970s. Most jobs in a refinery involve exposures to a wide variety of straight chain and aromatic hydrocarbons. Also, most of the decedents in this study began work before 1940 andsome as early as 1903. Some 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 t o specific chemicals. Several limitations are present in this study. No information on lifestyle factors, such as smoking, was available and could not be controlled. Two of the causes of death of interest, pancreas cancer and kidney cancer, have been associated with smoking. However, the SMRs for cancer of the lung and diseases of the respiratory system were very low for the totalcohort. No estimates of exposure to specific agents in the workplace are available; therefore, no dose-response trends could be examined, and only length of employ- ment trends are considered in the case-control studies. Possible differences in calendar time between the case groups and the controls are not taken into account. Some exposures encountered during the 1940s and before are likely to be different from exposures in the 1970s. The latency period in the younger and more recent employees may not be adequate to determine if the trends are changing over time. Many comparisons were performed on these data, and some statistically significant results are expected to be due to chance. The major strengths of the study are the large number of persons in the total cohort and the fact that there is a fairly complete characterization of the working lifetime of most of these men for all of the facilities in operation. Follow-up is nearly complete, and the work history records are believed to be accurate. Even with the large numbers in the total cohort, however, the number of persons in many of the job groups is small, particularly in the case-control analyses. The small numbers may preclude the possibility of reaching any definite conclusions. These results suggest several workplace locations which may be associatedwith increased risk of specific causes of death; however, as noted earlier in the text, most of the associations are not supported by a longer employment duration in the cases than the controls. Those associations which should be examined in future studies to determine if the trends continue over time are summarized below. '. Although nearly half of the pancreas cases ever worked in the manager job, the suggestion of an occupational link is not supported by a longer job duration at this time. Seven of the 43 brain tumor cases were ever employed in the quality control laboratories with five cases employed i n this unit the longest. Whether using the criteria of ever worked or longest job held, the rate ratios for this job category are consistently elevated. The brain tumor cases worked longer in the laboratories than the controls; 21 vs. 14 years for ever worked, and 28 vs. 25 years for longest job worked, respectively. The rate ratio remains elevated when research laboratory workers are also included, but the difference in average duration is less pronounced. The R R s for general laboratory job tasks are also elevated, although the adjusted average time spent in those tasks is slightly less for the brain tumor cases than the controls. 38 Although -the r~ate ratios for leukemia are elevated for three Drocess units, utilities, asphalt plants, and lu_ b- e- %i refinina". the averaae V iob I duration is considerably shorter for the cases than the controls. Also. ex- cept for the utilities. the same associations are not seen with longest job held. Twelve of the 58 lymphatic cancer cases were ever employed on fluid catalytic cracking units (RR 3.1), but there is no difference in the length of employment between the cases and the controls. Duration of employment trendsrdo not support any of the excess RRs for stomach cancer. Six of the 24 kidney cancer cases held a job at the manager level in maintenance and worked an average of nearly five years longer than the controls, but the time difference is not apparent in the longest job held analysis. It is possible that any association of workplace exposure and death from these causes may be due to procedures or workplace practices that have been discontinued. In order to evaluate this possibility, the cohort will need to be updated to see if the trends observed in the 1940s and 1950s continue through the 1970s and 1980s. Industrial hygiene measurements of actual workplace exposures to specific agents were begun in the 1970s and became an ongoing activity in the late 1970s. This information can be used in the future to evaluate exposures in any process units or job tasks indicated by future analyses. r 39 ma i # APPENDIX: JOB TASK AND PROCESS UNIT DESCRIPTIONS LA. JOB TASK GROUPINGS 1. MANAGER accountant, administrative assistant, chief chemist, clerical, computer operator, draftsman, employe relations, foreman, manager, nurse, secretary, storekeeper, supervisor-engineer,instrument specialist, profezsional field staff, salaried technical assistant- cook, gateman, nontechnical field staff, timekeeper, watchman-salesworker 2. CONTROLMAN, OPERATOR bargeman, barrel filler, bottomman, car loader and bracer, coke worker, compressorman, controlman, counterman, dockman, fillerman, fireman, gauger (except coke), helper on a unit (except specific craft), hydrotreater, operator (except driver), pumper, rackman, stillman, tank car tester, treater, truck loader, utilityman, water treater 3. LAB chemist, professional lab staff- lab helper, technical lab staff, tester-bottlewasher, nontechnical lab staff 4. CARPENTER, INSULATOR, MASON carpenter-bricklayer, concrete worker, mason- asbestos worker, gunitor, insulator 5. PAINTER painter, stenciler, sandblaster 6. ELECTRICAL electrician-electric instrumentman, instrument-meterman 7. METAL WORK boilermaker, car repairman-tinsmith-acetylene burner, welder-lead burner 8. STRUCTURAL STEEL .agent, repairman, rigger, steel erector, steel repairman-pipefitter, headupman, heat backup- blacksmith, handyman, machinist, sheetmetal worker, tool repairman 9. EQUIPMENT OPERATOR crane operator, heavy equipment operator, rail operator, special equipment operator, truck, tractor, trailer, railway driver 10. CLEANER drum, still, tank, tube cleaner 11. GENERAL MAINTENANCE & YARD construction, general maintenance-recoverylsalvage-boiler, craft helper, general helper (yard), laborer, maintenance, yard job-mechanic-clerk -1, designer, engineer helper, inspector, meterman, technical field staff 12. CRAFTS-MAINTENANCE Groups 4-11 41 I.B. PROCESS UNIT GROUPINGS 1 PRODUCT HANDLING bulk oil department, tank batteries, loading rack, product blending and movement, pipeline, pump house, receipt packing and shipping, RP&S docks, receipt storage and movement, water transport, transfer pumping and gauging 2. CRUDE UNITS crude pressure stills, crude unit, crude stills, desalter, pressure pipe stills, vacuum pipe stills, vertical stills, steam stills, lube stills, pressure stills, badger pipe stills 3. COKING delayed coking, pressure coke stills 4. REFORMING crude recovery, hydrotreaters, catalytic reforming, udex, H-2, platformer- aromatics 5. SULFUR sulfur plant, sulfur recovery 6. CATALYTIC CRACKING cracking, fluid catalytic cracking, recovery area 7. HOLMES-MANLEY PROCESS Homes-Manley vertical still, thermal cracking- visbreaking 8. LIGHT END alkylation, isomerization- polymerization- light hydrocarbons-AOSU, compression plant, gas plant, s t a b i Iizer 9. LUBE OIL REFINING motor oil unit- lube oil extraction (furfural)-lube oil extraction (propane), deasphalting-solvent dewaxing unit- paraffinplant, refrigeration plant- solvent refinishing, dewaxing plant- coal treating plantclay treater plant, clay filter plant, filter plant 10. GREASE grease plant 11. ADDITIVES additive plant .12. ETHYLENE & OLEFINS light olefins unit-ethylene 13. TREATING acid plant, agitation, caustic, engine oil treating, heavy oil agitation, heavy oil treaters, light oil treaters, treating, bulk oil agitators, Linde 14. PACKAGING case and canning, packaging and shipping, case and packing, barrel house-bulk oil, compounding, lube oils-storehouse, warehouse 15. LAB analytical lab, lab, quality control, product control-research labs 42