Document QMkqOV6or1LqwwQj5X3qvk7J4

Occupational and Environmental Medicine Papers 145 152 159 164 167 174 181 -- 191 197 202 206 Correspondence 215 215 215 216 Assessment of occupational exposures in a general population: comparison of different methods Erik Tielemans, Dick Heederik, Akx Burdorf, Roc1 Vermeulm, Hendrik Veulemmrr,Hans Kromhout, Karin Hartog Organochlorinein the serum of inhabitantsliving near an electrochemicalfactory Maria .Sa&, lordi Sunye Raquel otero, Mary sa?ltiago-silvcr,Gnlcs Gmrps, loan Grimuh Modulating influence of cytochrome P450 USpI p o l y m o r p h on serum liver functionprofiles in coke oven workers Mmg-Tmg Wu,Chi-Kvng Ho,Song-Lib Huang, Ya-Fan Yeb, Chia-LingLiu, I-Fang Mao, DrrvidCChristiani Mortality from nephritis and nephrosis in the fibrcgks manufacturingindustry Leonard C h k , Deborah K Watkins,Cheryl Ftyar, WilliamFaymeathq ]oel RBend- Michael Chiaze Update ofthe Texaco mortality study 1947-93: part 1. Analysis of overall patterns of mortality among refining, research,iumi pctrodnmical w& Barbara] Diybrc, ChidneM1Yllrtman,JudyKwkdt Update of the Texlco mortality etart).1947-93: part II. Analps of specific causes of for white men employed in refinin& reseaich, and parocbeaaiclls E u r h u J Diyinc, ChristineM Hartman, ]udy K W e d Mortality patterm among waktm exposed to acrylamide: 1994 follow up Gary M Mar&, h a i n e ] Lwas, Ada 0 Youk, Laam C Schall 20 Yearp of mediclls u r v a on exposureto allergenic and non-allngenic platinum compounds: the importance of c k d speciation Peter] Linnett, E Clyn Hughes Exposure-responserelations of a-amylase sensitisation in British bakeries and flour mills Murk ]Nieuwenbuijscn, Dick Heederik, Gcrt Doekcs, Katherine M V e ~ b k s , Anthony J Newman Taylor Efficacy of measures of hygiene in workers sensitised to acid anhydridesand the influence of sekction bias on the results H Drexler, K-H Scbalh, J Nieketa, A Weber, M Weihraucb, H W e l i k , S Skerfwng Distribution of rest days in 12 hour shift systems: impacts on health, wellbeing, and on shift alertness Philip Tucker, Lawrence Smith, Ian Macdonald, Simon Folkard Mortality of workers exposed to methylenechloride employed at a plant producing cellulose triacetate film base Elsebeth Lynge;]obn A Tomenson, Susan M Bonner, Colin G Heijne, David G Fatrar, Trevor F Cummings Inhalation of ammonium nitrate fuel oil explosive: and possible concomitant exposure A Michael Donoghue Correction Notices fi;...,'i * . ,"*. . j .:,* "* .. +I Update of the Texaco mortality study 1947-93: part I. Analysis of overall patterns of mortality among refining, research, and petrochemical workers Barbara J Divine, Christine M Hartman, Judy K Wendt Texaco, PO Box 1404, Houston, TX 77251, USA B J Divine C M Hartman Universityof Texas, hiD Anderson Cancer Center, 1100 Holcombe Boulevard, Pain Research Group, Box 221, HMB 15.556, Houston, TX 77030, USA J K Wendt Correspondence to: Dr B J Divme, Texaco, PO Box 1404, Houston, TX 77251, USA. Accepted 2 October 1998 Abstract Objective-To update information on the workers of the Texaco mortality study to determine if the patterns of mortality have changed with 16 additional years of follow up. Subjects and nzethods-AU workers were employed for 2 5 years at company refineries, petrochemical plants, and research laboratories &om 1947-93. The cohort now consists of 28 480 employees with an average of 2 2 0 years of follow up. Results-The overall mortality, and most cause specific mortalities were lower than or s i m i l a r to those for the general population of the United States. For white men (86% of the cohort), there were 8873 observed deaths and 11 181 expected resulting in a significantly lower standardised mortality ratio (SMFt) of 79. There were significant deficits for all the leading causes of death in the United States including all cancers, cancer of the lung, stroke, heart disease, respiratory disease, and accidents. Slightly increased mortality was found for cancer of the pancreas, cancer of the brain and central nervous system, leukaemia, and cancer of other lymphatic tissue. For cancer of the bone, the SlMR was 162 (95% confidence interval (950/, CI) 86 to 278), and for benign and unspecified neoplasms, it was 152 (95% CI 109to 206). Overall mortality patterns for non-white men and women were similar to those for white men. Mortality patterns for white men were also examined by duration of employment, time first employed, location, and by job and process unit. There were significantly increased SMRS for brain cancer for those people employed as laboratory workers and on units with motor oil and for cancer of other lymphatic tissue for people employed on the fluid catalytic cracking unit. Concltlsions--The results of the updated study showed a favourable mortality experience for employees in the Texaco mortality study compared with the United States population. There were a few increases found consistently including, but not limited to, brain cancer and cancer of other lymphatic tissue. These increases led to additional analyses that will be discussed in the accompanying paper. (Occup Emiron Med 1999;56 167-173) Keywords: petroleum; chemical industry; occupational cancer; mortality Texaco published two epidemiological reports in 1985 and 1986 on the patterns of mortality at company refineries, petrochemical plants, and research laboratories.' The Texaco mortality study cohort consisted of over 21 000 workers whose mortality experience was followed up from 1947 to 1977. Results fiom these earlier analyses were generally favourable in that the overall mortality and most cause specific mortalities were lower than or similar to those for the general United States population. White men experienced significant deficits for all causes of death combined, all malignant neoplasms, cancer of the lung, stroke, arteriosclerotic heart disease, respiratory disease, and accidents. Increased standardised mortality ratios (SMRs) were found for cancer of the pancreas, cancer of the brain, Hodgkin's disease, leukaemia, cancer of other lymphatic tissue, and benign and unspecified neoplasms, but none was significant. About half of the deaths in the benign and unspecified neoplasm category were from benign and unspecified neoplasms of the brain, and increased SMRs were found for brain tumours in people ever employed in the quality control and research laboratories. This study expands and updates the earlier cohort by adding people who had met the cohort eligibility requirements since the earlier study end date of 31 December 1977, adding current and former employees of three former Getty Oil Company refineries acquired by Texaco as part of its purchase of Getty Oil, and extending vital status follow up to 31 December 1993. The large size of the cohort and the additional 16 years of follow up add substantial power to the original cohort to determine if the earlier observed patterns of mortality have continued. Information on complete work histories for all cohort members made it possible to examine mortality patterns in detail and to compare the results to other studies of refinery workers which have reported increased mortality from leukaemia, malignant melanoma, benign and unspecified neoplasms, kidney cancer, and mesothelioma.'W The cohort includes employees from 15 refineries (all but seven of which have now been sold or closed), two former packaging facilities, five chemical plants (all of which have been 1 168 \ 4 Divine, Harnnan, W h i t sold or closed), and three research laboratories. The factories are in 11 states, began operations from 1903 to 1967, and range in size from 140 to 10 060 former and current employees. Materials and methods The original study cohort included all employees of Texaco who worked at selected refinery, petrochemical, and research establishments at least one day between 1 January 1947 and 31 December 1977; were employed at these for a cumulative total of 3 5 years; and were employed there on either their last day of employment or the original study end date.' Employees of other departments at these factories, employees who transferred to nonparticipating locations before the study end date, and employees at the former Jefferson Chemical Company who ended employment before 1955 were excluded. Because the duration of employment requirement was longer than usual for a cohort mortality study, information on people employed for 3 1 year was collected for the largest refinery in the study and for the three former Getty refineries. Comparisons of the patterns of mortality for the 1 year versus the 5 year cohorts showed no major differences. Original study participants were identified from personnel separation rosters listing all employeeswho left the company between 1947 and 1978 and from computerised personnel files of workers employed in 1978. One hundred and seventeen men and eight women listed on the separation rosters for whom no work history could be found were included in the overall analyses, but not in the job specific analyses. The current cohort consists of members of the original cohort as well as additional people employed at the Texaco mortality study factories who met the eligibility criteria by 31 December 1993. New cohort members were identified with computerised personnel information as well as by reviewing records of employees at the three Getty refineries that Texaco acquired in 1985. Data collected on each employee included name, social security number, race, sex, date of birth, and a complete history (if available) of all jobs held at any of the participating factories. Updated work history and employment status information on people employed in 1978 was obtained from the computerised personnel system as was information on new cohort members. Original records were requested if parts of the work history had not been computerised. People (1%) whose records lacked information about sex or race were assumed to be white men for the analysis. Vital status information to 31 December 1993 for former employees was obtained from Texaco files where available, the National Death Index, the Social Security Administration master beneficiary record file, and the Health Care Financing Administration. People identified as alive by the Social Security Administration in 1985 were assumed to be alive if no matching death record was found by the National Death Index. Copies of death certificates were obtained from company files or from the health departments in the states where the deaths occurred. The deaths were coded by a trained nosologist to the eighth revision of the international classification of diseases (ICD-8) as the eighth revision is the one used by the analysis program. Analyses of patterns of mortality were performed with the program developed by Monson.' The program uses the observed and expected numbers of deaths for specific causes to calculate SMRS, with the United States population as the comparison group. Ninety five per cent confidence intervals (95% CIS) were calculated assuming a Poisson distribution for the observed frequency in the numera- tor of the S M R . Person-years of observation were counted from the date the person attained 5 years of employment or the date the study began, whichever came last, until the study end date, the date of death, or the date lost to follow up, whichever came first. For subcohort analyses, person-years of observation were counted from the date the person attained the required duration of employment in that group or the date the study began, whichever came last. Total mortality was examined by race and sex. Because of the few non-white women (OB%), patterns of mortality for all women were examined with the mortalities for United States white women as the comparison. For white men, patterns of mortality were examined by duration of employment, by time first employed, by plant, by job and unit, and by grouped jobs and units (based on the potential for similar tasks and exposures) by duration of employment in the job or unit. For the analysis of mortality by date first employed, the year 1950was chosen as a dividing date between the early years of the petroleum industry and the later years that tended to have more complex process units and increasing production. Also, about half of the cohort was first employed before 1950, and half after 1950. People who had numerous jobs could be counted in more than one job and unit analysis. See table 1 for a brief description of the potential exposures associated with some of the jobs and units examined. Results The study cohort included 28480 people (table 2) with 738 454 total person-years of observation. Over 92% of the cohort members were men, and of these, over 93% were white. Over 85% of the women were white. Because the cohort overwhelmingly consisted of white men (86.4%), the following descriptive statistics refer only to that group. Thirty six per cent of the cohort was dead, only 2.5% were lost to follow up, and death certificates were obtained for all but 1.9% of the deaths. As of 31 December 1993, only 19% of the cohort was still employed, 44.6% had retired, and 2.5% had left because of a permanent disability. Thirty eight per cent were employed by the company at study locations for 330 years. Forty per cent of the cohort were born before 1920, and a quarter were first employed before 1940. Sixty seven per cent of the dead employ- i Table 1 Texaco mortality study:potenrial exposures byjob and process unitgroup d yob or process unit Desoiprion (pornrialfor exposure) 'Office, professional,and supervisor Operators Craft, maintenance Laboratory staff Fluid catalytic cracking unit (FCCU) Pipefitters and boilermakers Crude stills Motor oil units Delayed coking unit (DCU) Receipt, pumping and storage (RPS) Facility background exposure Operate all process units in facility (potentially exposed to all chemicals present at facility) Provide maintenance support; potentially exposed to all chemicals present at facility (certain crafts have potential for exposure to asbestos) Provide quality assurance checks, pilot plant operations, research activities;potentially exposed to all chemicalspresent at facility (no asbestos) Uses mid-distillatesas input, main product is gasoline;residual heavy ends give potential for exposure to polynuclear aromatics (PNAs) (no benzene) Could work in shop or any unit in factory; potentially exposed to all chemicals present at factory, including asbestos (under certain circumstances,potential for exposure is high) Uses crude oil as input; produces gasoline, middle distillates, and heavy bottoms such as asphalt (potential for exposure to PNAs) Used to remove wax, solids,and PNAs from motor oils; potential exposure to removed materials, motor oils, and methyl ethyl ketone-toluene (no benzene exposure) Potential for exposure to coke dust, PNAs, and heavy ends (no benzene or asbestos) Potential for high exposure to all factory s u e a m s and products (no asbestos) ees were 265 years at the time of death with primary bone tumours. If the ICD-9 had been half of the deaths occurring since the end date used to classify the underlying causes of death of the previous study, 31December 1977. for the bone cancer death certificatesinstead of Table 3 shows the patterns of mortality for ICD-8, seven of the deaths would not have white and non-white men and for women. For been coded to cancer of the bone. The SMR white men, there were 8873 observed deaths for bone cancer would then have been 75. and -11 181 expected giving a significantly For non-white men, there were 437 observed lower all causes SMR of 79. Significant deficits deaths and 563 expected giving a significantall were also found for the leading causes of death causes SMR of 78. Deficits were also found for in the United States population: all cancers many of the major causes of death-such as all (SMR=81), digestive system cancers cancers (SMR=81), cancer of the digestive sys- (SMR=77), lung cancer (SMR=67), arterio- tem (SMR=90), lung cancer (SMR=82), pros- sclerotic heart disease (ASHD, SMR=82), tate cancer (SMR=52), lymphohaematopoietic stroke (SMR=86), non-malignant respiratory cancer (SMR=36), stroke (SMR=77), and disease (SMR=65), and all external causes non-malignant respiratory disease ( S M R = 5 5 ) . (SMRz62). The observed and expected num- The SMRs >100 were for cancer of .the stom- bers of deaths were similar for leukaemia, can- ach (SMR=119), cancer of the pancreas cer of the pancreas, cancer of the brain and (SMR=l19), diabetes (SMR=137), and cancer central nervous system, and cancer of other of the brain and central nervous system t lymphatic tissue. Standardised mortality ratios (SMRz287). None of these were significant. >lo0 were found for benign and unspecified For women, there were 265 observed deaths neoplasms (SMR=152), and bone cancer and 325 expected giving a significantdeficit for (SMRz162). Only the SMR for benign and the all causes SMR=82. Deficits were also unspecified neoplasms was significantly in- found for all. the leading causes of death creased. including all cancers (SMR=76), digestive sys- The increases for the causes of death from tem cancers (SMR=50), colon cancer benign and unspecified neoplasms were exam- (SMR=20), lung cancer (SMR=79), breast ined further. Twenty one ( 5 1%) of the benign cancer (SMR=71), diabetes (SMR=38), and unspecified neoplasms were benign (n=3) ASHD (SMR=75), stroke (SMR=83), non- and unspecified brain tumours (n=18). Be- malignant respiratory disease (SMR=69), cause many of the unspecified brain tumours pneumonia (SMR=25), and all external causes could be malignant, it was decided to combine (SMR=83). Standardised mortality ratios all the brain tumours together. If the deaths >lo0 were found for several cancer sites from benign and unspecified brain tumours including cancer of the stomach (SMR=126), were combined with the malignancies of the leukaemia (SMR=129), uterine cancer brain and central nervous system, there were (SMR=143), and cancer of other lymphatic 85 observed deaths and 75 expected, giving an tissue (SMR=209). None of these was signifi- SMR for brain tumours of 113, which is not cant. significant. All further results apply only to the subco- An examination of the deaths from bone hort of white men and represent only a small cancer showed that 10 of the 13decedents died subset of the total analyses. In general, the after 1978 when the ninth revision of the ICD reported results are for the largest job unit codes (ICD-9) came into effect. In ICD-9, the groups studied or are for subgroups with coding rules were changed so that metastatic results that differed from those expected and bone tumours were no longer classified with led to further cause specific analyses. Table 2 Texaco mwtaliw study:wtal cohort by wital status (n ("A)) Table 4 shows the patterns of mortality by duration of employment. The patterns were Vicol s w w on 31 December 1993 White men Non-whirc m m lPbmm Total similar for all of the groups, and all had significant deficits for all causes of death, all cancers, Aliie Dead UnIulOwn TOM 15115 (61.4) 8873 (36.1) 616 (2.5) 24604 (86.4) 1303 (73.1) 437 (24.5) 43 (2.4) 1783 (6.3) 1691 (80.8) 265 (12.7) 137 (6.5) 2093 (7.3) 18109 (63.6) 9575 (33.6) 796 (2.8) 28480 lung cancer, ASHD, respiratory disease, and all external causes of death. There were significantly increased SMRs for benign and unspeci- fied neoplasms for those employed 2 3 0 years 170 Divine, Harman, W d t Table 3 Texaco mortality srudy: SMRs for selected causes of death, 1947-93 Na-white men n=l783, White men n=24604,py=653857 pv=36552 All W m e n n=2093,p-.?v=48045 Cause of death (ICDA-8) Observed deaths All causes All cancers (140-209) Cancer of digestive system (150-159) Cancer of stomach (151) Cancer of large intestine (153) Cancer of pancreas (157) Cancer of lung (162) Cancer ofbone (170) Cancer of skin (172-173) Cancer of breast (1 74) Cancer of cervix (180) Cancer ofuterus (181) Cancer of prostate (1 85) Cancer of bladder (188) -Cancer of kidney (189) Cancer of brain and CNS (191 132) Lymphatic and haematopoietic cancer (200-209) Lymphosarcoma and retidosarcoma (200) Hodgkin'sdisease (201) Leukaemia (204-207) Other lymphatic tissue (202,203,208) Benign and unspecified neoplasms (210-239) Diabetes mellitus (250) Arteriosclerotic heart disease (410-413) Vascular lesions of CNS (430-438) Non-malignant respiratory disease (460-5 19) Pneumonia (480-486) Cirrhosis of liver (571) AU external causes (800-998) 8873 1975 505 88 184 132 537 13 43 218 63 55 64 226 26 17 93 85 41 128 3209 645 549 226 101 457 SMIP 79 81 77 80 78 105 67 162 98 99 83 94 108 99 75 98 101 109 152 77 82 86 65 74 47 62 9S%CI 78to 81 77 to 84 70 to 84 64to 98 68to 91 88 to 125 61 to 73 86 to 278 71 to 132 Observed deaths ~~ 437 95 31 9 8 7 31 1 1 87 to 114 64 to 107 71 to 122 83 to 137 87 to 113 49 to 110 57 to 156 81 to 123 87 to 135 109 to 206 65 to 90 79 to 84 79 to 92 60 to 71 65 to 85 38 to 57 56 to 68 7 2 1 4 3 0 0 0 3 2 14 113 35 19 9 2 61 SMR ~~ 78 81 90 119 98 119 82 296 133 52 100 53 287 36 0 0 0 81 150 137 93 77 55 52 14 , 91 95%CI Observed dearhs 71 to 85 6 5 t o 98 61 to 128 54 to 226 4 2 t o 193 4 8 t o 244 56to 116 4 t o 1650 2 to 742 21 to 107 11 to 360 1 to 293 77 to 735 7 t o 106 010 417 Oto 615 O r 0 122 1 6 t o 237 17to 543 7 4 t o 230 76to 111 5 4 t o 107 33 to 87 24to 98 2 t o 50 7 0 t o 117 265 74 11 3 2 4 13 0 1 15 1 4 2 1 1 11 0 0 4 7 0 3 61 22 14 2 5 14 SMR 82 76 50 126 20 91 79 0 69 71 30 143 195 65 40 131 0 0 129 209 0 38 75 83 69 25 81 83 9S%CI 72to 92 60 to 96 25 to 90 25 to 367 2 t o 74 24to 232 42to 135 0 to1503 1 to 383 40to 118 Oto 167 3 9 t o 367 22 to 706 1 to 360 1 to 221 6 5 t o 234 Oto 309 Oto 618 35to 331 84to 431 0 to 267 8 t o 112 5 7 t o 96 52to 125 38to 116 3 t o 89 26to 188 45 to 139 py=Person-years. (SMRz184) and for cancer of other lymphatic tissue for those employed 20-30 years (SMRz165). The SMR for cancer of other lymphatic tissue for those employed 330 years was not increased. Table 5 shows the patterns of mortality by time first employed, either before or after 1950. The all causes SMR was 82 for the group first employed before 1950 and 64 for the group first employed after 1950. The observed and expected numbers of deaths were similar for cancer of the pancreas, cancer of the skin, can- cer of the brain and central nervous system, leukaemia, and cancer of other lymphatic tissue for the group first employed before 1950. The SMR for benign and unspecified neoplasms was 146 and is of borderline significance. For the group first employed in 1950 and later, there were slightly increased SMRs for cancer of the pancreas and Hodgkin's disease, and the SMR for benign and unspecified neoplasms was 176. Almost all of the analyses by job and process unit showed deficits for all causes of death, all Table 4 Texaco morrality study: SMRr for selected causes of &ah, 1947-93, by duration of employment Cause of & a h (ICDA-8) Whrte men emphyed <IO y n=4017,py=138057 Observed &ah SMR 95%CI White men emproyed 10-19 y n=5372,py=187250 Obserwd deaths SMR 9S%CI White men emproyed 20-29y n=5759,py=157474 Obserwd deaths SMR 95%Cl White men employed S30y n=9456,py=171076 Observed deaths SMR 9S%CI All causes AU cancers (140-209) Cancer of digestive system (I 50-159) Cancer of stomach (151) Cancer of large intestine (1 53) Cancer of pancreas (157) Cancer of lung (162) Cancer of bone (170) Cancerofskin (172-173) Cancer of prosrate (185) Cancer of bladder (188) Cancer of kidney (189) Cancer ofbrain and CNS (191-192) Lymphatic and haematopoieticcancer (2W209) Lymphosarcoma and retidosarcoma (200) Hodgkin'sdisease (201) Leukaemia (204-207) other lymphatictissue (202,203,208) Benign and unspecified neoplasms (210-239) Diabetes mellitus (250) Arteriosclerotic heart disease (41 0 413) Vascular lesions of CNS (430-438) Non-malignant respiratory disease (460-519) Pneumonia (480486) Cirrhosis of liver (571) AI1 external causes (800-998) 567 119 29 6 11 5 30 0 4 8 4 3 8 18 3 3 7 5 2 2 149 39 31 12 5 84 p-y=Person-years, 71 65 to 77 73 60to 87 71 48 to 102 88 32 to 192 77 39 to 139 64 21 to 150 57 38to 81 0 0 to 524 87 23 to 223 83 36 to 164 108 29 to 277 73 15 to 214 125 54 to 246 95 56 to 150 107 22 to 313 111 22 to 325 93 37 to 192 91 29 to 212 83 9 to 301 18 2 to 64 66 56to 77 103 73 to 141 69 47to 89 73 38 to 128 23 8 t o 55 60 48to 74 1263 267 68 10 24 19 63 1 5 23 6 8 14 39 8 5 17 8 9 13 439 74 74 34 20 110 80 76to 85 84 74 to 95 80 6 2 t o 102 66 32to 122 83 53to 124 117 71 to 183 63 48to 80 71 1 to 397 63 20 to 148 108 68to 162 71 26to 155 98 42 to 192 125 68to 210 114 81 to 156 136 58 to 267 116 37 to 271 125 73 to 200 80 34 to 158 196 89 to 371 58 31 to 98 86 78to 94 82 64 to 102 77 6 0 t o 96 92 64 to 128 47 29to 73 58 48 to 70 2324 514 132 28 40 40 142 4 12 52 13 18 16 70 10 4 23 31 6 32 930 I33 149 59 25 104 83 79to 86 4719 84 76to 91 1075 79 66 to 94 276 96 64 to 139 44 69 50 to 94 109 125 89 to 170 68 69 59to 82 302 182 49 to 465 8 108 56 to 189 22 105 78 to 138 135 71 38 to 122 40 117 69 to 185 26 100 57 to 162 26 124 97 to 157 99 109 52 to 200 5 89 24 to 228 5 101 64 to 152 46 165 112 to 234 41 86 31 to 187 24 79 54 to 112 81 93 87 to 99 1691 73 61 to 86 399 75 63 to 88 295 83 63 to 107 121 39 25to 58 51 57 47to 70 159 79 76to 81 80 75 to 85 76 67 to 85 74 54 to 100 81 67to 98 98 76 to 124 68 61 to 76 215 93 to 425 109 68 to 165 97 82 to 115 88 63 to 120 84 55 to 123 100 66 to 147 84 68 to 102 30 10to 70 85 27 to 197 94 69 to 126 94 68 to 128 184 118 to 274 91 72 to 113 77 73to 81 90 82 to 100 59 52to 66 68 56to 81 57 42to 75 69 59to 81 Table 5 Texaco mortality study: SMRs for selected causes of deathby time first employed, 1947-93 Cause of death (ICDA) Whitp men empbyed before 1950 n= I1649, py=383640 Observed deaths SMR 95%CI White men employed 19SOand aftpr n= 1295S,p-y=270302 Observed demhs SMR 9S%CI All causes 7644 82 81 to 84 1229 64 61 to 68 AU cancers (14&209) 1655 84 8010 88 320 68 61 to 76 Cancer of digestive system (150-159) 412 75 6810 83 93 84 68 to 103 Cancer of stomach (151) 75 79 6 2 t o 99 13 84 44 to 143 Cancer of large intestine (153) 152 78 66 to 92 32 80 54 to 112 Cancer of pancreas (1 57) 105 102 83 to 124 27 119 78 to 173 Cancer of lung (1 62) 449 72 65 to 78 88 50 4010 62 Cancer of bone (170) 13 196 104 to 335 0 0 0 to 267 Cancer of skin (172-173) 32 105 72 to 148 I1 82 41 to 147 Cancer of prostate (I 85) 195 100 8610 115 23 98 62 to 147 Cancer of bladder (1 88) 55 83 62 to 108 8 88 38 to 173 Cancer of kidney (189) 43 94 68 to 126 12 94 49 to 164 i CancerofbrainandCNS (191-192) 48 113 83 to 150 16 95 54 to 154 Lymphatic and haematopoietic cancer (200-209) 187 104 90 to 120 39 81 58to 111 Lymphosarcoma and reticulosarcoma (200) 20 70 43 to 109 6 99 36 to 215 Hodgkin'sdisease (201) 12 93 48 to 163 5 111 36 to 258 Leukaemia (204-207) 84 113 90 to 140 9 50 23 to 94 Other lymphatic tissue (202,203,208) 67 113 87 to 143 18 98 58 to 155 Benign and unspecified neoplasms (2 10-239) 32 146 100 to 207 9 176 80 to 335 Diabetes mellitus (250) 116 86 71 to 103 12 42 2210 73 Arterioscleroticbean disease (410-413) 2833 84 81 to 87 376 66 60 to 73 Vascular lesions of CNS (430-438) 596 88 81 to 95 49 65 48 to 86 Non-malignant respiratory disease (46&5 19) 495 68 62 to 74 54 49 37 to 64 F'neumonia (480-486) 206 76 6610 88 20 58 35 to 89 Cirrhosis of liver (571) 79 51 4010 63 22 36 22 to 54 All external causes (800-998) 318 66 59 to 74 139 54 4510 63 p-y=Penon-years. cancer, digestive system cancer, lung cancer, ASHD, stroke, and all external causes of death similar to those found for all white men. Therefore, the following table shows only the results for selected neoplasms for these job subgroups. Table 6 shows the patterns of mortality for those employed in selected job and process groups mentioned below for at least 5 years. For those who were employed in office jobs, as managers and supervisors, or in professional jobs, there were non-significantly increased SMRs for cancer of the pancreas, bone cancer, skin cancer, prostate cancer, bladder cancer, and benign and unspecified neoplasms. For those employed as operators and controlmen, the SMR for cancer of other lymphatic tissue were increased as was the SMR for benign and unspecified neoplasms for those ever employed in maintenance. For those ever employed as laboratory staff, there were increased SMRs for skin cancer, cancer of the brain and central nervous system, and benign and unspecified neoplasms. For people employed on a fluid catalytic cracking unit, there was a significantly increased S M R for all lymphohaematopoietic cancer that partly resulted from a significantly increased SMR for cancer of other lymphatic tissue. Although the SMRs for the other lymphohaematopoietic cancers were also increased, they were not signilicant. The higher SMRs noted for skin cancer, kidney cancer, cancer of the brain and central nervous system, and benign and unspecified neoplasms were not significant. For people employed as pipefitters or boilermakers, there were increased SMRs for bone cancer, kidney cancer, leukaemia, cancer of other lymphatic tissue, and benign and un- Table 6 Texaco mortality snuaj~:SMRsfor selected neoplasms,for selectedjobs and process units, 1947-93 *,pmfeswnal, ruperviror employed > Syearr white males n=7S86, p-y=168489 Observed deaths SMR 9S%CI C a m l m m , oprrarmr employed Sy wgite men ~1~9542, py=247983 Observed deaths SMR 9S%CI Cmfi, maintenanceemployed > Sy white men ~ 7 9 1 0 , py=220713 Obserwd deaths SMR 9S%CI AU cancers (140-209) Cancer of digestive system (150-159) Cancer of stomach ( I 51) Cancer of large intestine (1 53) Cancer of pancreas (157) Cancer of lung (162) Cancer of bone (170) Cancer ofskin (172-173) Cancer of prostate (1 85) Cancer of bladder (188) Cancer of kidney (189) Cancer ofbrain and CNS (191-192) Lymphatic and hematopoietic cancer (200-209) Lymphosarcoma and reticulosarcoma (200) Leukaemia (204-207) Other lymphatic tissue (202,203,208) Benign and unspecified neoplasms (210-239) p-y=Person-years 590 75 6910 81 80 1 81 7510 86 738 166 78 67 to 91 203 75 65 to 86 193 19 54 32 to 84 41 89 64 to 120 36 65 85 6 6 t o 108 63 66 5010 84 62 53 131 9810 171 52 101 75 to 133 56 140 54 45 to 63 223 69 6010 79 206 5 208 67 to486 3 91 18 to 264 4 18 133 7910210 I 1 63 32 to I14 17 82 111 89to 138 98 107 87 to 130 69 24 97 62 to 144 30 95 64 to 135 21 15 80 4510 131 26 110 72 to 161 22 15 84 47 to 138 20 85 52 to 132 22 54 75 57 to 98 93 101 82 to 124 71 8 77 33to 152 10 70 34 to 129 5 23 79 50 to 118 37 98 69 to 135 25 21 82 51 to 126 40 129 92 to 175 30 12 146 7510255 10 91 44 to 168 16 84 7810 91 82 70to 94 90 63 to 125 74 57 to 95 124 94 to 161 72 6310 83 139 37 to 357 111 65 to 177 87 6810 111 77 4 8 t o 1 1 8 105 66 to 159 105 66 to 160 88 6 9 t o l l l 40 1 3 t o 94 76 49 to 112 109 74 to 156 167 95 to 271 L 172 Divine, Hartman, W t Table 6 Cantinued Labommy staff empiuyed > 5 y whiu men n=2478, py=68286 Obsd &Lath. SMR 95%CI 171 67 57to 78 39 60 42to 81 5 50 16to117 25 104 67to 153 6 47 17 to 102 45 51 37to 68 0 0 0 to 504 7 149 60to307 21 97 60to 148 5 69 22to 161 1 16 Oto 89 11 169 84to302 20 84 51 to 129 2 60 7to215 8 85 37to 167 7 80 32 to 165 5 182 59to425 specified neoplasms. Because of increases related to asbestos found in pipefitters in other similar studies, deaths from mesothelioma were reviewed in detail in a further paper. For people employed on the crude stills, the SMR for cancer of other lymphatic tissue is 172. There were also non-significant increases for pancreas cancer and kidney cancer. For people who were employed on the motor oil unit, one of the lube oil extraction units, the clay filter plant, or the paraffin plant, there were non-significantly increased SMRs for stomach cancer, pancreatic cancer, bladder cancer, lymphosarcoma, leukaemia, and benign and unspecified neoplasms. The SMR for cancer of the brain and central nervous system was 314 (95% CI 115 to 684). Discussion The patterns of mortality found in this 1993 update of the Texaco mortality study cohort are generally similar to earlier findings on t h i s cohort.' As found in the 1977 study, mortality experience is more favourable for the cohort than the United States population. The cohort again experienced significantly fewer deaths from the leading causes of death in the United States, all causes combined, all cancers, heart disease, non-malignant respiratory disease, and all external causes. As in the earlier study, increased mortality was found for benign and unspecified neoplasms, and this result is now sigdicant. The increased mortality from cancer of the bone, which was not found in the earlier study, resulted from an ICD revision coding artifact. Several of these tumours were carcinomas and not sarcomas or tumours of bone tissue origin. Coding rules for the ICD-9 were modified to prevent metastatic bone tumours from being coded as primary bone tumours. As stated in the results section, under these rules there was no excess of bone cancer. A similar result was found by Satin et d 6 The overall results of this analysis support the conclusions of the meta-analysis of petroleum industry workers by Wong and Raabe.' They found that the petroleum industry, in general, had a significantly low cancer mor- tality for all cancer sites combined and for cancers of the digestive system, stomach, and lung. These findings are confirmed by the current study. Wong and Raabe found mortality to be similar to that for the United States population for cancers of the skin, brain, pancreas, prostate, and kidney. These cancers show small increases in various subgroups in the current study, but the increases are not consistent, nor are they linked to increasing duration of employment in most of the job or unit specific subgroups. The Wong and Raabe report did suggest that some of the refinery workers, especially those first employed before 1940, might have an increased risk of leukaemia. However, a follow up meta-analysis that looked specifically at leukaemia cell types" showed no increased SMRs for acute myelogenous leukaemia or any of the other leukaemia cell types. The SMR for leukaemia for this cohort is as expected (101) and is only slightly increased (113) for those first employed before 1950. None of the increases for total leukaemia in the current study is consistent among the subgroups studied, nor is any linked to increasing duration of employment in most of the job or unit specific subgroups. The SMRS for brain cancer in laboratory workers have decreased compared with those found in the earlier mortality analysis for this cohort.' The earlier study showed SMRs of 2 10 and 221 for those employed as laboratory workers for 3 1 year or 2 5 years, respectively. For the most recent update, the SMRs for the same groups have declined to 164 and 169. The original finding may reflect a chance cluster which has not continued. Even if the finding is not due to chance, there is no known exposure to chemical or physical agents for laboratory workers in the Texaco mortality study that has been causally associated with brain cancer. Also, laboratory practices have changed dramatically over the past 30 years, and exposure potentials have declined steadily through the automation of test procedures and advances in the design of laboratory ventilation. FCCUpracess unu employed > 5 y whir0 men n=1432, py=34975 Observed && SMR 95%CI Rpq'iner,,hihmaker employed > 5 y whue men ~ 1 9 7 4 , py=53028 Obsd deah SMR 95%CI 141 93 78to 110 204 88 76to101 38 95 67 to 130 48 75 56to 100 4 62 17to159 11 100 50to 180 14 96 53to 161 15 67 37to 110 10 129 62to237 10 83 40to153 32 63 43 to 90 64 85 66to 109 0 0 Ot0801 2 259 29to935 5 191 62to446 4 102 27to260 13 94 50to161 21 97 60to 149 1 21 Oto 119 6 80 29to 175 7 193 77to398 7 127 51 to261 5 140 45 to327 3 56 11 to 165 25 179 116to264 22 104 65to157 3 143 2 9 ~ 0 4 1 7 2 61 7 t o 2 1 9 8 142 61 to280 14 160 88to269 13 264 140to452 3 42 8 t o 1 2 2 3 186 37 to543 6 237 87to516 Crude s& employed >Sywhiumenn=1418, gy=3S270 Observed de& SMR 95%CI 136 69 57to 81 37 65 1 6 t o 90 5 49 16 to 114 7 35 14to 72 13 124 66to212 35 58 41 to 81 0 0 Oto525 1 33 Oto 184 19 89 53to 138 6 83 3 0 t o 180 7 154 62to317 2 51 6 t o 183 16 89 5 1 to 144 2 69 8to250 3 39 8 t o 114 10 172 82to317 2 91 l o t 0 3 2 8 Mom oil units em> 5y white men n-519, p y = I7782 Ob& &ah SMR 9S%CI 86 98 79to 121 24 99 63to 147 6 142 52to 309 6 70 25to 152 6 131 48to 285 19 68 41to 107 1 336 4 t o 1869 1 72 1 to 400 10 117 56to 215 5 170 55to 396 2 98 11 to 352 6 314 l l 5 t o 684 7 88 35to 181 1 79 1 to 437 4 121 33to 310 2 76 9 t o 274 2 208 23to 750 Cancer of other lymphatic tissue is increased unit combinations is not available nor is there in several job unit specific analyses including any sampling data on industrial hygiene cover- workers employed on the fluid catalytic crack- ing the first 30years of the study.Jobs and units ing units, the crude units, and as pipefitters. with similar responsibilities were grouped This cause of death category is composed of together for the analyses as a surrogate for three different types of lymphohaematopoietic exposures, but provide little information to link cancers including lymphoma, multiple my- exposure to outcome. eloma, and polycythemia vera. The conmbu- tion of each of these causes of death to this result will be examined in the accompanying 'Conclusion paper- This paper reports the overall results of the we have no explanation for the significant Texaco mortality study update. The results deficit for mortality &Om leukaemia for those showed a favourable mortality experience for first employed after 1950 ( S m = 5 0 ) . It is refinery, petrochemical, and research employ- unlikely that insufficientlatency is responsible. ees compared with the general United States The SMR for leukaemia is much lower than population. Most causes of death showed those for many of the solid turnours that are either similar mortality or a significant deficit often aSSumed to have a longer k h ~ t i o n among the Texaco employees. There was a period than leukaemia. 21% deficit for all deaths for the total cohort The Texaco mortality study update has sev- which translates into an increased life expect- era1 strengths, especially compared with other ancy of 2.9 years for Texaco refinery, petro- Company dies Of Petrolem workers. The chemical, and research workers compared with cohort is the largest in the petroleum industry United States white men." including over 28 000 people from many o n l y a few increases were seen consistently Places. More than half of the cohort was in several of the analysis subgroups, which were employed in the industry for >25 Years, and its significantly increased, or which have been mortality experience has been studied for 47 found in other studies of petrolem industry Years (1947-93). vital Status is unknown for workers. These include increases for cancer of ~ 3 %for the entire cohort, and the number of f i S S i n g death Certificatesis C2%. N S O , it is the O d Y reported large Study of refinery workers which includes information about the cohort members' completework histories and the only One in which many analyses by both individual other lymphatic tissue in workers employed on the fluid catalydc crackingunit and crude stills, and for benign and unspecified neoplasms and brain cancer in workers employed in &e l a b e ratories and on the u n i t s related to motor oil. Additional analyses of these results as well as jobs and units as well as grouped jobs and units results specific to leukaemia cell types and have been performed. In recent years, it has mortality &om mesothelioma will be discussed become common in the petroleum industry to in a later report. use contract workers for maintenance activities and for the remaining workers to work on mul- tiple crafts or units. However, this practice did 1 Divine BJ, Monplity Banun Vw,hKitaeplan SrDe.fiTneexrya,cpoemnoorctahleitmy sitdu,dpynId. not begin untilthe 1980s, and it is thus unlikely that it would have had an influence on the causes of death of interest which have 320 years of latency. The study does have several limitations. As with other mortality studies, it has the problems associated with diagnoses &om the death certificates for the of death-for research workers.3 Omp Med 1985;27:445-7. 2 Divine BJ, Barron V. Texaco mortality study II. Patterns of mortality among white males by specific job groups.Am J Ind Med 1986;10:371-81. 3 Wong 0,Raabe GK. Critical reviewof cancer epidemiology in p w l e u m indusny employees, with a quantitativemefp- analysis by cancer site.A m J I n d M e d 1989;15:283-310. 4 McCraw DS,JoynerRE, Cole P. Excess leukemia in a refin- ery populati0n.J Omp E n v i m Med 1985;27:22&2. 5 Nelson NA, Van Peneen PFD, Blanchard AG. Mortality in z:r2nt Oil refinery cohort.J OC'W ~ ~ ' r o n 198799: example, diagnostic accuracy and specificity, and comparability of ICD codes over time. Although the cohort is one of the largest stud- ied in the petroleum industry, many ofthe subgroups analysed are small in size, and for many 6 Satin W,Wong 0,Yuan LA, et d.A 50-year mortality follow-up of a large cohon of oil refinery workers in Turps. Envim Med1996;38:492-506. 7 ShallenbergerLG, AcquaveUa JF,Donaleski D. An updated mortality study of workers in three major United States refineries and chemical plants. BrJ Ind Med 1992;49:345- 4. of the causes of death, the number of deaths is ~ ~ ~ ~ $ n; s ~ ~ ~ ~ ~& small making the results inconclusive and diffi- a refinery and peuochemical plant.AmJIndMed 1996;29: cult to Because Of the analyses done, some isolated increases in 89-98. 9 Monson RR.Analysis of relative survival and proportionate mortality. Cornpurer Bwmed Res 1974;7:325-32. SMRs would be expected by chance alone. 10 Wong 0,h a b e GK. Cell-type specific leukemia analyses in a combined cohort of more than 208 000 petroleum work- Although the employee's complete work his- ers in the United States and the United Kingdom. Reg tory was used, information about the 'pecific chemicals associated with each of the job and T&l Pharmacol1995;21:307-21. 1 1 Tsai Sp, Hardy RJ, Wen Cp, The mortality ratio and life expectancy.~rn~~plden1r9i9o2~;13~:~24--31. 174 Occup Emiron Med 1999;56174-180 Update of the Texaco mortality study 1947-93: part 11. Analyses of specific causes of death for white men employed in refining, research, and petrochemicals Barbara J Divine, Christine M Hartman, Judy K Wendt Abstract &om employment at one of the places Objective-To examine patterns of mor- included in the Texaco mortality study. tality for specific causes of death with (Occup Emiron Med 1999;56174-180) increases in the Texaco mortality study to determine if the patterns are related to employment in the petroleum industry. Methods--Mortality patterns by duration Keywords: petroleum industry; mesothelioma; asbestos; non-Hodgkin's lymphoma; multiple myeloma; brain tllIIlOUrS of employment in various job groups were examined for mesothelioma, non- A recent report presented the 1947-93 pat- Hodgkin's lymphoma, multiple myeloma, terns of mortality for the workers of the Texaco cell type specific leukaemia, and brain mortality study employed for 2 5 years at com- hUllOUrS. pany refineries, petrochemical plants, and Results-Mortality from mesothelioma research laboratories.' The study cohort con- 1 \ was examined for the total cohort and for two maintenance groups with the greatest potential for exposure to asbestos. The insulator group had a standardised mortality ratio (SMR) of 3029, and a larger group consisting of insulators, carpenters, labourers, electricians, pipefitters, boilermakers, and welders had an SMR of 411. The mortalities from mesothelioma in- sisted of over 28 840 workers, and the results showed that the overall mortality and most cause specific mortalitieswere lower than those of the general United States population. The patterns of mortality for the job and process units were similar to those for the overall cohort. However, there were a few increases that were found consistently in several of the creased with increasing duration of em- analysis subgroups, were significantly in- ployment. Mortality was lower for those creased, or have been found in other studies of first employed after 1950. An analysis of petroleum industry workers. all brain tumours for the total cohort and Significantincreases for cancer of other lym- some job and unit subgroups resulted in phatic tissue, which includes both lymphoma an SMR of 178 for those employed on the and multiple myeloma, were reported in the units related to motor oil and 166for those earlier study of workers employed on the fluid employed as laboratory workers. Mor- catalytic cracking unit and on the crude stills. tality fkom brain tumours in both of these Non-significant increases were also found for job groups was higher for those employed several other job and unit combinations. Mor- 2 5 years in the group. An analysis of non- tality from non-Hodgkinl lymphoma and Hodgkin's lymphoma showed no consist- multiple myeloma were examined separately to ent patterns among the various reflect more recent classifications of these employment groups. Mortality from mul- diseases. Texaco, PO Box 1404, Houston, TX 77251, USA B J Divine C M Hamnan University of Texas, MD Anderson Cancer Center, 1100 Holcombe Boulevard, Pain Research Group-Box 221, HMB 15.556, tiple myeloma was non-significantly increased among people employed on the crude (SMR=155) and fluid catalytic cracking units (SMR=198). Leukaemia mortality was not increased for the total cohort, and a cell type analysis of leukaemia mortality for the total kohort showed no significant increases for the major cell types. However, there were significant increases for acute unspecified leukaemia (SMRz276) and leukaemia of unknown The total cohort of white men showed a sig- nificant excess for benign and unspecified neoplasms, and the excess was also noted in some of the job and unit analyses, especially for those ever employed in laboratories or on the motor oil units. Twenty one of the 41 deaths in the benign and unspecified neoplasm category were from benign and unspecified neoplasms of the brain. Brain cancer mortality was also increased for those ever employed in laboratories or on the motor oil units. Therefore, Houston, TX 77030, USA J K Wendt cell type (SMR=231). further analyses of all brain tumours (malig- Conclusions-Analyses of specific causes nant, benign, and unspecified combined) by of death by duration of employment in job and unit combinations were carried out. Correspondence to: various job and process units did not show A recent publication by Tsai et a12 showed Dr B J Divine, Texaco, PO Box 1404, Houston, TX 77251, USA. any patterns which suggest that, other that a group of maintenance workers defined than for mesothelioma, any of these by their potential for exposure to asbestos had increases in mortalities were likely to have a significant excess of mesothelioma. There- Accepted 2 October 1998 resulted from workplace exposures or fore, it was decided to examine mortality from mesothelioma for several job groups with the, the recent report by Tsai et a f and are based on potential for exposure to asbestos. the pattern of incidences from the Connecticut An investigation of the mortality for leukae- tumour registry.6 mias of specific cell types was also performed. Mortality from mesothelioma for the total Refinery workers have the potential to be cohort of white men was examined as well as exposed to benzene which has been associated the mortality in two job groups with the great- with an increased risk of acute myelogenous est potential for exposure to asbestos. The first leukaemia at high exposure levels.' Wong and job group consisted of people ever employed as Raabe` recently conducted a meta-analysis of insulators. The second group included people leukaemia by cell type in petroleum industry ever employed as insulators, pipefitters, weld- cohorts but found no significantly increased ers, boilermakers, carpenters, electricians, and risks for any cell type, including acute myelog- labourers, a grouping similar to that reported enous leukaemia. Satin et td found an excess of recently.* The mesothelioma patterns were acute lymphocytic leukaemia in a recent cohort examined for these groups by duration of mortality study at a large refinery next to the employment and by time first employed in the largest factory included in the Texaco mortality job group. study. However, there was no exposure- response relation, and the authors also found a ANALYSIS OF LYMPHOHAEMATOPOIETIC DEATHS large deficit for chronic lymphocytic leukae- The combined categories of lymphohaemat- mia. opoietic cancers in the published United States mortalities prevented analyses of these causes i Methods of death as separate entities. For example, lym- The expected numbers of deaths were obtained phosarcoma and reticulum cell sarcoma com- by multiplying the person-years for each prise one category, whereas multiple myeloma, subgroup analysed by the United States lymphoma, and polycythemia vera are grouped mortalities. Standardised mortality ratios in another category. Haematologists recom- (SMRs) were calculated as the ratio of the mend that lymphosarcoma and reticulum cell observed and expected numbers of deaths for sarcoma should be grouped with lymphoma to specific causes multiplied by 100. The SMRs make a non-Hodgkin's lymphoma category were tested for statistical significance with and that the other causes of lymphohaemat- Poisson based 95% confidence intervals (95% opoieticcancers should be examined separately CIS) assuming a two sided test. The SMRs (Irons R, 1995; personal communication). were calculatedfor those people ever employed Also, all of the leukaemias are grouped together in the job and unit group of interest, and if in the published United States mortalities there were sufficient numbers of deaths, for although the different cell types are different those people who were employed in the job and diseases and can have different aetiologies. unit group for 3 5 years. The SMRS were also Mortalities for the period 1964-92 for mul- calculated by the period in which first employ- tiple myeloma, leukaemia by cell type, and ment occurred. lymphoma were obtained from the mortality and population data system maintained by the ANALYSIS OF DEATHS FROM MESOTHELIOMA University of Pittsburgh Department of Mesothelioma does not have a specific code in Biostatistics.' The rates for the years before the eighth revision of the international classifi- 1964 were assumed to be the same as those for cation of diseases (ICD-8) that can be used to 1965-9. The rates for unspecified leukaemia identify the deaths from this cause. Instead, for the years 1980-92 were assumed to be the depending on the phrasing of the death certifi- same as those for 1975-9. The rates for cate, mesothelioma can be coded as a respira- lymphosarcoma and reticulum cell sarcoma torycancer (ICD-8 codes 162.lY163,163.0,or from Monson's program* were added to the 163.9), a malignant neoplasm without mention lymphoma rates to give mortalities for non- of site (ICD-8 codes 199, 199.0, or 199.1), a Hodgkin's lymphoma. The mortalities for each benign respiratory disease (ICD-8 codes of the causes of death of interest were then 212.3,212.4, or 228), or a malignant neoplasm multiplied by the person-years for the sub- of the peritoneum (ICD-8 code 158). Death groups to give the expected numbers of deaths certificates for the Texaco mortality study for the specific cause of interest. cohort with any of these ICD-8 codes either as the underlying or as a contributory cause of ANALYSIS OF DEATHS FROM BRAIN TUMOURS death were reviewed. Those death certificates Because several of the deaths from benign and with any mention of mesothelioma were unspecified neoplasms in the previous report defined as a case. on this cohort' were categorised as benign and Because mortalities for mesothelioma have unspecified brain tumours, United States mor- not been published, the 1973-80,1985-8, and talities for these two categories of death were 198&9 1 surveillance, epidemiology, and end also obtained from the University of Pittsburgh results programme of the National Cancer mortality and population data system. The Institute mesothelioma incidences were used. rates were added together, and the rates for the As mesothelioma is almost always fatal, the years before 1964were assumed to be the same incidences were assumed to be the same as the as those for 1965-9. These rates were then mortalities. It was assumed that the rates in the multiplied by the person-years for the various 1950s and 1960s were one quarter and one groups of interest. Observed and expected half, respectively,the rates for the 1970s.These numbers of deaths from brain tumours were assumptions were the same as those used for added to those for cancer of the brain and 176 Divine,Harmon, It%?& Table 1 Taaco mortality s d y : SMRs for mesothelioma and lung cancer,white men, 1947-93 Mesothrlioma Obsarpd SMR 95%CI Lung cancer Observed SMR 9S%CI Total cohort Insulators (duration of employment): Ot04y 5to19y 220 y Before 1950 1950 And after Insulators, pipefitters, welders, labourers, boilermakers, electricians,carpenters (duration of employment): Oto4y 5to19y 220 y Before 1950 1950 And after 44 4 1 10 9 2 29 14 7 16 25 4 297 216to 399 57 1 714 9916 4586 1200 411 153 IO 1462 910 3974 447 to 18237 2092 to 8706 134 to 4335 275to 590 239 322 1053 470 230 131 to 129 to 601 to 304to 62 to 402 664 1710 694 591 537 31 10 7 18 5 219 91 75 238 52 67 61 to 73 77 52 to 110 133 63 to 245 128 51 to 264 159 94 to 251 61 20 to 141 68 5910 78 75 60to 92 89 70 to 112 78 69 to 89 61 45 to 80 central nervous system to calculate an overall brain tumour SMR. The SMRS were calcu- lated for the total cohort and for the job and unit groups where increased SMRs for either brain cancer or benign and unspecified neoplasms were found in the overall analysis.' Results MORTALITY ANALYSES OF MESOTHELIOMA There were 44 deaths with any mention of mesothelioma on the death certificate. For all but one of these, mesothelioma was the underlying cause of death. All but eight of the decedents were first employed before 1January 1950, and for all, at least 30 years had elapsed between the time they were first employed and the date of death. The mesothelioma SMR for the total cohort was 297 which was significant. There were 11 deaths from mesothelioma for those employed as insulators for 2 1 year (SMR=3029, 95% CI 1510 to 5420) and 29 deaths for those employed as insulators, pipefitters, welders, labourers, boilermakers, electricians, or carpenters for 21 year (SMR=411,95% CI 275 to 590). Table 1 shows the SMRs for mesothelioma for all white men and for the two exposure groups by duration of employment in the group and by time first employed. Almost all of the SMRs were significantly increased and corresponded with increasing duration of employment in each group. The highest SMRS are found among people employed as insulators for 3 2 0 years (SMRz9916). The SMRs were higher for people first employed in either of the job groups before 1950. Increased SMRs were Table 2 Tarno mortaliv smay:SMRC for brain rumours, white men, 1947-93 Total cohort: Ever Employed before 1950 Employed 1950 and after Maintenance: Ever Laboratory staff: Ever Employed 3 5 y Receiving, pumping, and storage: Ever Pipefitter Ever Motor oil unit: Ever Employed 2 5 y Observed 85 62 23 61 19 15 11 15 11 8 Expad SMR 75 113 54.8 113 20.2 113 55.1 110 11.4 166 8.1 185 10.9 101 14.9 100 6.2 i 178 2.5 326 9556CI 90 to 140 86 to 145 72 to 170 84 to 142 99 to 259 103 to 305 50 to 181 56 to 165 88 to 319 140to643 also found for those people first employed after 1950 and were significant for the group of insulators. Several of the deaths from mesothelioma were originally coded as primary lung cancer deaths (an ICDA-8 code of 162.9 for the underlying cause of death), and increased mortality from lung cancer has also been associated with exposure to asbestos.' Table 1 also shows the SMRs for lung cancer with the mesothelioma deaths with an ICDA-8 of 162.9 excluded for the two exposure groups defined by duration of employment. Only the subcohort of insulators had a mortality that increased slightly with duration of employment in the group. None of these SMRs was significant. For the larger exposure group, the SMRs increase with duration of employment, but all are <loo. There was a non-significant increase for lung cancer for those first employed as insulators before 1950 (SMR=l59) but a large deficit for those first employed in 1950 and after (SMR=60). No information on the smoking histories of the cohort was available. MORTALITY ANALYSES OF BRAIN TUMOURS There were 21 observed deaths and 15.5 expected from benign and unspecified brain tumours, resulting in an SMR of 135 (not significant). There were 64 observed deaths and 59.5 expected from cancer of the brain and central nervous system. Combining all such tumours gave 85 observed deaths and 75 expected, for an SMR of 113 (table 2). Observed and expected deaths from brain tumours were compared for persons employed in maintenance (SMR=llO); in receiving, pumping, and storage (SMRz101); as a pipefitter ( S M R = 100); as laboratory stafF (SMR=166); and on the motor oil units (SMR=178). The total cohort SMRs for those first employed before 1950 and for those first employed in 1950 or later were the same ( S M R = 113).Mortality from brain tumours for people employed either in laboratory jobs or on the motor oil units for 3 5 years were higher than for people ever so employed, and the SMRs were significant (SMRs=185, and 326, respectively). Table 3 Texaco mortality study: SMRs for non-Hodgkin's.lymphomaand muln'ple myehma, white men, 1947-93 Total cohort: Ever Employed before 1950 Employed 1950 and after Maintenance: Ever Operators: Ever Fluid catalytic cracking unit: EVer Employed 2 5 y Receiving, pumpmg and storage: EVa Pipefiners, boilermakers Em Delayed coking unit: Ever Employed 2 5 y Crude stills: Ever Employed 2 5 y Non-Hodgkin'r lymphoma Observed SMR 95%CI 74 88 69 to 111 56 87 65 to 113 18 95 56 to 151 54 87 65to114 37 73 51 to 101 13 155 82 to 265 6 118 43 to 258 14 118 64 to 198 15 72 40 to 119 6 86 31 to 188 16 122 70 to 199 Multiple myeloma Obserwd SMR 95%CI 36 101 70 to 140 30 103 69 to 147 6 91 33 to 199 30 112 76 to 161 17 79 46 to 126 7 198 79 to 408 6 270 90 to 589 6 118 43 to 258 14 153 84 to 258 6 196 71 to 427 1 156 2 to 873 9 155 70 to 294 5 172 55 to 401 MORTALITY ANALYSES OF NON-HODGKIN'S LYMPHOMA Table 3 shows the results of the analysis of deaths from non-Hodgkin's lymphoma. Ob- seen for those first employed before 1950 (SMR=103) and first employed in 1950 and after (SMR=91). For those employed on the fluid catalytic cracking unit, the SMR was 198, served and expected deaths from lymphosar- and it rose to 270 for those employed on the coma, reticulum cell sarcoma, and lymphoma fluid catalytic cracking unit 35 years. For those combined were similar for the total cohort employed on the crude unit, the SMR was 155, (SMRz88) and for those employed in mainte- increasing to 172 for those employed on the nance (SMR=87), in receiving, plumbing, and crude still 2 5 years. Four of the decedents were storage (SMR=118), on a process unit employed on both the fluid catalytic cracking (SMR=73), on the delayed coking unit unit and the crude unit. For those employed as (SMR=86); on the crude unit (SMR=122); a pipefitter or boilermaker, the SMR was 153; i and as a pipefitter or boilermaker (SMR=72). however, there were no deaths from multiple The SMRs are similar for those first employed myeloma among people employed as pipefit- before 1950 (SMR=87) and first employed ters or boilermakers for 2 5 years. For those 1950 and after (SMR=95). For those ever employed on the delayed coking unit the SMR employed on the fluid catalytic cracking unit, was 196, but there was only one death among the SMR was 155, but it decreased to 118 for those employed on the delayed coking unit for those employed on the fluid catalytic cracking 2 5 years. These SMRs for multiple myeloma unit 2 5 years. None of these SMRs was were not significant. significant. MORTALITY ANALYSES OF LEUKAEMIA BY CELL Table 4 Texaco &ity MORTALITY ANALYSES OF MULTIPLE MYELOMA Table 3 also shows the results of the analysis of multiple myeloma. The observed and expected deaths were about the same for the total cohort ( S M R = 101); people employed in maintenance (SMR=112); employed in receiving, plumbing, and storage (SMR=118); and employed on a process unit (SMR-279). Similar results were sncdy, SMRs for leukaemia by cell type, white men, 1947-93 TYPE The SMRs were calculated for the cell type specificleukemias where there are at least four deaths (table 4) for the total cohort and for those first employed before and after 1950. All but one of the cell type groups showed deficits in the group first employed in 1950 and after. The mortality for acute lymphocytic leukaemia was the same as expected (SMR=lOl), and all of the acute lymphocytic leukaemia deaths Total cohort: Acute lymphocytic (ALL) ALL, employed before 1950 Chronic lymphocytic (CLL) CLL, employed before 1950 CLL, employed 1950 and after Acute myelogenous (AML) AML, employed before 1950 AML, employed I950 and afrer Chronic myelogenous (CML) CML, employed before 1950 CML, employed 1950 and after Acute unspecified (AUL) AWL, employed before 1950 AUL, employed 1950 and afrer Cell rype unspecified (UL) WL, employed before 1950 WL, employed 1950 and affer Observed 5 5 15 13 2 20 19 1 12 10 2 15 13 2 15 14 1 E x ~ ~ t e d SMR 4.9 101 3.8 133 18.6 80 15.8 82 2.8 72 15.5 129 12.2 155 3.3 30 11.4 105 8.6 116 2.8 70 5.4 276 4.7 276 0.7 273 6.5 231 5.3 262 1.1 87 95%CI 32 to 235 42 to 311 45 to 133 43 to 140 8 to 262 78 t O 199 93 to 243 0 to 168 54 to 183 55 to 214 7 to 254 154 to 455 147 to 472 30 to 986 129 to 381 143 to 440 1 to 486 occurred in people first employed before 1950 (SMR=133). There was a deficit for chronic lymphocytic leukaemia (SMR=80) with 13 of those deaths occurring in people first employed before 1950 (SMR=82) and two in those employed in 1950 and after (SMR=72). The SMR for acute myelogenous leukaemia was 129.Nineteen of the 20 deaths occurred in people first employed before 1950 (SMR=155), and there was a deficit among people first employed in 1950 and after (SMR=30). Mortality from chronic myelogenous leukaemia was essentially the same as expected (SMR=105). Ten of the 12 deaths occurred in people first employed before 1950 (SMR=l16), and there was a deficit of chronic 178 &vine, Hanman, Wend1 myelogenous leukaemia among people first employed in 1950 and after (SMR=70). There were significant excesses for acute leukaemia, unspecified (SMR=276), and for leukaemia cell type unspecified (SMR=231). For acute leukaemia,unspecified, the SMR was essentially the same among people first employed before 1950 and those first employed in 1950 and after (SMRs=276 and 273 respectively). For leukaemia, cell type unspecified,the SMR was increased for those first employed before 1950 (SMR=262), but for those first employed in 1950 and after, there was a deficit (SMR=87). Discussion There was a significant increase in mortality from mesothelioma for the total cohort which was higher for people employed in maintenance subgroups based on their potential for exposure to asbestos. Although the increase was highest for people employed as insulators, mortality rose in all of the maintenance groups with increasing employment in that group. An examination of mortality by time first employed showed that the SMRs were highest for people first employed in any of the subgroups before 1950. Almost all of the cohort's increased mortality from mesothelioma was limited to the maintenance subgroups, especially insulators. There were only seven deaths from mesothelioma versus 5.3 expected (SMR=130) for those never employed in the maintenance subgroup and only two versus 3.9 (SMR=51) for those never employed in any maintenance job. The finding of excess mortality from mesothelioma in this cohort was not unexpected. Materials containing asbestos were used in the past as thermal insulation for various vessels and pipes in refineries. There were no industrial hygiene sampling data about the levels of exposure to asbestos for the various crafts until the 1 9 7 0 ~an~d by that time work practices for asbestos handling had already been modified. However, insulators would have had the highest potential for exposure, followed by workers in other crafts who had to work with insulation to accomplish their tasks. Several other petroleum industry studies have also shown increased mortalities for mesothelioma in refinery workers.'-' Other studies of refinery workers have not identified excess deaths from mesothelioma. Collingwood et all3 found two deaths from mesothelioma versus 2.3 expected at the Mobil Paulsboro refinery, and Raabe et all' found one death from mesothelioma versus 3.2 expected at the Mobil Beaumont refinery. Although mesothelioma is known to be associated with asbestos exposure in thc shipbuilding industry," it is unlikely that much of the excess found in this cohort was associated with exposures in this industry during the second world war. No information on previous employment was available. However, even if all people who had the potential to be employed in shipbuilding during the second world war were excluded, the SMRs are still significantly increased. Because of the lack of mortalities for mesothelioma and of a specific ICD code for this cause of death, the comparability of the mesothelioma SMRs among all these studies is unknown other than with the study by Tsai et al. Our comparison mortalities were the same as those used in that study as was the definition of the maintenance subgroup. It is unknown why some of the refinery studies show lower than expected SMFb for mesotheliomas. No increase in mortality from lung cancer was found in the larger maintenance subgroup. There was, however, a small increase in lung cancer mortality for those employed as insulators which rose with increasing duration of employment. The increases were small and not significant. There was a deficit of lung cancer mortality among those first employed in 1950 and after for both groups. No histories on smoking were available to indicate whether these increases were primarily found in smokers. The SMR for all brain tumours for the total cohort (1 13) was similar to that reported for cancer of the brain and central nervous system (108) in the earlier report.' The highest SMRs for all brain tumours combined were 178 for people ever employed on any of the motor oil units and 326 for people employed on these units for 35 years.These results were similarto those for cancer of the brain and central nerv- ous system where the SMRs were 164 and 314, respectively. An SMR for brain tumoursof 166 was found for people ever employed in research or quality control laboratories and it was 185 for people employed as laboratory staff for 2 5 years. Again, these SMRs were similar to those for cancer of the brain and central nervous system where the SMRs were 142 and 169, respectively. Thus, other than increasing the number of deaths, combining all brain tumours together did not change the pattern of results found for the category of cancer of the brain and central nervous system alone. There were no specific chemical exposures that might be common to the motor oil unit and the laboratory, and that have been associated with brain tumours. In any event, laboratory practices have changed dramatically over the past 30 years and exposure potentials have declined steadily through the automation of test procedures and advances in the design of laboratory ventilation. Similar decreases in exposure levels have also been achieved on operating units through the use of engineering controls and enhanced work practices. With Texas death rates as the comparison, Satin et a t found an increase in benign and unspecified brain tumours at the GulEIChevron Port Arthur refinery. The SMR was much lower when United States death rates were used. A previous report on brain tumours in the Port Arthur refinery ~ o h o r td'i~scussed the problems associated with diagnostic errors of brain tumours. Errors occur more often among inaccessible tumours-such as those occurring in the liver, brain, or pancreas. A potential for more accurate and complete reporting of brain tumours in occupational groups as a result of diagnostic sensitivity bias has also been dis- i cussed in several previous investigations. \ tinational cohort of over 250 000 petroleum Diagnostic sensitivity bias is one reasonable workers. The SMR for multiple myeloma was explanation for the increased SMR of 113 for 93, and there was no pattern by duration of brain tumours for the overall Texaco mortality observation. study cohort. The analysis of leukaemia by cell type The causes of malignant and benign brain showed a non-significantly increased SMR for tumours are still largely U ~ ~ I I O W W~ .i'th~in the acute myelogenous leukaemia for white men general population of the United States, the overall, which was limited to those people first recorded cases of, and deaths from, cancers of employed before 1950. Because many of the the brain and central nervous system have been deaths from leukaemia do not have a specific rising slightly since the 1 9 7 0 ~ .S' ~ome or all of cell type mentioned on the death certificate, these increases are thought to result from the this resulted in significant excesses of both advent of computed tomography, magnetic acute leukaemia and leukaemia, cell type resonance imaging, and other techniques that unspecified. More detailed information about have led to improvements in the detection of the specific cell type of leukaemia would reduce previously undiagnosed or misdiagnosed cases the increases in the unspecified categories, and of brain tumours. increase the SMRs in some or all of the catego- The few established risk factors for brain ries specific for cell type. However, although tumours account for only a small proportion of death certificates are considered to be highly 1 cases. Genetics plays a part," and ionising accurate in their diagnosis of leukaemia,26the radiation to the head is a strong, but rare, risk accuracy of the information specific to cell type ''factor." The only occupational exposure is much lower. It would be better to have access linked to increased risk of brain cancer is expo- to hospital and pathological information, but sure to high concentrations of vinyl ~hloride,'~that is not possible in a study of this type. A which was not present at any of the Texaco meta-analysis of leukaemia by cell type in mortality study locations. For dozens of workers in the petroleum industry4 did not occupational groups, slight, inconsistent excess show increased SMRs for acute myelogenous risks of brain cancer have been reported by leukaemia or any of the other leukaemia cell various investigators, but no convincing asso- types. ciations, certainly not specific causative agents, Although employees' complete work histo- have been identified.lg ries were used for these cause specific analyses, The analysis of the SMRs for non-Hodgkin's information about the specific chemicals asso- lymphoma for the total cohort and for the job ciated with each of the job and unit combina- and process unit groups shows primarily tions was not available, nor was there any deficits and slight increases. The only job industrial hygiene sampling data covering the group with a noticeable increase for non- first 30 years of the study. (For information Hodgkin's lymphoma is the fluid catalytic about potential exposures associated with the cracking unit (SMR=155), but the SMR jobs and units, see table 1 in the previous decreases to 118 for those employed on the report'.) Jobs and units with similar responsi- fluid catalytic cracking unit for 2 5 years. Thus, bilities were grouped together for the analyses the increased SMRs found in the previous as a surrogate for exposures, but provided little report' for cancer of other lymphatic tissue information to link exposure and outcome. among people employed on the fluid catalytic Patterns of mortality were examined by time crackingunit or the crude stills were not related first employed to determine if the patterns had to increased mortality from non-Hodgkin's changed over time. Although many groups lymphoma. showed much lower SMRs for those more However, the S M R for multiple myeloma for recently employed, in many cases the number people ever employed on the fluid catalytic of observed deaths was so small that the results cracking unit was 198, and it increased to 270 were difficult to interpret. for those people with 2 5 years on this unit. For people ever employed on a crude unit, the Conclusion SMR was 155, increasing to 172 for those peo- Further analyses of specific causes of death by ple with 2 5 years on this unit. The increases duration of employment in various job and for the gro.ups employed for 2 5 years in the job process units did not show any patterns which are based on only six and five deaths, would suggest that, other than for mesothe- respectively, and some of these were employed lioma, any of these increases in mortalities were on both units. likely to result from either workplace exposures The SMR for multiple myeloma among or employment at one of the locations included those employed on the delayed coking unit for in the Texaco mortality study. The brain 3 5 years was lower than for the 1 year group, tumour rates in people employed in the motor and there were no deaths from multiple oil units and the laboratories and the rates for myeloma in the group of pipefitters and boiler- multiple myeloma in those employed on the makers employed for 2 5 years. There were no fluid catalytic cracking units rose with increas- specific exposures associated with employment ing duration of employment in these jobs. on the fluid catalytic cracking unit or crude However, there was no known exposure stills which have been associated with multiple common to these jobs which had been myeloma. Also, there was no evidence that associated with brain cancer or multiple multiple myeloma was increased in petroleum myeloma, nor has an increased risk of these workers overall. Wong and Raabe" carried out causes of death been consistently found in a rneta-analysis of multiple myeloma in a mul- other studies of petroleum workers. There was 180 Divine,Harzman, UZ?ndt only a slight and non-significant increase in the SMR for acute myelogenous leukaemia which is the only leukaemia cell type associated with benzene exposure. The increase in mesotheliomas has been found in other studies of refinery workers, and it rose steadily with duration of employment. The rate for mesothelioma was higher among those first employed >45 years ago and those employed in job groups with greater potential for exposure to asbestos. 1 Divine BJ, Hamnan CM,Wendt JK.An updated report on the Texaco mortalitv studv 1947-1993 Pan I. Analvsis of overall patterns 6f m&nality. Occup Environs Med 1999;%.167-73. 2 Tsai SP,Waddell LC,Gilstrap EL, et ai. Mortality among maintenance employees potentially exposed to asbestos in a refinery and petrochemical plant. Am 3 Ind Med 199629: 89-98. 3 Wong 0. Risk of acute myeloid leukaemia and multiple myeloma in worlrers exposed to benzene. Ocncp Etlyimn Med 1995;52:380-4. 4 Wong 0,Raabe GK. Cell-type specitic leukemia analyses in a combined cohort of more than 208,000 petroleum workers in the United States and the United Kingdom. Reg T& B ~ r m a c d1995;21:307-21. 5 Satin KF, Wong 0, Yuan LA, et d.A 50-year mortality follow-up of a large cohort of oil refinery workers in Texas. J Occup Envimn Med 1996~38492-506. 6 Fraumeni JF,Blot W.Lung and pleura. In: Schonenfeld D, Fraumai JF,e&. Cancer epidnniorogv and prewnrion. Philadelphia:W B Saunders, 198256442. 7 Marsh GM, Ehland J, Sefcik S.M d i t y andpopdzim daza syrrnn (MPDS). Pittsbuqh, PA University of Pittsburgh, Lkparnnent of Biostatistics Techniul Rcport, 1987 8 Monson RR. A d y 8 i s of relative survival and proportionate mortality. CompuBiomod Res 1974;7:325-32. 9 Doll R Pet0 R Thc cawes of c a w . New York:Oxford Uni- VersiGPress, 1981:1243. 10 Schnatter AR, Theriault G, Kaa AM, et d.A retrospective mortality study within operating segments of a petruleurn company. A m 3 Ind Med 1992;22:209-29. 1 1 Kaplan SD. Update of a mortality study of workers in petroleum rdineries.J OccupMed 1986;28:514-6. 12 Homstra M. A mortality study of Whiting refinery employ- ees. Chicago, E.Report for Amoco Corporation, 1993. 13 Collingwood KW, b a b e GK, Wong 0. An updated cohort mortality study of workers at a northeastern United States petroleum refinery. Inr Arch OccupEm'ronHedrh 1996;68: 277-88. 14 Raabe GK, Collingwood KW, Wong 0.An updated mortality study of workers at a petroleum refinery in Beau- mont,TX. Am3lndMed 1998;33:61-81. 15 Wen CP,Tsai SP,Gibson RL.A report on brain tumors from a retrospective cohort study of refinery workers. Ann NYAcad Sa.1981;381:130-8. 16 Greenwald P, Friedlander BR, Lawrence CE, et d. Diagnostic sensitivity bias: an epidemiologic explanation for an apparent brain rumor excess. 3Occup Med 1981;23: 690-4. 17 Wong 0,Morgan RW,Bailey WJ,et d.An epidemiological study of petroleum refinery workers. BrJ Ind Med 1986;43 6-17. 18 Wong 0,Whonon MD. Diagnostic bias in occupational epidemiologic studies: an example based on the vinyl chloride literature.AmJlndMed 1993;24251-6. 19 Inskip PD, b e t MS, Heineman EF. Etiology of brain tumorsin adults. EprdemiOrRew 1995;15:382414. 20 Ries LAG, Kwary CL, Hankey BF, et d,eds. SEER cancer statistics rmiew 1973-94. Bethesda, MD: National Cancer Institute, 1997. (NIHpub1 no 97-2789). 2 1 Chung RY, Seizinger B R Moleculargenetics of neurological tumors. J M e d G a t 1992;29361-7. 22 Ron E, Modpn B, BoiceJD Jr, et d.Tumorsof the brain and central nervous system after radiotherapy in childhood. N End3 Med 1988;3191033-9. 23 Neglia JP, Meadows AT, Robison LL, et al. Second neoplasms after acute lymphoblastic leukemia in childhood.NEngiJMed 1991;325:1330-6. 24 International Agency for Research on Cancer. URC Monographs of rhL RKllunrion of the r n ' n o p z i c risk of chemuals w humans. Chmd d u a r i o n r of coninogmticiry:an updating of URC momgraph. Vols 1-42,suppl7. Lyon, France: International Agency for Research on Cancer, 1987. 25 Wong 0,Raabe GK. Multiple myeloma and benzene exposure in a multinational cohort of more than 250,000 petroleum workers. Rsg ToxicdPharmord 1997;%.188-99. 26 pafy C, Smnek III E, Gloeckler L. Accuracy of cancer death certi6cates and its effect on cancer monality statistics.AmJAcblic Health 1981;71:242-50. Mortality patterns among workers exposed to acrylamide: 1994 follow up Gary M Marsh, Lorraine J Lucas, Ada 0 Youk, Laura C Schall Department of Biostatistics, Graduate School of Public Health,University of Pittsburgh,PA, USA G MMarsh A 0 Youk L C Schall 22 Brier Road, Whitehouse Station,NJ 08889, USA L J Lucas Correspondence IO: h G M Mprsh, A-410 Crabtree Hall, Graduate School of Public Health, University of Pittsburgh, Pinsburgh,PA 15261, USA. email: grnarsh+@pitt.edu Accepted 17 September 1998 Abstract Objective-To update the mortality experience of a cohort of 8508 workers with potential exposure to acrylamide at three plants in the United States &om 1984-94. Methodt+-Analyses of standardised mortality ratios (SMR) with national and local rates and relative risk (RR) regression modelling were performed to assess site specific cancer risks by demographic and work history factors, and exposure indicators for acrylarnide and muriatic acid. ResuZts-For the 1925-94 study period, excess and deficit overall mortality risks were found for cancer sites of interest: brain and other central nervous system (CNS) (SMR 0.65,95% confidence interval (95% CI) 0.36 to 1.09), thyroid gland ( S h l R 2.11,95% CI 0.44 to 6.17), testis and other male genital organs ( S M R 0.28, 95% CI 0.01 to 1.59), and cancer of the respiratory system (SMR 1.10, 95% CI 0.99 to 1.22); however, none was significant or associ- ated with exposure to acrylamide. A previously reported excess mortality risk of cancer of the respiratory system at one plant remained increased among workers with potential exposure to muriatic acid (RR 1.50,95% CI 0.86 to 2-59), but w a o~nly slightly increased among workers exposed or unexposed to acrylamide. In an exploratory exposure-response analysis of rectal, oesophageal, pancreatic, and kid- ney cancer, we found increased SMRS for some categories of exposure to acryla- mide, but little evidence of an exposureresponse relation. A significant 2.26-fold risk (95% CI 1.03 to 4.29) was found for pancreatic cancer among workers with cumulative exposure to acrylamide >0.30 mg/m'.years; however, no consistent exposure-response relations were detected with the exposuremeasures consid- ered when RR regression models were adjusted or time since first exposure to acrylamide. Conclusion-The contribution of 1115 ad- ditional deaths and nearly 60 000 personyears over the 11 year follow up period corroborate the original cohort study findings of little evidence for a causal relation between exposure to acrylamide and mortality from any cancer sites, including those of initial interest. This is the most definitive study of the human carcinogenic potential of exposure to acrylamide conducted to date. (Omp Environ Med 1999;56:181-190) Keywords: acrylamide; muriatic acid; cohort mortality study In 1989, Collins et all reported the mortality experience of 8854 workers with potential exposure to acrylamide, a substance widely used in the manufacture of water soluble polymers used for water treating, paper mining, and sugar processing, at four Cytec Industries (formerly the chemical division of American Cyanamid Company), in three United States plants (Fortier, LA; Kalamazoo, MI; and Warners, NJ), and a plant in Botlek, The Netherlands. The original acrylamide study was prompted by animal studies that suggested acrylamide's carcinogenic potential based on an increased incidence of cancers of the brain and central nervous system (CNS), thyroid 'gland, other endocrine glands, and reproduc- tive organs,' and by limited epidemiological data. Sobel et al` in a small cohort mortality study of 371 workers exposed to acrylamide, reported 11 observed cancer deaths versus 7.9 expected due to excess cancers of the digestive tract and respiratory system among workers exposed to organic dyes. Among workers not exposed to organic dyes, four cancer deaths were observed versus 6.5 expected. The original acrylamide study found no significant excesses in total or cause specific mortality between 1925 and 1983 in the four plants. An exposure-response analysis of 2293 workers with exposure to acrylamide showed no trend of increased risk of mortality from several cancer sites. A significant excess in res- piratory cancer (SMR 1.31, p < 0.05) was found at the Warners factoryywhich was largely confined to two groups: men who worked in the muriatic acid operations ( n = l l deaths) between 1925 to 1956, and men hired between 1940 to 1949 who had worked less than 1 year in various departments (n=52 deaths). The investigators, including two of the current authors (GMM and LJL) concluded that the results did not support the hypothesis that acrylamide is a human carcinogen.' An extended and updated investigation was undertaken to examine the acrylamide cohort mortality experience from malignant neoplasms relative to exposure to acrylamide, and to investigate the apparent cluster of respiratory cancers at the Warners factory with emphasis on exposure to muriatic acid. we report here the results of an 11 Year follow UP (1984-94) of the original acrylamide cohort. Methods STUDY POPUIATION The original acrylamide cohort included 8854 male employees with full time work experience 182 Marsh, Lucas, Y a k ,et al Table 1 Characteristicsof study plants and popularion used in 1994 update Characteristics Plant start up date AMD production dates Total sublects: White men Non-white men Year of birth: < 1900 1900-19 1920-39 3 1940 Year of hire (entry into study): 1925-39 1940-49 1950-59 1960-73 Duration of employment (y): c1 1-4.9 5-9.9 10-19.9 Z20 Person-years (1 925-94): Total Unexposed to AMD Exposed to AMD' 0.001-0.029( mg/m'.y) 0.03-0.29 20.30 Person-years (1950-94): Total Unexposed to AMD Exposed to AMD* 0.001-0.029 ( mg/m'.y) 0.03-0.29 Z0.30 Tobacco smoking: Never smoker Ever smoker unknown Fortier Kalamazw 1951 1966 Dresent 1295 1172 123 1930 1967 vresent 60. 57 3 W a r m AU plants 1917 1954-85 7153 6013 1140 - 8508 7242 I266 15 197 15 724 17 373 23 633 2602 2908 1010 639 2814 3649 1406 05 0 12 730 17 565 26 1584 2471 1566 1532 1589 2483 2313 2123 230 6 285 12 133 3 248 8 399 31 3932 1537 434 453 797 4168 1834 570 709 1227 40592 21320 19272 6710 11850 712 1884 726 1158 328 632 198 245255 201781 43474 6274 12296 24904 287731 223827 63904 13312 24778 25814 40592 21320 19272 6710 11850 712 1778 644 1134 328 608 198 214615 174467 40148 6248 11991 21909 256985 196431 60554 13286 24449 22819 196 4 550 25 549 31 421 1558 5174 621 2133 5754 *Person-years refer to follow up time not exposure time. Person-years among unexposedworkers includes follow up time of workers before their firstexposure to AMD (started employment in an unexposed job). Table 2 Vital sfntus of study population by plant:follow up to the end of 1994 Vitals f n m Alive: Assumed Confirmed Dead: Death certificate (%) N o death certificate Unknown (%) Fodr 987 582 405 288 280 (97) 8 20 (2) Kalamazm 36 19 17 24 24 (100) 0 0 (0) Warnen 3777 2861 916 2970 2807 (95) 163 406 (6) AB &nu 4800 3462 1338 3282 3111 (95) 171 426 (5) at any of the four study plants between 1 January 1925 and 31 January 1973. In the current update, we did not include the 346 workers from the Botlek plant because the follow up was incomplete at the time of analysis. For the United States plants, company records were reviewed to update the work histories and the exposures to acrylamide and muriatic acid of study members actively employed at the end of 1983. The United States study cohort includes 8508 workers. Table 1 shows selected characteristicsof the United States study population by plant. The Warners factory is the oldest and largest plant in the study contributing 84% of the total study members and person-years. About half of the cohort was hired before 1950 and employed at the Warners or Kalamazoo plants, and about half were short term workers (defined as <1 year of employment). Most of the cohort members in each plant are white. Because of incomplete data, tobacco smoking history had limited use as a covariable in the original exposure-responseanalysis and now. COHORT TRACING The vital status of the United States study members was determined as of 31 December 1994 with Cytec personnel and pension files and based on a two stage tracing protocol that uses several conventional tracing sources, including the Social Security Administration and the National Death Index.5 Consistent with our protocol, all people who died afcer the 1994 follow up period or who were traced and not identified as deceased during the study period were assumed to be alive as of the study end date. Those with unknown vital status were untraceable due to missing Social Security numbers, a methodological limitation of the original study due mostly to Warners employees who started work before the establishment of Social Security. Death certificates were obtained from the corresponding state health department, and to conform with the original study, were coded by a nosologist to the underlying cause of death with the 8th revision rules of the international classification of diseases (ICD-8). Table 2 shows the plant specificvital status of the United States study population at the end of 1994. A total of 3282 deaths as identified through 1994, an increase of 1115 deaths from the original 1983 follow up of the United States plants. People lost to follow up decreased from 513 (6%) in the original study to 426 (5%). Death certificates were obtained for 3111 (95%) of all deaths. EXPOSURE ESTIMATES Amylamide The general methods of the quantitative estimates of exposure to acrylamide detailed by Collins et UP considered a worker as exposed to acrylamide if his cumulative exposure value is >0.001 mg/m'.years, the approximate equivalent exposure of a 1 day average concentration to the current permissible exposure limit of 0.3 mg/m'. From the individual worker job histories and the job and time specific exposure estimates, three time dependent summary measures of exposure to acrylamide, were computed for each worker: 0 Duration of exposure=the sum of the days spent in jobs with non-zero exposure to acrylamide (y) Cumulative exposure=the product of the number of days in each job and the estimated average daily exposure to acrylamide, summed across all jobs (mg/m'.y) 0 Average intensity of exposure=the ratio of cumulative to duration (mg/m') Exposure estimates for the 1984-94 period were assigned with the 1983exposures to acrylamide developed in the original study. Few employees in the follow up period would have potential exposure to acrylamide, as the Warners plant stopped acrylamide operations in 1985 and any extrapolation error would be small with minimal impact on our findings. This conservative operating assumption was also supported by a review of industrial hygiene monitoring data and by plant personnel, including one coauthor (LJL),knowledgeable Table 3 Summary statistics for exposure measures to AMD and muriatic acid by plant, - all wrkers, 1925-94 plants used wet and d r y acrylamide in the formulation of other products. Kalamazoo used Exposure indicator* Acrylamide: Duration of exposure (y): Minimum 25th percentile Median 75 th percentile Fortier Kalamazoo Warners AU plants 00 00 0 2.62 5.40 15.58 00 00 00 00 only wet acrylamide. Potential exposure may have occurred during monomer and polymer production from inhalation of dry powder, acrylamide monomer, or aerosols of acrylamide solution and dermal absorption of the anylamide monomer and solutions. Maximum Mean 34.62 4.16 41.67 7.62 40.99 41.67 1.20 1.69 SD 7.25 9.68 4.79 5.42 STATISTICAL ANALYSES Coefficient of variation Average intensity of exposure (mgim'): Minimum 25 th percentile Median 75 th percentile Maximum Mean SD Coefficient of variation Cumulative exposure (mgim'.y) 174.3 0 0 0 0.010 0.055 0.007 0.01 1 157.2 127.1 0 0 0.003 0.008 0.124 0.009 0.019 201.8 400.5 319.9 00 00 00 00 2.200 2.200 0.115 0.098 0.364 0.336 315.8 342.7 Descriptive analysis of SMRs We examined the total and cause specific mortality experiences of the United States acrylamide cohort from 1 January 1925 to 31 December 1994. Cohort analyses were performed with a modified life table procedure from the occupational cohort mortality programme (OCMAP).'T8 Person-years at risk Minimum 25 th percentile Median 75 th percentile Maximum Mean SD Coefficient of variation Muriatic acid; Duration of exposure (y): Minimum 0 0 0 0.06 0.81 0.04 0.08 190.7 0 0 0.02 0.08 0.72 0.08 0.15 180.8 0 0 0 0 32.31 0.29 1.34 458.3 0 0 0 0 32.31 0.25 1.23 487.6 0 contributed by each study member were jointly classified by plant, race, age group, calendar time, duration of employment (DOE), and the time since first employment (TSFE). Personyear counts began at the date of hire and continued until date of death or the end of the 1994 study period. For workers lost to follow up, person-year counts stopped at the last date 25 th percentile Median 75 th percentile Maximum 0 of known vital status, which was always the date 0 0 of the end of employment. 26.53 We computed expected numbers of deaths Mean SD Coefficient of variation 0.08 1.02 1256.4 *Computed from date of hire to earliest of date of ending work,death, or 31 December 1994. by multiplying average annual race, age, and time specific standard population death rates by the person-years at risk in the corresponding race age time intervals. To coincide with of acrylamide processes during the follow up update periods, 1980-4 rates were applied to period. 1980-3 person-years and 1985-9 rates were applied to 1984-9 person-years. As in the Muriatic acid original study, expected numbers of deaths for Muriatic acid production operated only at the the United States plants were computed with Warners plant between 1422 and 1937; the total United States male population as the between 1937 and 1956 muriatic acid exposure standard population. United States male death was restricted to trans-shipment and repackag- rates covering the 1925-89 observation period ing. Duration of exposure was the only were obtained from the cohort analysis soft- muriatic acid exposure measure computed as ware developed by M ~ n s o nF. ~or the 1990-4 only a qualitative estimate was assessed (pres- period, correspondingUnited States rates were ence or absence of potential exposure for each obtained for comparable categories from the job). mortality and population data system (MPDS) Table 3 shows selected summary statistics maintained at the University of Pittsburgh." for the three measures of exposure to acryla- As an enhancement to the original study, we mide computed for all workers in the 1994 computed expected numbers of deaths based acrylamide cohort and for duration of exposure on MPDS rates for the male populations of the to muratic acid at Warners. For all workers counties surrounding the factory (for Warners- combined, 60 554 (24%) of the total person- Middlesex and Union Counties, NJ) or the years are associated with exposure to acryla- local county in which the subcohort largely mide (table 1). The values of each exposure resides (for Fortier-Jefferson Parish, LA; for measure are generally small, consistent with Kalamazoo-Kalamazoo County, MI). Due to the observation in table 1 that nearly half of the limitations of the MPDS data, expected study members were short term workers. For numbers of cancer deaths were limited to the example, at Warners, the mean intensity of period 1950-94; non-cancer deaths were lim- exposure to acrylamide was 0.115 and the ited to 1960-94 (with 1962-4 rates applied to mean duration of exposure to muratic acid was 1960-4 person-years). MPDS rates were used only 0.08 years. only for those cause of death categories defined Exposure to acrylamide varied, as each of the by identical or nearly identical ICD codes in domestic plants produced or used acrylamide the Monson and MPDS rate files. Because differently. Acrylamide is produced in two local death rates usually provide the most valid forms, wet acrylamide in aqueous solution, and external mortality comparisons (as they help to dry acrylamide that can be in either a powder adjust for the social, cultural, and economic or a solid pellet. The Warners plant produced factors related to disease),all but the aggregate both forms of acrylamide. The Fortier factory analysis were based exclusively on local rates. manufactured wet acrylamide monomer. Both Moreover, because the counties involved repre- 184 Marsh, Lucas, Yack,et al sent large population areas, the local rates are measured with good precision. Mortality excesses and deficitswere expressed as standardised mortality ratios (SMRs) along with their 95% confidence intervals (95% CIS). The SMRs were computed for subgroups of the cohort defined by plant, race, follow up period, calendar period, year of hire, duration of employment, and the time since first employment. The SMRs were also computed for selected causes of death for the measures of exposure categories of acrylamide and muriatic acid with and without exposure lagging.'" Here, person-year counts in the unexposed baseline categories include the observation time of workers before their first exposure. The sigmficance of SMRs was assessed with Poisson probabilities.All tests were done at the 0.05 sig- nificance level and no adjustment was made for multiple comparisons. Relative risk regression analysis Relative risk regression modelling was used to investigate the dependence of the internal cohort rates (modelled as time to death) for selected cancer sites on combinations of the categorical acrylamide or muriatic acid exposure measures, with adjustment for potential confounding factors. Study data from the entire 1925-94 period were modelled. For each cancer site examined, risk sets were explicitly constructed from the cohort data file with age as the primary time dimension, with the RISKSET program module in O C W - P L U S . * To adjust for year of birth (cohort) effects, risk sets were caliper matched on year of birth. The time dependent exposures were evaluated for each person at each event time they were at risk. Multiplicative relative risk (RR) models of the form h(t) = h, (t) exp{x(t)p}were fitted to the internal cohort rates. Mathematical details of the models are given elsewhere."-" The conditional logistic regression program in EGRET" was used to estimate p from the explicitly constructed risk sets. Categorised forms of the covariates were considered to parallel the descriptive SMR analysis of mortality relative to exposure. The demographic and exposure variables were first considered univariately as categorical variables to idenufy patterns of univariate associations with the outcome and sparse data problems. Possible exposure-disease associations were then evaluated with a forward stepwiseapproach to adjust for possible confounders. Effect modification was assessed as far as possible. The significance of each main effect (expressed as a global p value) and interaction was assessed with a likelihood ratio statistic.For the quantitative exposure variables that had a monotonic pattern in the parameter estimates, a test for linear trend was conducted (expressed as a trend p value). All tests were done at the 0.05 significance level with no adjustment made for multiple comparisons. Results GENERAL MORTALITY BY STUDY PERIOD Table 4 presents SMRs by cause and study period for the combined 1994 United States acrylamide cohort (some findings for the 1925-83 period differ slightly from those Tiable 4 Observed deaths and SMRs for selected causes by folknu up penod, all workers, nariaal comparisons Cause of death (lCDA-8) f 1925-83 Obs SMR 95%CI 1984-94 Obi SMR 95%CI 1925-94 Obs SMR 95%CI All causes (000-999): AU malignant neoplasms (140-209) Buccal cavity and pharynx (140-149) Digestive organs and peritoneum (150-159) Oesophagus (150) Stomach (151) Large intestine (153) Rectum (154) Liver (155, 156) Pancreas (157) Respiratory system (160-163) Larynx (161) Lung (162, 163) Bone (170) Skin (172, 173) Prostate (185) Testis and other male genital organs (186,187) Bladder (188) Kidney (189) Brain and other cenual nervous system (19 1,192) Thyroid gland (193) All lymphopoietic tissue (200-209) Lymphosarcoma and retidosarcoma (200) Hodgkin's disease (201) Leukaemia and aleukemia (204-207) Other lymphatic tissue (202,203,208) Benign neoplasms (210-239) Diabetes mellitus (250) Diseases of the circulatory system (390458) Non-malignant respiratory disease (460-519) Cirrhosis of the liver (571) All external causes of death (80&998) Unknown causes (999.9) People (n) Person-years 2167 496 13 141 16 35 38 16 5 27 202 8 194 2 4 29 0 13 12 5 2 39 6 8 14 11 8 26 1019 105 68 199 101 0.91** 0.87 to0.95 1.06 0.96 to 1.15 0.83 0.44to 1.42 1.07 0.90 to 1.26 1.15 0.66 to 1.87 1.34 0.94 to 1.87 0.94 0.67 to 1.29 1.20 0.69 to 1.95 0.51 0.16 to 1.20 1.09 0.72 to 1.59 1.25" 1.08 to 1.44 1.10 0.48 to 2.18 1.27- 1.10 to 1.46 0.88 0.11 to3.18 0.48 0.13 to 1.23 0.96 0.64 to 1.38 - 0.00to 1.23 1.06 0.56 to 1.81 1.06 0.55 to 1.86 0.36* 0.12 to 0.85 2.32 0.28 to 8.37 0.88 0.62 to 1.20 0.70 0.26 to 1.53 1.39 0.60 to 2.74 0.78 0.43 to 1.31 0.92 0.46 to 1.66 1.24 0.54 to 2.44 0.77 0.50 to 1.12 0.90** 0.85 to 0.96 0.75" 0.62 to 0.91 1.08 0.84 to 1.37 0.70*' 0.61 to 0.81 8508 228816 1115 357 8 85 15 12 28 8 5 17 139 6 133 0 6 38 1 14 10 9 1 21 0 0 9 11 2 15 434 74 12 43 70 0.76'' 0.72 to 0.81 0.89* 0.80 to 0.99 0.96 0.41 to 1.89 0.89 0.71 to 1.10 1.30 0.73 to 2.14 0.95 0.49 to 1.66 0.78 0.52 to 1.13 1.26 0.55 to 2.49 0.58 0.19 to 1.35 0.91 0.53 to 1.46 0.94 0.79 to 1.11 1.25 0.46 to 2.71 0.94 0.78 to 1.11 - 0.00 to 6.19 0.89 0.33 to 1.93 0.82 0.58 to 1.13 1.92 0.05 to 10.70 1.38 0.75 to 2.31 1.11 0.53 to 2.04 1.15 0.53 to 2.19 1.80 0.04 to 10.01 0.60' 0.37 to 0.92 - 0.00 to 2.35 - 0.00 to 4.04 0.68 0.31 to 1.29 0.65 0.32 to 1.16 0.60 0.07 to 2.15 0.53'' 0.30 to 0.87 0.61**0.56 to 0.67 0.53** 0.42 to 0.67 0.54* 0.28 to 0.94 0.65**0.47 to 0.87 5942 58916 3282 853 21 226 31 47 66 24 10 44 34 1 14 327 2 10 67 1 27 22 14 3 60 6 8 23 22 10 41 1453 179 80 242 171 'p<0.05; ** p<O.Ol. tMonson life table programme ICD-8 categories, labels, and codes for US plants for 1925-89; correspondingMPDS rates for 19904. 0.85" 0.98 0.88 0.99 1.22 1.22 0.87 1.22 0.55 1.01 1.10 1.16 1.11 0.70 0.66 0.88 0.28 1.20 1.08 0.65 2.11 0.76* 0.59 1.20 0.74 0.76 1.02 0.66** 0.79** 0.64** 0.94 0.69** 0.82 to 0.88 0.92 to 1.05 0.54 to 1.34 0.87 to 1.13 0.83 to 1.73 0.89 to 1.62 0.67 to 1.10 0.78 to 1.82 0.26 to 1.00 0.74 to 1.36 0.99 to 1.22 0.63 to 1.95 0.99t01.24 0.08 m 2.52 0.31 to 1.22 0.68 to 1.1 1 0.01 to 1.59 0.79 to 1.75 0.68 to 1.64 0.36 to 1.09 0.44 to6.17 0.58 to 0.97 0.22 to 1.29 0.52 to 2.37 0.47 to 1.10 0.48 to 1.16 0.49 to 1.87 0.47 to 0.89 0.75 to 0.83 0.55 to 0.74 0.74to 1.17 0.61 to 0.79 8508 287731 i Table 5 Observed deaths and SMRs for selected cancer sites by plant, US workers, 1950-94, local county comparison Cause of death (ICDA- 8)t firrier Obi SMR 95%CI Kalamazoo Obs SMR 95% GI Warnm Obs SMR 95% CI All malignant neoplasms (14Cb209) 99 0.93 0.76to 1.14 6 1.03 0.38 to 2.24 736 0.91** 0.84 to 0.97 Buccal cavity and pharynx (14Cb149) Digestive organs and peritoneum (150-159) 4 23 1.32 0.36 to 3.39 1.04 0.66 to 1.56 0 0 - Oesophagus (150) Stomach (151) 4 1.41 0.38 to3.61 0- 4 1.37 0.37 to 3.56 0 - 0.00 to 24.72 0.00 to 2.49 0.00 to 33.01 0.00 to 11.91 17 0.75 198 0.83.' 27 1.oo 40 0.89 0.44 to 1.20 0.72 to 0.96 0.66 to 1.45 0.63 to 1.21 Large intestine (1 53) 8 1.01 0.44 to 1.99 0 - 0.00 to 8.1 1 58 0.69** 0.52 to 0.89 Rectum (154) 1 0.58 0.01 to 3.22 0 - 0.00 to 20.91 21 0.93 0.58 to 1.42 Pancreas (157) 6 1.36 0.50to 2.96 0 - 0.00 to 12.58 38 0.94 0.66 to 1.28 Respiratory system (160-163) 36 0.78 0.55 to 1.08 4 1.95 0.53 to 4.99 298 1.11 0.98 to 1.24 Bronchus, trachea, lung (I 62)$ 34 0.77 0.54to 1.08 4 2.02 0.55 to 5.16 284 1.11 0.99 to 1.25 Prostate (185) 3 0.52 0.11 to 1.52 1 1.86 0.04 to 10.34 63 0.85 0.66 to 1.09 Testis and other male genital organs (172.5, 173.5, 186, 187)$ Bladder and other Urinary organs (1 88, 189.9)$ 0 2 1.08 0.00 to 10.88 0.13 to3.89 0 0 - 0.00 to 129.95 0.00 to 19.71 2 0.81 0.10 to 2.91 25 1.02 I 0.66 to 1.50 Kidney (189.0, 189.1,189.2)$ Cennal nervous system (191,192) 3 0.98 0.20 to 2.87 1 6.95 0.17 to 38.70 5 1.80 0.59 to4.21 0 - 0.00 to 22.57 18 0.89 9 0.49. 0.53 to 1.41 0.22 to 0.93 Thyroid gland and other endocrine glands (193, 194N All lymphatic and haematopoietic tissue (200-209) i Leukaemia and aleukaemia (204-207) 1 3.93 0.10t021.91 0 - 11 1.30 0.65 to 2.32 0 2 0.63 0.08 to 2.27 0 - 0.00 to 291.70 0.00 to 5.96 0.00 to 17.75 5 2.17 0.71 to 5.07 46 0.66** 0.48 to 0.88 21 0.80 0.50 to 1.23 Other lymphopoietic tissue (202,203,208, 2W$ 8 2.31 0.99 to4.54 0- 0.00 to 13.68 14 0.50* 0.28 to 0.85 ~~ *p<O.OSI; **pCO.Ol. tMPDS ICD-8 Categories, labels and codes for US plants. W D S ICD codes M e r h m correspondingMonson codes. reported by Collins et aZ' due to the cohort data revisions noted above). During the 1984-94 update period there were 1115 deaths, yielding a s i e c a n t @<0.01) 24% overall deficit in total mortality compared with the general United States Population. For the Overall 1925-94 study period, there was a significant (pc0.01) 15% deficit in total mortality based on 3282 observed deaths and 287 731 personyears. The SMRs for all maknant neoplasms combined and for many cancer site specific categories also decreased in the 1984-94 update period. In particular, cancer of the respiratory system decreased from a significant (p<O.Ol) 25% excess in mortality for the 1925-83 period to a 6% deficit (based on 139 cancer of the respiratory system deaths), resulting in a non-significant 10% excess for the combined 1925-94 study period. Most of the deaths fiom cancer of the respiratory system were due to lung cancer. For the other cancer sites of initial interest in the combined study period (table 4), we found deficits in deaths for cancer of the brain and other parts of the CNS (SMR 0.65, 95% CI 0.36 to 1.09) and cancer of the testis and other male genital organs (SMR 0.28,95% CI 0.01 to 1.59). A non-significant excess in cancer of the thyroid gland was based on three deaths ( S M R 2.11, 95% CI 0.44 to 6.17). The SMR for thyroid cancer decreased to 1.80 in the 1984-94 update period. Most of the nonmalignant cause of death categories shown in table 4 showed deficits in deaths for the update and combined study periods and many deficits were significant (pCO.05). The SMRs for some cancer site specific categories increased in the update period; however, none of the resulting excesses in the update or combined study periods w a s significant. Although not shown here, SMR analyses by race, period, year of employment and duration of employment were essentially unremarkable for the initial cancer sites and the other cause of death categories examined in table 4. GENERALMORTALITY BY p m Table 5 shows the 1950-94 cancer mortality experience of workers from each United States plant compared with the local male populations. Restricting the observationperiod excluded only 12 cancer deaths, all from the Warn- plant. Shown are the cancer site categories of initial interest (brain and other parts of the CNS,thy- roid and other endocrine glands, and testis and other male genital organs) or sites with at least 20 observed deaths across the three plants. For certain categories (thyroid and other endocrine glands), the observed number of deaths differs between tables 4 and 5 due to the slight variations in the ICD codes used in the MPDS and Monson rates. Table 5 shows that 736 (88%) of the observed cancer deaths occurred at Warners (SMR 0.91, 95% CI 0.84-.97), whereas only six occurred at Kalamazoo (SMR 1.03, 95% CI 0.38 to 2.24). None of the plants show significant excess in mortality at any specific cancer site. Among the initial sites, non-significant excesses in deaths occurred at Fortier for brain and other parts of the CNS (SMR 1.80, 95% CI 0.59 to 4.21) and thyroid and other endocrine glands at Fortier ( S M R 3.93, 95% CI 0.10 to 21.911, and at Warners (SMR 2.17, 95% CI 0.71 to 5.07). The two deaths from cancer of the testis and other male genital organs occurred at Warners resulting in a 19% deficit ( S M R 0.81,95% CI 0.10 to 2.91). CANCER MORTALITY BY EXPOSURE TO ACRYLAMIDE Table 6 depicts the 1950-94 cancer mortality experience of all United States workers compared with those with exposure to acrylamide for the selected cancer sites examined in table 5 . Consistent with original study results, we found deficits compared with local county 1'86 Marsh, Lucas, Youk,et d Table 6 Observed deaths and SMRs fw selected cancer sites by exposure to acylamide, US workers, 1950-94, local county comparison Cduse of death (ICDA4)f All malignant neoplasms (14C-209) Buccal cavity and pharynx (140-149) Digestive organs and peritoneum (150-159) Oesophagus (1 50) Stomach (151) Large intestine (153) Rectum (154) Pancreas (157) Respiratory system (160 to 163) Bronchus, trachea, lung (1 62) Prostate (185) Testis and other male genital organs (172.5, 173.5, 186, 187) Bladder and urinary organs (188, 189.9) Kidney (189.0, 189.1,189.2) Central nervous sptem (191,192) Thyroid gland and other endocrine glands (193, 194) All lymphaticand haernatopoietictissue (20e209) Leukaemia and alaeukaemia (204-207) Other lymphopoietic tissue (202,203,208,209) People (n) Person-years *p<0.05; **pCO.Ol. tMPDS ICD-8, labels, and codes for US plants. Unexposed (<O. 001 mg:m`.y) Exposed (Z 0.001 mg1d.y) Obs SMR 95%CI -Obs SMR 95%CI 681 0.89'* 0.83 to 0.96 160 18 0.85 0.50 to 1.34 3 177 0.81** 0.70 to 0.94 44 24 0.96 0.61 to 1.42 7 39 0.95 0.68 to 1.30 5 55 0.72* 0.54 to 0.93 11 17 0.82 0.48 to 1.32 5 30 0.80 0.54 to 1.14 14 276 1.07 0.95 to 1.20 62 260 1.07 0.94 to 1.20 62 60 0.88 0.67 to 1.14 7 0.98 0.65 1.03 1.41 0.69 0.71 1.33 1.79 1.04 1.09 0.58 0.83 to 1.14 0.13 to 1.89 0.75 to 1.38 0.57 to 2.90 0.22 to 1.60 0.3610 1.28 0.43 to 3.09 0.98 to 3.01 0.79 to 1.33 0.84 to 1.40 0.23 to 1.19 2 0.86 0.10 to 3.10 22 0.98 0.61 to 1.48 16 0.84 0.48 to 1.37 11 0.64 0.32 to 1.14 4 1.91 0.52 to 4.88 43 0.67** 0.48 to 0.90 17 0.70 0.41 to 1.12 16 0.63 0.36 to 1.02 7532 202409 0- 0.00 to 7.09 5 1.19 0.39 to 2.77 6 1.36 0.50 to 2.96 3 0.74 0.15 to 2.15 2 4.27 0.52 to 15.42 14 0.98 0.54 to 1.64 6 1.15 0.42 to 2.51 6 1.00 0.37 to 2.18 2004 54576 cancer mortality among both exposed and unexposed workers for all cancers combined and many specificcancer sites, including brain and other parts of the CNS, and testis and other male genital organs. Although the original study reported a significant (p<0.05) 31% excess for cancer of the respiratory system among unexposed workers (and 14% among those exposed), we now found that up to the end of 1994 only slight non-significant 7% and 4% excesses for cancer of the respiratory system among the unexposed and exposed groups, respectively. These decreased SMRs follow from the overall reduction in mortality from cancer of the respiratory system during the 1984-94 follow up period noted in table 4. For the initial site, cancer of thyroid and other endocrine glands, a non-significant increased mortality risk was found among both unexposed (SMR 1.91, 95% CI 0.52 to 4.88) and exposed ( S M R 4.27, 95% CI 0.52 to 15.42) workers. EXPLORATORY EXPOSURE-RESPONSEANALYSIS OF SMRS Four cancer sites-rectum, oesophagus, pancreas, and kidney-were selected for more detailed exploratory exposure-response analyses based on a 220% excesses in mortality in exposed workers and deficits in unexposed workers (table 6); none of these excesses was significant.Table 7 shows observed deaths and SMRs based on local rates (1950-94) for the four cancer sites by duration of employment, the time since first employment, and the three exposure measures, duration of exposure to acrylamide, intensity of exposure to acrylamide, and cumulative exposure to acrylamide. The exposure categories for cumulative expo- Toble 7 Observed deaths and SMRs for selected cancer sites by duration of empkyment, time sincefirst employment, and measures of exposure w acrykamide, all US workers, 1950-94, locd county comparisons Oesophaps Obs SMR 95%CI Duration of employment (y): < 1 12 0.84 1-14 b 15 . 9 10 0.95 1.60 Time since first employment (y): z. 20 3 0.69 20-9 6 0.80 2 3 0 22 1.21 Duration of exposure (y): Unexposed 24 0.96 0.001-4.999 4 1.63 5-19 3 1.80 b 20 0 - Cumulative exposure (mg/m'.y): <.001 24 0.96 0.001-0.029 2 2.58 0.03-0.29 3 1.70 20.30 2 0.82 Mean intensity of exposure (mum'): Unexposed 24 0.96 0.001-0.019 2 1.37 0.02-0.29 3 1.53 20.30 2 1.26 *p<0.05. 0.43 to 1.46 0.43 to 1.SO 0.77 to 2.94 0.14 to 2.01 0.29 to 1.73 0.76 to 1.83 0.61 to 1.42 0.45 to 4.18 0.37 to 5.26 0.00 to 4.10 0.61 to 1.42 0.31 to 9.30 0.35 to 4.97 0.10 to 2.98 0.61 to 1.42 0.17 to 4.95 0.32 to 4.47 0.15 to 4.53 Rectum Obs SMR 95%CI 5 0.51 0.1610 1.18 12 1.37 0.71 to 2.39 5 0.86 0.28 to 2.02 3 0.7 1 0.15 M 2.07 5 0.85 0.28 to 1.98 14 0.98 0.53 to 1.64 17 0.82 0.48 to 1.32 3 1.92 0.40 to 5.61 1 0.71 0.02 to 3.96 1 1.20 0.03 to 6.70 17 0.82 0.48 to 1.32 1 2.31 0.06 to 12.88 2 1.73 0.21 to 6.23 2 0.92 0.11 to 3.31 17 0.82 0.48 to 1.32 2 2.12 0.26 to 7.65 0- 0.00 to 2.03 3 2.89 0.60 to 8.43 Pancreas Obs SMR 95%CI 17 0.87 0.51 to 1.39 15 0.95 0.5310 1.57 12 1.19 0.61 to2.08 4 0.66 0.18to 1.68 11 1.08 0.5410 1.92 29 1.00 0.6710 1.44 30 0.80 0.5410 1.14 5 1.46 0.47 to 3.41 5 1.79 0.58 ro4.17 4 2.42 0.66 to6.19 30 0.80 0.54to 1.14 3 2.77 0.57 to 8.09 2 0.73 0.09102.64 9 2.26* 1.03to 4.29 30 0.80 0.54 to 1.14 4 1.69 0.46 to 4.32 5 1.50 0.49103.49 5 2.31 0.75 to 5.40 Kidney Obs SMR 95%CI 8 0.79 0.34 to 1.55 7 0.86 0.35 to 1.78 7 1.36 0.55 to 2.79 2 0.58 0.01 to 2.09 3 0.54 0.11 to 1.58 17 1.18 0.69 to 1.89 16 0.84 0.48 to 1.37 2 0.99 0.12 to 3.59 3 1.88 0.39 to 5.48 1 1.18 0.03 to 6.56 16 0.84 0.48 to 1.37 1 1.45 0.04 to 8.08 2 1.17 0.14 to 4.23 3 1.49 0.31 to 4.35 16 0.84 0.48 to 1.37 1 0.66 0.02 to 3.66 3 1.71 0.35 to 5.01 2 1.68 0.20 to 6.08 I i sure to acrylamide are those used in the original study.' To help discern trends in S M h , categories for the other variables were chosen to most evenly distribute the numbers of observed deaths simultaneously across the cancer sites examined. For duration of exposure and time since first exposure to acrylamide, SMRs are generally smallest in the lowest categories. The Sh4Rs for oesophageal cancer increased monotonically with increasing values of both duration of exposure and time since first exposure to acrylamide. Table 7 shows no consistent evidence of an increased risk for cancer mortality associated with exposure to acrylamide, except for pancreatic cancer, where we found a signrficant (p<0.05) 2.26-fold risk (95% CI 1.03 to 4.29) in mortality among workers with a cumulative exposure to acrylamide of 20.30 mg/m'.years. Although a monotonically increasing risk with duration of exposure to acrylamide was found, none was found with cumulative exposure to acrylamide. An increased, non-significant 2.77-fold risk (95% CI 0.51 to 8.09) among workers was detected in the lowest cumulative exposure category. Lagged cumulative exposure to acrylamide (not shown) for periods of 5, 10, and 20 years produced a similar pattern of findings as noted in table 7. RISK OF CANCER OF THE RESPIRATORY SYSTEM REIATIVE TO EXPOSURE TO ACRYIAMJDE AND MURIATIC ACID We fitted RR regression models for cancer of the respiratory system with the subset of the cohort that had potential exposure to muriatic acid (Warners plant workers hired before 1957). All models were based on 276 cases of cancer of the respiratory system (16 exposed to muriatic acid) and 5923 controls, and were adjusted for age and calendar time. In the univariate models, only smoking (characterised as: never, ever, unknown) was a significant predictor of cancer of the respiratory system risk with RRS of 1.00, 8.72, 7.34, respectively. None of the variables considered, including the categorical muriatic acid or exposure to acrylamides, was a significant predictor of risk of cancer of the respiratory system. Workers exposed to muriatic acid had a non-significant 1.50-fold (95% CI 0.86 to 2.59) risk of cancer of the respiratory system compared with workers unexposed to muriatic acid. Models that included an exposure variable representing the joint effects of muriatic acid and exposure to acrylamide had no significant RRS or heterogeneity in risks of cancer of the respiratory system across the joint effects considered (results not shown). Table 8 Summary of relatiw risk regression analysisfor cancer of the pancreas, US plants, 192.5-94 Vdl2 ca%Tv Deaths Ruknuio 9S%CI Global terr (dfi Test* d *{&I Univariate modelst Race Year ofhire Duration of employment (y) Time since first employment (y) Smoking$ Time since first exposure (y) Durationof exposure (y) Cumulative exposure (mp/m'.y) Mean intemsity of exposure (mdm') Bivariable models:T Cumulativeexposure (mp/m'.y) Mean intemsity of exposure (mum') White Non-white 1925-39 194M9 1950-59 1960-73 <1 1-14 2 15 <20 20- 29 5 30 Never Ever unknown < 20 20-29 5 30 eo.001 0.00 1-<5 5-19 320 eo.001 0.001-<0.03 0.0><0.30 30.30 <0.001 0.001-0.02 0.02-<0.30 30.30 co.001 0.001-CO.03 0.03-CO.30 30.30 ~0.001 0.001-c0.02 0.02-<0.30 20.30 39 5 10 18 12 4 17 15 12 4 11 29 0 12 32 34 5 5 30 5 5 4 30 3 2 9 30 4 5 5 30 3 2 9 30 4 5 5 1.00 0.86 1.00 1.09 1.10 1.59 1.oo 1.11 1.39 1.oo 1.48 1.28 1.00 4.96 3.59 1.oo 2.19 2.54 1.oo 1.73 2.01 2.37 1.oo 3.14 0.77 2.63' 1.oo 1.77 1.67 2.86' 1.00 2.51 0.57 1.75 1.oo 1.27 1.05 1.94 0.34 to 2.20 0.105 (1) p=0.746 0.48 to 2.49 0.44 to 2.78 0.40 to 6.36 0.405 (3) p=0.939 0.55 to 2.24 0.66 to 2.96 0.738 (2) p=0.692 0.38 to 5.73 0.32 to 5.15 0.378 (2) p=0.828 0.82 to infinity 0.64 to infinity 3.79 (2) p=0.145 0.84 to 5.73 0.95 to 6.80 4.58 (2) p=0.101 0.64 to 4.60 0.78 to 5.23 0.82 to 6.82 4.161 (3) p=0.245 0.92 to 10.71 0.18 to 3.26 1.23to 5.60 7.69 (3) p=0.053 0.61 to 5.12 0.64 to 4.34 1.08 to 7.56 4.74 (3) p=0.192 - ~~ - - 4.22 (1) p=0.040 3.91 (1) p=0.048 - - 0.63 to 9.92 0.1 1 to 2.99 0.48 to 6.39 0.34 to 4.78 0.26 to 4.18 0.52 to 7.22 3.77 (3) p=0.287 1.18 (3) p=0.759 - - *p c.05. tTrend tests performed only on exposure variables exhibiting a monotonic increase or decreasein P-eter estimates.tRisk sets matched on exact age and year of birth, SLogXact used to model smoking due to sparse data problems. w k sets matched on a c t age and year of birth, models adjusted for time since first exposure to acrylamide. A 88 Marsh, Lucas, Ymk,et al EXPLORATORY RR REGRESSION ANALYSIS Table 8 shows the RR regression models relating internal cohort rates for pancreatic cancer to the potential confounding variables of interest and the categorical measures of exposure to acrylamide. The number of deaths observed, the estimated RR (95% CI), and the global test (of main effect) p value and a trend test p value, if warranted (monotonic increasing or decreasing pattern of exposure estimates) are shown for each category of the variables considered. The category used as baseline always has an RR value of 1.O; individual RRs are significant if their 95% CISdo not contain 1.O. All models are based on 44 cases and 10 261 controls and adjusted for age and calendar time. The top section of table 8 shows that none of the potential confounding variables considered is a significant predictor of risk of pancreatic cancer. Smoking history and time since first exposure to acrylamide had relatively low global test p values (0.145 and 0.101, respectively) and RRs >2.0 in the non-baseline categories. Due to sparse data, the model for smoking history was fitted with the exact conditional (stratified)logistic regression program LogXact.'6 Time since first exposure to acrylamide had a significant (p=0.040) positive monotonic trend in the parameter estimates. The univariate models of the three measures of exposure to acrylamide show a pattern of RRs qualitatively similar to the pattern of SMRs in table 7.The RRs are generally larger than the corresponding SMRs, due to the 20% deficit in risk of pancreatic cancer among workers in the lowest exposure categories (table 7) that serves as the baseline for the RR models. Table 8 shows that duration of exposure to acrylamide has a significant (p=0.048) monotonic increase in the parameter estimates with increasing duration of exposure to acrylamide.Table 8 also shows that cumulative exposure to acrylamide is a predictor of pancreatic cancer risk of borderline significance @=0.053), although the increase in RRs with increasing exposure to acrylamide is not monotonic. The RRs for the highest exposure categories of both cumulative exposure to acrylamide and intensity of exposure to acrylamide are significant @<0.05). The bottom of table 8 shows models for cumulative exposure to acrylamide and intensity of exposure to acrylamide, adjusted for the time since first exposure to acrylamide. Because of the sparse data for smoking history (zero non-smoking cases), models adjusted for smoking history would not converge. The adjusted models show a pattern of RRs qualitatively similar to their univariate counter- parts; however, RRs are uniformly lower, suggesting positive confounding by time since first exposure to acrylamide. Cumulative exposure to acrylamide is no longer a significant predictor of pancreatic cancer in the adjusted models, and the RRs for the highest exposure categories of both cumulative exposure to acrylamide and intensity of exposure to acrylamide are no longer significant. Lagged analyses (5, 10, and 20 years) with other categories of cumulative exposure to acrylamide produced a pattern of RRs for cumulative exposure to acrylamide qualitatively similar to those shown in table 8. Discussion and conclusions The overall mortality patterns found in the 11 year follow up period indicated a significant reduced mortality risk from all causes combined and all cancer sites combined. A similar pattern was noted for the entire 1925-94 study period. This favourable mortality pattern is probably influenced in part by the healthy worker effect, a relative absence of deleterious health risks relative to employment, and the effects of continuing employment with its many benefits-such as improved health care and quality of life. For the 1925-94 study period, we found no significant overall risk or plant specific risk for the cancer sites of initial interest: cancer of the brain and other parts of the CNS, thyroid and other endocrine glands, testis and other male genital organs, and cancer of the respiratory system. However, among the workers exposed to acrylamide, except for cancer of the respiratory system, the few observed deaths for the sites of initial interest resulted in low statistical power to detect important excesses and precluded a detailed examination of exposureresponse. Although not significant, we found increased SMRs for cancer of the thyroid gland (and the larger category, thyroid gland and other endocrine glands) among several cohort subgroups, including a fourfold risk among workers exposed to acrylamide and a twofold risk among unexposed workers. This pattern, based on six cases, with a decreasing S h 4 R across the update periods does not support a causal association between exposure to acrylamide and cancer of the thyroid and suggests that other unmeasured occupational or nonoccupational factors may be involved with these excesses. Additional cohort follow up will help to elucidate these possibilities. An examination of the cluster of suspected cases of cancer of the respiratory system among Warners employees' identified ah additional five cases of cancer of the respiratory system among workers with potential exposure to muriatic acid before 1957. Although there is some suggestion that employees with exposure to muriatic acid with and without exposure to acrylamide are at increased risk of cancer of the respiratory system, the limited number of observed deaths (n=16) and incomplete data on smoking history preclude definitive conclusions. Our modelling of cancer of the respiratory system risk showed no evidence of a joint effect of muriatic acid and exposure to acrylamide. In an exploratoryexposure-responseanalysis of SMRs, we found a non-significant monotonically increased risk of pancreatic cancer relative to duration of exposure to acrylamide. No increased risk pattern was found for either cumulative exposure or intensity of exposure to acrylamide, although a significant 2.26-fold risk was found among workers in the highest i category of cumulative exposure to acrylamide. workers-skewed distribution of person-years (20.30 mg/m3.years). All 14 cases of pancre- with respect to duration of exposure, low expo- atic cancer exposed to acrylamide were born sures, and incomplete smoking data, are found before 1939 and hired before 1966. No in many occupational cohort studies. As such, consistent pattern was readily apparent when many studies restrict cohort entry to long term individual work history records were exam- employees. Had short term employees been ined. A non-significantly increased risk of pan- excluded in the 1994 update, valuable data creatic cancer was found in the original study would have been lost-for example, the cancer among workers exposed to acrylamide, but of the respiratory system cluster. Potential there was no trend with increasing exposure.' selection bias from workers lost to follow up Modelling results showed a pattern of RRs may be operating, but the effects would be qualitativelysimilar to those found in the S M R minimal due to the small percentage of workers analysis and suggested potential confounding involved. Because we did not adjust p values for with time since first exposure to acrylamide multiple comparisons, some of our significant 1 and with a history of smoking. Smoking is the SMRs and RRs may be simply chance most consistent risk factor for pancreatic occurrences. cancer, associated with a greater than twofold Our study also has low statistical power to increased risk." In models adjusted for time detect excess risks for many cancer sites of ini- since first exposure to acrylamide, we found tial interest, especially in the subgroup analy- less evidence of an exposure-response relation ses. For example, for the combined 1925-94 for cumulative exposure and intensity of expo- study period, our study has the followingpower sure to acrylamide; RRs for these measures to detect twofold or greater risks in mortality decreased to non-sigdicant levels suggesting (at the 5% one sided significance level): cancer that this variable acted as a positive confound- of the brain and other parts of the CNS (0.25), ing factor.Although our inclusion of time since thyroid gland (0.18), and testis and other male first exposure to acrylamide as a co-factor in genital organs (0.41). On the other hand, the models with cumulative exposure to acryla- corresponding power to detect cancer of the mide and intensity of exposure to acrylamideis respiratory system of 0.87 is excellent. Al- statistically appropriate and biologically rel- though power is not relevant for the excess of evant, some bias may have resulted in the mod- pancreatic cancer identified in the exploratory els for cumulative exposure to acrylamide due analysis, a future study of this size would have to the collinearity between the two time a power of nearly 100% to detect a twofold or dependent factors. greater excess. Given the absence of a significant absolute By comparison, our acrylamide study has risk, the weak evidence for an exposure- many methodological strengths that include: i response relation and the inability to consider large cohort size; long observationperiod; good smoking adequately, as a potential confound- death certificate ascertainment for the United ing factor, our findings for cancer of the States plants; sufficient statistical power to pancreas should be interpreted with caution, in detect meaningful excesses for many categories the context of an exploratoryanalysis to gener- of cause of death; ample duration of exposure ate hypotheses. This is supported by a recent (many years of acrylamide production); quan- review by Anderson et a1 who concluded that titative measures of exposure to acrylamide an occupational aetiology was probably unim- and the use of local county comparison rates portant, given the breadth of workplace experi- and robust modelling of internal cohort rates, a ences examined with few positive findings." methodological improvement on the original Additional followup may elucidatewhether the 1989 report. association is spurious or warrants further In summary, this study has many strengths examination. Among the other cancer sites and is the most definitive study of the human examined in the exploratory analysis (oesoph- carcinogenic potential of exposure to acryla- agus, rectum, and kidney), we found increased mide conducted to date. The contribution of SMRS for some categories of exposure to acry- 1115 additional deaths and nearly 60 000 lamide, but little evidence of an exposure- person-years over the 11 year follow up response relation. corroborate the original cohort study Nearly half of the cohort were short term findings-namely, little evidence of a causal workers, a characteristic not presented in the relation between exposure to acrylamide and original study findings. About 50% of the acry- mortality &omany cancer sites, including those lamide cohort worked >1 year and only 1936 of initial interest. An increased risk of mortality (23%) worked 2 10 years. Consequently, most from cancer of the respiratory system, noted in of the cohort had minimal potential for the original study, decreased considerably in exposure to acrylamide or were exposed to low the update period and remains only slightly levels. Contrary to many other occupational increased among workers exposed and unex- cohort studies, short term workers did not posed to acrylamide. Additional follow up of show a differential mortality pattern often the cohort may elucidate whether the excess of associated with increased mortality for both thyroid cancer and the association found malignant and non-malignant diseases. The between exposure to acrylamide and pancre- long duration of follow up in this study may atic cancer in an exploratoryanalysis are spuri- have mitigated the mortality influence of short ous or warrant further examination. term workers. The limitations of our acrylamide study- Thls work was supported by Cytec Indusmes.We acknowledge Steve Sefcik, Consultant and Al Johnson of Cyrec Indusmes namely a large proportion of short term who provlded computer programmingsuppon. Marsh, Lucas, Youk,et al 1 Collins JJ,Swaen GH, Marsh GM, et d.Mortality patterns among workers exposed to acrylamide. 3 Oenrp Med 1989;31:614-17. 2 Johnson KA,Gorzinski SJ, Bodner Kh4, et d.Chronic tox- icity and oncogenicity s ~ d oyf acrylamide incorporated in the drinking water of Fischer 344 rats. Tmicol Appl F'harmacd 1986;85: 154-68. 3 Bull RJ, Robinson M,Laurie RD, et d.Carcinogeniceffects of acrylamidein Sencar and A/J mice. Cancer Res 1984;U 107-11. 4 Sobel W,Bond GG,Parson TW,et d.Acrylamide cohort mortality study. BrJ IndMcd 1986;43:785-8. 5 Schall LS,Marsh GM, Henderson VL.A two-stageprotocol for verirying vital status in large historical cohort Studies.J OccupEm'mn Med 1997;391097-102.6Marsh GM, Prei- nhger ME. O C W a user-oriented occupational cohort mortality analysis program. American Statisrhbn 1980;34: 245-6. 7 Marsh GM, E h h d J, Paik M, et d. OCMAPiPC: a user-oriented cohort mortality analysis program for the IBM PC. A-an Storirrhbn 1986;40:30@-9. 8 Marsh GM,Youk AO, Sefcik SS, et d. OCMAP-PLUS: a program for the comprehensive analysis of occupational cohort data.3Orcup Em'nm Med 1998;40:351-62. 9 Monson RR.Analysis of relative survival and proportional mortality. Compuun and Biomedical Research 1974;7:325. 10 Marsh GM, Ehland J, Sefcik SS. Mmdio and PW@h data system [technical report]. Rttsburgh, PA University of Pittsburgh, Department of Biostatistics, 1997. 11 Caplan RJ, Marsh GM, Enterline PE. A generalized effective exposure modeling program for assessing doseresponse in epidemiologic investigations. Compuun and Biomedical Research 1984;16587-96. 12 Breslow NE, Day NE.77u &sign and andysis o f w h s d ies. Starirricd methodc m cancer research. Vol II. Lyon: International Agency for Research on Cancer, Oxford University Press, 1987. (IARC Sci Pub1 No 82.) 13 Cox DR. Regression models and life tables [with discus- sion].jburnd oj the Royal Sra&rical Socieo 197ZW.187- 220. 14 Cox DR. Panial likelihood. Eiomenika 1975;62:269-76. 15 EGRE'I; k i o n 1.02.09. Seattle, WA: Statistics and Epidemiology Research Corporation, 1991. 16 Cpel SoIiware. fofXact turbo: sofirvcln fm enact Mditinol logisric regression.Cambridge, MA: Cytel Soliware, 1993. 17 Anderson KE, b t e r JD, Mack TM.Pancreatic cancer (chapter 35). In: Schonenfeld D, Fraumeni JF Jr. e&. Cancer epidemwbgy and preumnim, 2nd Ld New York: Oxford University Pms,1996. 1 Correspondence and editorials Occuparional and Environmental Medicine wel- comes correspondence relating to any of the material appearingin the journal. Results &om preliminary or small scale studies may also be published in the correspondence column if this seems appropriate. Letters should be not more than 500 words in length and contain a minimum of references. Tables and figures should be kept to an absolute minimum. Letters are accepted on the under- standing that they be subject to editorial revision and shortening. The journal also publishes editorials which are normally specially commissioned. The Editor welcomes suggestions regarding suit- able topics; those wishing to submit an editorial, however, should do so only after discussion with the Editor. COWRESPONDENCE Mortality of workers exposed to methylene chloride employed at a plant producing cellulose triacetate film base EorroR-Tomenson et a? reported results from a cohort study of men exposed to methylene chloride in the manufacture of cellulose triacetate film base. One of the findingsin the study was four observed deaths from brain and other central nervous system (CNS) tumours compared with 2.8 expected cases based on national mortality rates. This is an interesting finding, as Heineman et a[-` previously reported an association between exposure to methylene chloride and astrocytic brain cancer in a case-control study of men from southern Louisiana and northern New Jersey who had died. However, Tomenson et al dismissed the finding as two of their cases "had minimal exposure to methylene chloride". They furthermore argued that "no SUppOrt for an association between methylene chloride and brain cancer was provided by the other three retrospective cohort ~tudies".~H'owever, only one of the three studies that Tomenson et al referred to reported data on brain cancer. Hearne et all found two observed deaths from brain cancer compared with 1.7 expected based on reference rates from New York State, excluding New York City, and 2.0 expected based on interval rates from Kodak, Rochester. Brain cancer data were not reported in the studies by Lanes et ClT and Gibbs et al.' The four other available cohort studiesw that included workers exposed to methylene chloride also did not report data on brain cancer. It would have been prudent for Tomenson et al to bring to the attention of the reader the lack of data on the association between methylene chloride and brain cancer. Not supported is different from not reported! ELSEBETH LYNGE Danish Cancer Society,Instinm of Cancer Epuiemwlogv, Srrandbodevarden 49, DK-2100 Copenhagen, Denmanh 1 Tomenson JA, Bonner SM, Heijne CG, et d. Mortality of workers exposed to methylene chloride employed at a plant producing cellulose triacetate film base. Occup Envimn Med 1997;544706. 2 Heineman EF, Cocco P, Gomez MR, er d. Occupanonal exposure to chlorinated aliphatic hydrocarbons and risk of astrocytic brain cancer. Am 3 Ind Med 1994;26 15549. 3 Hearne FT, Pifer JW, Grose F. Absence of adverse mortality effects in workers exposed to methylene chloride: an update. 3 O a p Med 1990;32:23440. 4 Lanes SF, Rothman KJ, Dreyer NA, er d.Mor- tality update of cellulose fiber production workers. ScandJ U%rk Emimn Hedrh 1993;19: 42f-8. 5 Gibbs GW, Amsel J, Soden K. A cohort mortality study of cellulose triacetate-fiber workers exposed to methylene chloride. 3 Occup Envimn Med 1996;38:693-7. 6 Sinks T, Lushniak B, Haussler BJ, er al. Renal cell cancer among paperboard printing workers. EprdPmwlogy 1992;3:483-9. 7 Cragle DL, Wells SM, Tankersley WG. An occupational morbidity study of a population potentially exposed to epoxy resins, hardeners and solvents. Appl Occup Emimn Hyg 1992;7: 826-34. 8 Shannon HS, Haines T, Bernholz C, a al. Cancer morbidity in lamp manufacturing workers. A m y Ind Med 1988;14:28l-90. 9 OK MG, Carlo GL, Steinberg S, et ai. Mortality among employees engaged in chemical manu- facturing and related activities. Am 3 E p d m i d 1985;122:311-22. Authors' reply-Lynge is correct to point out that Gibbs et all and Lanes et aF did not report brain cancer mortality in two cohorts of United States cellulose triacetate fibre workers in Cumberland, Maryland, and Rock Hill, South Carolina. At the time of writing our paper,' the paper by Gibbs et al' had not been published. However, we had access to a more detailed report` of the mortality experience of workers at the Cumberland facility. For comparison purposes,Gibbs' also reports some mortality findings for workers at the Rock Hill factory.When the paper of Gibbs et al` was published, we removed the reference to the unpublished report.` Unfortunately we overlooked the fact that the published paper did not include the brain cancer findings which we now give. Gibbs et d`and Gibbs` reported mortality results for workers in two groups exposed to methylene chloride. Among workers in the high exposed group (350-700 ppm) there was one death from cancer of the central nervous system (CNS) (2.01 expected) and three deaths from cancer of the CNS (2.54 expected) in the low exposure group (50-100 ppm)? Exposure levels of workers in the Rock Hill cohort studied by Lanes et al,'were similar to those of workers in the Cumberland cohort, but it was not possible to separate them into exposure categories. There was one death from cancer of the CNS (1.52 expected) in the Rock Hill workers.` Hence these studies do not provide support for an association between methylene chloride and brain cancer. However, we apologise for not ensuringthat readers had access to the relevant data. When discussing the supporting evidence for an association between brain cancer and exposure to methylene chloride, we focused on the information provided by the four cohorts of workers producing cellulose triac- etate fibres or film base.'.' ' These workers had high and well characterised exposure to methylene chloride compared with workers in the four other cohort studies cited by Lynge,"' which did not report brain cancer mortality, but they have limited relevance to an assessment of the human health effects of exposure to methylene chloride. JOHN A TOMENSON SUSAN M BONNER ICI Epldemwlogv Unit, Nonhwich, Cheshire, CW8 4D3;UK COLIN G HEIJNE DAVID G FARRAR ICI Chemicals and Polymers,Runcorn, Cheshire, WA7 4QE UK TREVOR F CUMMINGS ICI Chemicals and FWymers,Middlebomugh, Cleveland, TS90 83A, UK 6 Sinks T, Lushniak B, Haussler BJ, et al. Renal cell cancer among paperboard printing workers. Epidemiologv 1992;3:483-9. 7 Cragle DL, Wells SM, Tankersley WG. An occupational morbidity study of a population potentially exposed to epoxy resins, hardeners and solvents.Appl Occup Enoron Hyg I992;7: 826-34. 8 Shannon HS,Haines T, Bernholz C, et ai. Cancer morbidity in lamp manufacturing workers. AmJInd Med 1988;14281-90. 9 Ott MG, Carlo GL, Steinberg S, et al. Mortality among employees engaged in chemical manufacturing and related activities. A m 3 Epidemwl 1985;122:311-22. Inhalation of ammonium nitrate fuel oil explosive: and possible concomitant exposure Author's reply-Sostrand`has suggested that the nitrogen dioxide component of diesel engine exhaust fumes may have caused the symptoms detailed in the case of ammonium nitrate fuel oil explosive (ANFO) inhalation that I reported recently? This is unlikely for the following reasons: 0 Symptom onset after inhalation of the ANFO plume was immediate 0 At the time of the ANFO inhalation the miner was charging a face at ground level and the diesel powered platform truck had been switched off for about 15 minutes 0 There were no other diesel units operating in the area 0 Three other miners in the same area who were not exposed to the ANFO plume were unaffected 0 The affected miner had regularly worked in very similar conditions for 9 years without symptoms. Sostrand has correctly indicated that diesel fuel is commonly used in ANFO. As I indicated in my previous report either the hydrocarbon solvent fuel oil, or the ammonium nitrate, or possibly both of these components of ANFO, may have caused the symptoms described.' A MICHAEL DONOGHUE The Medical Centre,Mounr Isa Mines Limited, Mounr Ira, Queensland4825,Australia 1 SostrandP.Inhalation ofammonium nitrate fuel oil explosive (ANFO): and possible concomitant exposure. Occup Emiron Med 1998;55:647. 2 Donoghue AM. Inhalation of ammonium ni- uate fuel oil explosive (ANFO).Occup E m i m Med 1998;55:144. CORRECTION 1 Gibbs GW, Amsel J, Soden K. A cohort mortality study of cellulose triacetate-fiber workers exposed to methylene chloride. 3 OccupEmiron Med 1996;38:693-7. 2 Lanes SF, Rorhman KJ, Dreyer NA, et al. Mortality update of cellulose fiber production workers. ScandJ Work Emiron Health 1993;19: 426-8. 3 Tomenson JA, Bonner SM, Heijne CG, er al. Mortality of workers exposed to methylene chloride employed at a plant producing cellulose niacetate film base. Occup Envrmn Med 1997;54:470-6. 4 Gibbs GW. The mortality of workers employed at a cellulose acetate and triacetate fibers plant in Cumberland, Maryland: a 1970 cohort followed 1970-89. A Final Report to the Hoechst Celanese Corporation, May 1992. 5 Hearne FT, Pifer JW,Grose F. Absence of adverse mortality effects in workers exposed to methylene chloride: an updare. 3 OcCuP Med 1990;32:234-40. Risk of pulmonary tuberculosis relative to silicosis and exposure to silica dust in south African gold miners by E HNIZDO, J MURRAY (1998;55:496-502). T h e cumulative dust exposure meaSWementS given in this paper were in units of ghm-'and not in mg/m'-y as incorrectly stated. The cumulative dust exposure in ghn-'units was calculated as a sum of products between the number of shifts spent in an occupational category, the average number of hours spent underground in the occupational category, and the average respirable dust concenvatjon (mg/m') for the occupational category. TO convert the cumulative dust fromghm4 into mg/m'-y one needs to multiply the average 216 Correspondence;Correction, Notices exposure of 14.3 given in the paper by 1000 and divide by (270x8) to obtain 6.6 mg/m'-y. We are grateful to Professor HJ Woitowitz from the Institute of Work and Social Medicine in Giessen, Germany, for pointing out this error. NOTICES International Course of Molecular Epidemiology. 19-24April 1999. V i a Gualino, Torino,Italy. The Institute for Scientific Interchange Foundation, the International Agency for research on cancer, the University of Torino, the Centre for Oncologic Prevention, and the Italian Molecular Epidemiology Group are running this course. Keynote lectures 0 Current perspectives in cancer research and prevention The aetiology of common diseases and their pathogenetic pathways Session 1-molecular dosimetry: techniques and methods 0 DNA adducts and protein adducts 0 From adduction to DNA damage Session 2-genetic susceptibility 0 high and low penetrance cancer genes 0 metabolic polymorphisms 0 DNArepair Session 3-epidemiological methods Study design: transitional and formal studies 0 Case-control,cohort studies 0 Bias, confounding, and effect modification Session +DAN damage 0 Acquired mutations and mutational spema 0 Cytogenetic damage Session S-statistical analysis Exploratory data analysis 0 Univariate analysis 0 Multivariate analysis Session &issues in molecular epidemiology 0 Ethical aspects 0 Combined evidence, meta-analysis 0 Causality assessment Course directors: P Vineis (Torino), P Boffetta (Lyon). Faculty (provisional list): N Rothman (Bethesda), M Berwick (New York), D Phillips (Sutton), M IngelmanSundberg (Stockholm), F Real (Barcelona), P Hainaut (Lyon), R Montesano (Lyon) Keynote speakers: P Kleihues (Lyon), F Perera (New York) No registration fee is required. The IS1 Foundation has reserved single rooms for participants from April 18 to 24 at Vila Gualino. The full board package is ITL. 780.000. Deadline for registration: 28, February 1999 For information contact: IS1 Secretariat, Villa Gualino, Viale Semmio Severo 65, Torino, Italy. TelOO39 11 6603090, fax 0039 11 6600049; email: isi@isi36a.isi.it - ww- w.isi.it EPICOH 14* InternationalConference on Epidemiology in Occupational Health. Henlia, Israel. 1 b 1 4 October 1999. chophysiological Studies on Combined Envimnmental Factors, Methodological Issues and Cmpurer Simulation as weU as Modelling. The programme will include introductory lectures, oral and poster presentations. The working language of the Conference is English. Further details from: Dr Olavi Manninen, The International ISCES Society, c/o Labour Protection Depzrtment, PO Box 272, FIN33101 Tampere, Finland.-Tel 00358 3 260 8820; fax 00358 3 2608899; email: Olavi.Manninen@tsp.stxn.vn.fi The ICOH together with the Occupational Health and Rehabilitation Institute in Israel is organising the EPICOH 14th International Conference on Epidemiology in Occupational Health. The main topics of the conference are: 0 Occupational work related diseases 0 Occupational cancer 0 Occupational epidemiology 0 Molecular epidemiology 0 Biological markers 0 Value of pre-employment examinations 0 Exposure assessment 0 Reproductive health studies 0 Communicable diseases For information contact:DrJudith Shaham, Occupational Health and Rehabilitation Institute, PO Box 3, Ra'anana, Israel. Tel 00 972 9 771 0092; fax 00 972 9 771 2212; email judiths@trendline.co.il International Conference on The Combined Effects of Environmental Factors ICCEF 2000.26-29 August 2000, Savonlinna, Finland. The 9th international conference on the combined effects of environmental factors will take place in Savonlinna,in the beautiful eastern lake area of Finland. The conference is organised under the auspices of the International Society of Complex Environmental Studies (ISCES) in cooperation with national and international research institutes and universities. The ICCEF 2000 invites all scientists who are interested in basic or applied studies related to complex interactions and combined effects of physical, chemical, psychosocial, and biological environmental factors on human health and wellbeing, or on biological systems. Contributions cover the following main subject areas Complex Environmental Studies, Combination E@mioIo#, Combinarion Toxicology, Combined Mutagens, Gemurnins, Carcinogens and Teratogens, Combined Environmental and Psychosocial Factors, Psy- I i