Document Exp0e3eVr6wpNVwx751jZZQLn

Reprinted from Journal of Occupational Medicine Volume 25. No. 5, May 1983 R&S 041165 A Case-Control Study of Brain Tumor Mortality at a Texas Chemical Plant Gregory G. Bond, M.P.H.; Ralph R. Cook, M.D., Peter C. Wight, Ph.D.; and George H. Flores A plant-based case-control study was undertaken to in vestigate a possible excess of brain tumor mortality identi fied at a Texas Chemical plant from a sample-based cohort study. Work histories and presumptive exposures of 28 former employees who had died of primary intracranial neoplasms were contrasted with those of two matched com parison groups in an effort to identify a possible etio/ogic agent. Because the sample-based cohort study suggested they might be at increased risk, those employees hired prior to 1945, those employed from one to four years and those employed for 20 or more years were studied separately. No statistically significant associations were found, although an elevated risk was suggested for employment in the machine shop prior to 7 945 based on three exposed cases. ixesults from a recent retrospective mortality study of current and former employees of the Dow Chemical U.S.A., Texas Division have indicated that, while an excess risk for brain tumor probably does not exist for the overall popula tion, an apparent excess risk may exist for those first employed prior to 1945.1 Among all hires, a possible excess appeared confined to those employed from one to four years and to those employed greater than 20 years. A plantbased case-control study was undertaken in an attempt to localize the risk to a particular process area, chemical or physical exposure agent, or job classification. The present investigation differs somewhat from a case-control study reported earlier,2 due to more extensive case ascertainment and a more refined case definition. In addition, the present study considers separately the various subsets of cases From Dow Chemical U.S.A., U.S. Area Medical, 1803 Building, Midland, Mich. (Mr. Bond, Research Epidemiologist; Dr. Cook, Director of Epidemiology); the Texas Division, Dow Chemical U.S.A., Freeport (Dr. Wight, Director of Epidemiology and Mr. Flores, Industrial Hygienist). Dr. Wight Is now in Portland, Ore. Address correspondence to Health and Environmental Sciences, 1803 Building, Dow Chemical U.S.A., Midland, Ml 48640 (Mr. Bond). Journal cf Pccuoational Wsdicine/Vo1. 25. No. .c7Mav 1983 identified from the retrospective mortality study as pos sibly at higher risk. There are few known risk factors for brain neoplasms in humans. A review of the epidemiology of brain tumors has been recently published3 and is recommended for a more thorough discussion. Incidence and mortality rates rise steeply after age 30 to a peak between 55 and 64 years, and are relatively higher among whites and males.4,5 Within the past few years, several toxicologic and epidemiologic re ports have focused on the possible etiologic role of chemi cals and, to a lesser extent, viruses. Experimental systems for inducing gliomas, a common histological type of brain tumor, by direct injection of chemical carcinogens and viruses have been summarized.6 Animal studies have suggested that acrylonitrile can induce CNS tumors,7 and controlled exposures to extremely high levels of vinyl chloride (2,500 to 10,000 parts per million) have produced brain tumors in rats.8 Other experimental brain carcino gens, such as some of the nitrosoureas, are recognized and have been summarized elsewhere.9 The epidemiologic reports have been less definitive. A study of records maintained by the Oil, Chemical and Atomic Workers Union noted a proportionate excess of brain tumors among union members, but was unable to link the observations to specific exposures.10 Excess mor tality due to brain neoplasms has been reported among other petroleum refinery and petrochemical employees11,12 and among a cohort of Swedish chemists.13 Other mor tality studies of employees in these industries have failed to demonstrate an excess risk for brain neoplasms.14"17 Thus, it Is not clear whether the elevated brain tumor risks ob served in some petrochemical plants, if real, are isolated incidents or indicative of the effects of a widely dissem inated brain carcinogen. It was the purpose of this investigation to determine whether or not a risk for brain tumor mortality could be associated with a specific job assignment, with employment in a particular process area, or with presumptive exposure to a chemical or physical agent. Because no a priori hy pothesis existed regarding a specific exposure agent, this ^77 Table 1 - Characteristics of Brain Tumor Cases Year Year Age Case of of at No- Hire Death Death Usual Job at Dow Duration of Employment* Time Interval Since Entry* Tumor Type {Source)^ 01 1940 1975 61 Machinist 02 1941 1958 39 Machinist 03 1943 1951 48 Operator 04 1943 1960 56 Laborer 05 1950 1974 48 Mechanic 06 1941 1952 36 Operator 07 1941 1965 49 Chemical engineer 08 1942 1973 68 Metal dipper 09 1941 1962 53 Painter 10 1943 1966 62 Roustabout 11 1942 1969 58 Clerk 12 1971 1977 69 Research scientist 13 1941 1977 57 Manager 14 1943 1961 67 Paint shop foreman 15 1941 1971 59 Production foreman 16 1941 1961 49 Operator 17 1943 1964 60 Pipefitter 18 1950 1977 61 Material handling technician 19 1941 19S8 55 Boilermaker-welder 20 1943 1977 51 Boilermaker 21 1968 1976 38 Operator 22 1943 1965 62 Oiler 23 1946 1970 46 Operator 24 1940 1945 27 Pipefitter 25 1944 1970 64 Metal dipper 26 1942 1976 65 Machine shop foreman 27 1953 1963 36 Rigger 28 1945 1973 67 Laborer 13;3 3;3 8;11 1;1 1;6 4,-5 4;2 1:2 0;1 17* 0;3 Q;11 35:3 14* 30;4 19;4 21 ;S 27* 27;11 31:2 8:2 0:5 23;11 2;7 1:5 34 ;2 10 ;2 0;1 34;5 17;10 8:11 17,-4 23:6 10;10 23:1 Q 30;1Q 21:1 23;5 26;11 5.6 35;4 17* 30;6 19:4 21* 27* 27;11 34;4 8:2 21:4 23;11 4;1 25;1Q 34;6 10;5 27* Astrocytoma, grade IV (SPR) Glioblastoma multiforme (SPR) Brain tumor (MR) Glioblastoma multiforme (DC) Oligodendroglioma (SPR) Carcinoma of brain (OC) Astrocytoma, grade HI (Aut) Astrocytoma, grade IV (MR) Glioblastoma multiforme (MR) Brain tumor (DC) Glioblastoma multiforme (DC) Glioblastoma of frontal lobe (MR) Medulloblatoma (MR) Astrocytoma, grade Hi (Aut) Malignant glioma (MR) Glioblastoma,Rt temporal lobe (MR) Brain tumor malignant (DC) Malignant brain tumor (MR) Astrocytoma, grade IV (MR) Carcinoma of brain (MR) Glioblastoma multiforme (MR) Glioblastoma multiforme (SPR) Glioblastoma multiforme (MR) Cancer of brain (DC) Astrocytoma (MR) Glioblastoma multiforme (MR) Malignant brain tumor (MR) Glioblastoma (OC) Years; months * Time interval between date of hire and date of death t MR indicates medical records; OC, death certificate; SPR, surgical pathology report; AUT, autopsy Death certificate did not indicate presence of intracranial neoplasm Control group A and the cases were distributed quite dif ferently, and this no doubt reflects the weighting of these controls toward employees with longer company service. There is a larger percentage of persons who attained the craftsmen and foremen category in this control group, and a smaller percentage represented at the higher (manager and administrator, professional and technical) and lower (cleri cal and laborers) ends of the job classification scale. A total of 79 specific chemical and physical agents were listed as presumptive exposures for cases or controls by the industrial hygienists. Tables 6 and 7 present the distribution of cases and control groups A and B, respectively, by pre sumptive exposure to specific agents. None of the matched odds ratios for exposure to any of the agents was signifi cantly elevated. Those agents to which cases and controls had the most frequent potential for exposure were chlorine, hydrogen chloride, and sulfur dioxide. Exposure to these agents was common among employees who were assigned to magnesium production and the frequency of exposure among cases and controls reflects the relatively large pro portion of these employees who had been assigned to the Magnesium Department. Exposure to solvents, such as carbon tetrachloride (no longer used for such a purpose), mineral spirits and toluene was common among employees who were engaged in a specific craft, e.g., machinists. Ex posure to the majority of the remaining chemical agents was less frequent- This is not surprising given the large num ber of individuals employed and the diversity of chemical operations over the years. As mentioned earlier, a retrospective mortality study of a 5% random sample of the ever employed population suggested that any brain tumor mortality excess, if present, was confined to two subgroups of the population: those employed from 1 to 4 years and those employed 20 or more years. There was also a suggestion that the risk was further confined to those individuals first employed prior to 1945, although there were too few observed and expected deaths among those hired later to state this unequivocally. Each of the subgroups suggested to be at increased risk was further explored by the case-control method in an attempt to identify a specific etiologic agent in the work environment that may have been responsible. e.c 041168 380 Brain Tumor in a Chemical Plant/Bond at al cally significant, elevated risk found earlier for having been a salaried employee at termination of employment,2 job classification titles were grouped into six socioeconomic categories following those developed by the Bureau of the Census. A total of 269 job titles were reduced to six cate gories ranked from presumed highest to lowest socio economic group: managers and administrators (high-level management); professional, technical and kindred (chem ists, research personnel, and engineers); craftsmen, foremen and kindred (skilled trades and first-line supervision); operative (inorganic and organic chemical production personnel); clerical and kindred (clerks); and laborers and kindred (unskilled). Cases and controls were compared for highest job category achieved and for ever having been assigned to any of the job categories. Classification of subjects by work area, or by job title can be a somewhat crude method for comparing exposure histories and may obscure an association with a chemical or physical agent exposure that may be common to several work areas or jobs.20 Thus, subjects in this study were also classified by presumptive exposure to specific chemical or physical agents. In the previous study,2 the focus had been mainly on comparisons of presumptive exposures to process chemicals. This was expanded in the present investigation to compare cases and controls with respect to presumptive exposure to the main agents, including those materials not directly related to chemical production. On-site industrial hygienists established chemical and physical agent-specific exposure profiles for each subject based on the work history and job function. The presump tive exposures for the job function were developed through interviews with employees knowledgeable of the process and the job, and knowledgeable of process changes and the use of available survey data. This was done with the indus trial hygienist being "blind" to whether the subject was a case or a control, although the identity of the subject was known to the industrial hygienist. Identity of the subject permitted the industrial hygienist to question former co-workers concerning job duties, whenever there were ambiguities in the work history. In some instances, the department and/or the job titles were too broad to establish an individual's work experience; interviews with former co-workers offered the only identi fiable source of information to determine the subject's department assignment. While it is quite possible the coworkers were aware of the case-control status of the subject when provided his identity, it is not believed a systematic bias in misclassification of exposure resulted. No attempt was made to reveal to the interviewee the reason for the inquiry, and the size of the plant and number of individuals questioned would make a systematic bias unlikely. Analytical Strategy -- Both department assignment and presumptive chemical exposures were regarded as dichoto mous variables (ever or never assigned, ever or never ex posed) for analytical purposes. Since for both control groups A and B a full complement of four controls could not be matched to each case, an analysis for a data set with a variable matching ratio was performed.21 Point estimates of the odds ratio were calculated,22 and confidence limits for the odds ratio were taken as test-based limits,23 based on the point estimates and the x from hypothesis testing. In some instances, the matching was dissolved and the data Journal of Occupational Medicine/Vol. 25, No. 5/Mav 1983 were stratified by variables into 2 x R tables to assess interaction and to control confounding.24 Pooled point estimates and x values were calculated and an asymptotic likelihood ratio test25 was employed to test the hypothesis that the odds ratio was uniform across the strata (test for homogeneity). All analyses were done on a programmable calculator with a packaged set of programs.26 Results Twenty-eight decedents met the criteria for having a primary intracranial neoplasm and constituted the case population. Selected demographic and employment in formation is presented for each of the cases in Table 1. The median age at death was 57 years, with a range from 27 to 69 years. The deaths occurred over a 32-year period, 19451977, and no additional brain tumor deaths were observed among former employees through 1979, even though case ascertainment has been updated through this period. The distribution of cases by tumor type reveals the majority (71%) of the neoplasms to be of glial origin. Nearly 58% of all the intracranial neoplasms were indicated to be glio blastomas (this includes the astrocytomas grades III and IV) and 29% were of an unspecified nature. The matching technique employed resulted in 110 decedents in control group A and 111 individuals in control group B successfully matched to the 28 cases. Table 2 pre sents some descriptive data by case-control status. The cases and controls were quite similar with respect to all factors on which they were matched. The only possible exception is that the cases were employed on the average 13.4 manmonths longer than control group B. This difference would have likely been smaller if the matching had been more stringent among those with more than 20 years of employ ment and had consideration been given to incorporating service breaks in the matching process. There was an even larger difference in the mean duration of employment between the cases and control group A; (41 man-months) the greater duration observed among the latter reflects the proportionately large share of vested employees among the deaths known to the company. The mean year of hire among the cases and the two control groups was similar, although the controls not matched on length of employment (control group A) began employ ment, on the average, later than did the cases. Matched odds ratios for department assignment are presented for the comparisons with control groups A and B in Tables 3 and 4, respectively. None of the odds ratios are significantly elevated, and there is a lack of consistency in direction between the odds ratios for the comparisons with the two control groups. For instance, there is nearly a two fold risk of brain tumor and having been assigned to a mag nesium department in the comparison with control group A. It would appear, however, that perhaps length of em ployment is a confounding variable, because the odds ratio is close to unity in the comparison with control group B which was matched to the cases on length of employment. The distribution of cases and both control groups by highest job classification achieved may be found in Table 5. The distributions by ever having been assigned to a job classification were similar. Generally, the cases and controls from group B were distributed across job classification cate gories nearly identically. 379 R&S 041167 UZ.M.PU b 'V ti Table 5 - Distribution of Highest Job Classification Achieved by Case-Control Status Job Classification Cases Control A Control B Managers and administrators Professional and technical Craftsmen and foremen Operatives Clerical Laborers and other unskilled 2 (7.1)* 2 (7.1) 13 (46.4) a (28.6) 1 (3.6) 2 (7.1) 4 (3.6) 3 (2.7) 66 (60.0) 30 (27.3) 2 (1.8) 5 (4.5) 5 (4.5) 8 (7.2) 49 (44.1) 33 (29.7) 4 (3.6) 12 (10.8) Total 28 (100.0) 110 (100.0) * Numbers in parentheses are percent of total 111 (100.0) Table 6 -- Matched Odds Ratios for Presumptive Exposure (Ever) to Specific Chemical and Physical Agents: Comparison With Control Group A* Chemical or Physical Agent No. of No. of Exposed Exposed Cases Controls Odds Ratio ao% Confidence Limits Chlorine 14 48 1.33 0.65-2.71 Hydrogen chloride 13 42 1.40 0.70-2.80 Sulfur dioxide 11 26 2.02 0.994.11 Carbon tetrachloride 5 20 1.00 ...t Noise 5 40 0.26 0.10-0.67 Mineral spirits 4 9 NC* NC Welding fumes 4 24 NC NC Cutting oils 4 9 NC NC Sodium hydroxide 3 12 NC NC Crystalline silica 3 9 NC NC Toluene 3 11 NC NC * There were 68 other chemical or Physical agents n which cases or controls had presumptc exposure; hovver, the total number of cases and controls with presumptive expo sure to each was less than 20 and no more than two cases had presumptive exposure to any of these agents * Test based confidence intervals will not compute when xmh20 * NC indicates not calculated because there were fewer than five exposed cases 382 Table 7 -- Matched Odds Ratios for Presumptive Exposure (Ever) to Specific Chemical and Physical Agents: Comparison With Control Group 8* Chemical or Physical Agent No. of Exposed Cases No. of Exposed Controls Odds Ratio 90% Confidence Limits Chlorine 14 55 1.03 0.81-1.31 Hydrogen chloride 13 51 1.02 0.81-1.29 Sulfur dioxide 11 39 1.19 0.58-2.43 Noise 5 22 0.92 0.35-2.45 Carbon tetrachloride 5 14 1.50 0.60-3.74 Welding fumes 4 14 NC* NC Cutting oils 4 10 NC NC Toluene 3 16 NC NC Sodium hydroxide 3 15 NC NC Crystalline silica 3 10 NC NC There were 68 other chemical or physical agents to which cases or controls had presumptive exposure; however, the total number of cases and controls with presumptive expo sure to each was less than 20 and no more than two cases had presumptive exposure to any of these agents * NC indicates not calculated because there were fewer than five exposed cases cases had a work history that was suggestive of this expcvsure. Another subcohort suggested to be at an increased risk for brain tumor was that group of employees employed from 1 to 4 years. There were seven cases who were em ployed for from 1 to 4 years, so again the relatively small sample size precludes many meaningful comparisons. Table 9 presents the exposure variable comparisons between these cases and their matched controls from group B. Only the odds ratio (OR) for having been assigned to a Magnesium Department is elevated for this group (OR = 2.67, 90% confidence limits = 0.74 - 9.62); however, this was not statistically significant. As mentioned earlier, the greatest risk for mortality from brain tumor appeared confined to the subcohort hired prior to 1945. This association with employment in the early years of the Texas Division was explored further by restricting the case-control analysis to department assign ments held prior to 1945 ind by dissolving the matching among the cases and controls. There were only four depart ments in which at least three cases had held an assignment prior to 1945; Magnesium, Administrative Services, Ma chine Shop and Pipe or Boiler Shop. Odds ratios were sightly elevated, but not significant, for having been as signed to Magnesium (COR - 1.20, Fisher27 exact upper p value = 0.46) or to the Pipe or Boiler Shop (COR ~ 1.38, Fisher exact upper p value = 0.46) prior to 1945. The asso ciation with having been assigned to the Machine Shop prior to 1945 was even stronger (COR - 5.3, Fisher exact upper p value = 0.09); however, there were only three cases with such a history. According to available records, all Brain Tumor in a Chemical Plant/Bond et al R&S 041169 Table 2 - Descriptive demographic Data by Case-Control Status Demographic Variables Cases (N=28) Mean - SO Control Group A (N=110> Year of birth Year of death No. of man-months employed Year of hire 1912.6i 9.6 1967 i 8.6 143.0 i 141.5 1945.1 i 7.6 1912.0* i 9.5 1967* i 8.5 184.0 i121.3 1947.2 28.I * Controls were matched to cases on these variables * Only 40 of 111 controls were known deceased as of Dec. 31, 1977 Control Group B (N=111) 1913.0* 2 9.4 1968* 2 8.3 129.6* 126.2 1945.5 2 5.7 First, since it was apparent there might be an increased risk for having been employed for more chan 20 years we examined the department assignments, job classification distribution and presumptive exposure to chemical and physical agents among this subset of cases and their con trols from group B. Comparisons with control group A were considered less valid due to the weighting of this group toward employees with longer service and have been ig nored. Of the nine observed deaths noted among this sub cohort in the sample-based cohort study, two were deter mined to not have been primary intracranial neoplasms on review of available medical records and histopathology and Table 3 - Matched Odds Retios for Oepcrtment Assignment (Ever): Comparison With Control Group A* Department No. of Exposed Cases No. of Exposed Controls Odds Ratio SOS Confidence Limits Maintenance 6 45 0.39 0.18-0.87 Magnesium 11 27 1.93 0.95-3.93 Pipe and Boiler Shop 5 27 0.66 0.27-1.61 Administrative Service 5 20 0.97 0.75-1.25 Paint Shop 2 14 NC* NC Power 1 15 NC NC Styrene 3 11 NC NC Chlorine 2 12 NC NC Chlorinated Methanes 0 12 NC NC Materiel Handling 2 9 NC NC Yard and Utility 2 9 NC NC Machine Shop 3 7 NC NC * There were 25 other departments in which at least one case or control had worked; however, the total number of cases and controls assigned to each was fewer than 10 and no more than two cases had one of these departments in common * NC indicates not calculated because fewer than five cases had exposure Journal of Occupational Medicine/Voi. 25, No. 5/May 1983 have been excluded from the case-control study. An addi tional case whose death certificate indicated a lesion of the brain stem and who had been employed greater than 20 years was included when medical records revealed a primary intracranial neoplasm. Table 8 presents the comparisons of cases employed greater than 20 years and control group B with respect to several exposure variables. There were elevated risk esti mates associated with having been employed in the Pipe or Boiler Shop, in a Maintenance area, or in Administrative Services. The small number of subjects with a common exposure history makes interpretation of these data diffi cult. There was an elevated risk for having been potentially exposed to carbon tetrachloride; however, only two of the Table 4 -- Matched Odds Retios for Department Assignment (Ever): Comparison With Control Group B* Department No. of No. of Exposed Exposed Cases Controls Odds Ratio SOS Confidence Limits Magnesium 11 40 1.15 0.56-2.36 Administrative Services 5 27 0.68 0.28-1.66 Maintenance 6 25 0.92 0.41-2.06 Pipe and Boiler Shop 5 15 1.45 0.55-3.82 Paint Shop 2 16 NC* NC Machine Shop 3 11 NC NC Chlorine 2 12 NC NC Caustic 2 10 NC NC Yard and Utility 2 10 NC NC Styrene 3 6 NC NC Material Handling 2 7 NC NC * There were 24 other department* in which at least one case or control had worked; however, the total number of cases and controls assigned to each was fewer than nine and no more than two cases had one of these departments in common * NC indicates not calculated because fewer than 5 cases had exposure 381 R&S 041172 Table 10 - Exposure Variable Comparisons of Glioblastoma Cases With Matched Controls Prom Group B Exposure Variable No. of Exposed Cases (N=16) No. of Exposed Controls (N=63) Department assignment (ever)* Magnesium Administrative Services Maintenance Pipe or Boiler Shop Machine Shop 7 22 (140; 0.6-3.4)t 4 14 3 12 7 34 Presumptive exposure to chemical or physical agent (ever) Chlorine Hydrogen chloride Sulfur dioxide Noise Carbon tetrachloride 8 28 (1.2S; 0.5-3.1) 7 28 (0.97;0.4-2.4) 7 22 (1.40:0.6-3.4) 2 12 36 Job classification (highest achieved) Managers and administrators Professional, technical and kindred Craftsmen, foremen and kindred Operatives Clerical Laborers 2 5 6 4 5 21 (0.91:0.3-2.5) 21 (1.17; 0.5-2J) 2 10 There were 25 other departments in which at least one case or control had worked; however, the total number of cases and controls assigned to each was fewer than seven and no more than one case had one of these departments in common * Numbers in parentheses are odds ratio; 90% confidence inter val, computed only when there were five or more cases with a history of exposure The apparent elevated risk for an assignment in a Mag nesium Department seen in the comparison of cases with control group A was noticeably absent in the comparison with control group B. This appears to have been due to the difference between the cases and control group A with respect to length of service with the company. The Mag nesium Department is often an entry level area for unskilled labor. During the war years, and in the immediate postwar period there appears to have been a great turnover in personnel in this production area. Thus, there were many employees with relatively short duration of service to the company, most or all of it in the magnesium production 384 areas. Because control group A was somewhat biased toward employees with longer service, it was overrepre sented by skilled employees (e.g., craftsmen) and under represented by the unskilled. This bias apparently led to a positive association with having been employed in a mag nesium department. Restricting the case-control analysis to subgroups of cases identified from the sample-based cohort study as being at increased risk appeared warranted. There were several of these subgroups, such as those employees with from one to four years of service, those with more than 20 years of service and those employees first employed prior to 1945. Analyses of these subgroups failed to identify a significant risk for assignment to any particular department, or presumptive exposure to a chemical or physical agent; although in some subsets the number of cases was small. The apparent increased risk for having been employed in the Machine Shop prior to 1945 is interesting and may be indicative of an exposure or other factor which was unique to having been employed in this area during the period. Since only three cases had a history of employment in this area, it is difficult to evaluate this association. Decoufle28 noted a slight excess mortality from CNS neoplasms (five observed v 3.3 expected) among a cohort of machinists employed five or more years in oil mist-exposed jobs; although none of these deaths were among those considered to have been heavily exposed. jarvholm et al29 examined cancer incidence and mortality among 792 Swedish men employed at sometime between 1950 and 1966 and who had worked at least five years in jobs with oil mist expo sures. There were no brain tumor cases observed among the 43 incident cancers noted over the follow-up period 19581976. While excess cancer of other siies has been reported among machinists by others,30-31 CNS neoplasms were not noted to have been observed in excess of expected. It is generally assumable that industrial hygiene condi tions in the chemical industry in the early 1940s were quite different than they are presently. Other than the three cases noted to have been employed in the Machine 5bop, the lack of a common documented work experience among the subgroup of cases employed during this period does not allow one to focus the risk on any particular depart ment, job classification or presumptive exposure. The effects of a general exposure that pervaded all operations might not be discriminated with the use of the case-control methodology, however. The early 1940s also represented an unusual period in terms of hiring practices in American industry. As World War II progressed and young men were drafted into military service, industry began hiring those not eligible to serve, including females and males who were selected out of the military. The effect of this selection process on the mor tality experience of industrial cohorts is not clear, but could be playing some role. As further studies are done on other industrial cohorts, it will be interesting to note whether or not this relationship, with first employment in the early 1940s, also emerges. Due to extensive benefits programs offered by many large employers, industrial cohorts usually have access to more sophisticated medical care than does the general population. Differences between the mortality experience of these cohorts and the general population have typically Brain Tumor in a Chemical Plant/Bond et al R&S 041171 Table 8 - Exposure Varieble Comparisons of Cases Employed > 20 Years With Matched Controls From Group 8 Exposure Varieble No. of Exposed Cases No. of Exposed Controls Department assignment (ever}* Maintenance Magnesium Pipe and Boiler Shop Administrative Services Caustic Chlorine (N=8) 4 2 3 2 2 2 (N=31> 13 10 7 13 4 2 Presumptive exposure to chemical or physical agent (ever) Chlorine Hydrogen chloride Sulfur dioxide Noise Welding fumes Carbon tetrachloride Cutting oils 5 19 (1.13; 0.34.1)* 4 17 2 12 3 11 35 23 22 Job classification (highest achieved) Managers and administrators Professional, technical and kindred Craftsmen, foramen and kindred Operatives Clerical laborers 1 1 4 2 0 0 4 2 IB 7 0 0 * There were 25 other departments in which at least one can or control had worked; however, the total number of cases and controls assigned to each was fewer than five and no more than one case had one of these departments in common * Numbers in parentheses are odds ratio; 90% confidence inter vals computed only when there were five or more cases with a history of exposure three of these individuals had malignant intracranial tumors and these were indicated to have been glioblastomas. Restricting the case-control analysis to the 25 malignant intracranial neoplasms had no material effect on the results. Further restricting the analysis to cases indicated to have been glioblastomas (including the astrocytomas, grades III and IV) yielded the findings in Table 10. Elevated, but not significant, odds ratios were noted for having been assigned to a Magnesium Department (OR = 1.40) or the Machine Shop (OR = 4.16). There were elevated odds ratios for having been potentially exposed to chemicals closely associated with these departments; chlorine, hydrogen chloride, and sulfur dioxide and carbon tetrachloride, respectively. Table 9 - Exposure Varieble Comparisons of Cases Employed One to Four Years With Matched Controls From Group B Exposure Variable No. of Exposed Cases No. of Exposed Controls Department assignment (ever)* Magnesium Paint Shop Administrative Services Chlorine Machine Shop Maintenance (N=8) 5 0 (N=32) 12 (2.67; 0.7-9.6)* 6 4 5 3 3 Presumptive exposure to chemical or physical agent (ever) Chlorine Hydrogen chloride Sulfur dioxide Carbon tetrachloride Nois8 4 19 5 15 (2.02:0.5-8.1! 5 12 (2.67; 0.7-9.6) 26 3 Job classification (highest achieved) Managers and administrators Professional, technical and kindred Craftsmen and foremen Operatives Clerical laborers 1 0 3 3 0 1 0 0 13 13 2 4 * There were 10 other departments in which at least one case or control had worked; however, the total number of cases and controls assigned to each was fewer than four and no more than one case had one of these departments in common * Numbers in parentheses are odds ratio; 90% confidence inter val, computed only when there were five or more cases with a history of exposure Discussion This case-control study of 28 cases of primary intra cranial neoplasm failed to identify a significantly elevated risk with a job assignment in any particular department, or with presumptive exposure to any specific chemical or physical agent. Furthermore, the distribution of the cases and controls by job classification category did not differ materially. This lack of association is not surprising, given that the estimation of expected deaths derived from the sample-based cohort study suggests that overall there is, at most, only a slight excess of brain tumor deaths among the Dow, Texas Division work force. Without an overall excess, one would expect the job and department assignments and exposures of the cases to be distributed much the same as that of the entire work force. Journal of Occupational Medicine/Vol. 25, No. 5/May 1983 383 11. Theriault G, Goulet L: A mortality study of oil refinery workers. JOM 21: 367-370, 1979. 12. Glioblastoma cluster in a chemical plant -- Texas. MMWR 29:359-360, 1980. 13. Olin GR, Ahlbom A: The cancer mortality among Swedish chemists graduated during three decades. Environ Res 22:154-161, 1980. 14. Pell S, O'berg, MT, Karrh BW: Cancer epidemiologic surveil lance in the DuPont Company. JOM 20:725-740, 1978. 15. Wen CP, Tsai SP, Gibson RL: A report on brain tumors from a retrospective cohort study of refinery workers. Proceedings of the New York Academy of Sciences, New York City, 381:130-138, 1982. 16. Rushton L, Alderson MR: An epidemiological survey of eight oil refineries in Britain. Br j ind Med 38:225-234, 1981. 17. Hoar SK, Pell S: A retrospective cohort study of mortality and cancer incidence among chemists. JOM 23:485-494, 1981, 18. Reeve GR, Lloyd )W, Alexander V, et al: A progress report: The investigation of brain tumors at the Texas Division of Dow Chemical. Proceedings of the New York Academy of Sciences, New York City, 381:62-72,1982. 19. Buffler PA, Wood S, Eifler C, et al: Mortality experience of workers in a vinyl chloride monomer production plant. JOM 21: 195-203, 1979. 20. Hoar SK, Morrison AS, Cole P, et al: An occupation and exposure linkage system for the study of occupational carcino genesis. JOM 22:719-748, 1979. 21. Miettinen OS: individual matching with multiple controls in the case of all or none responses. Biometrics 25:339-354, 1969. 22. Miettinen OS: Estimation of relative risk from individually matched series. Biometrics 26:75-86, 1970. 23. Miettinen OS: Estimability and estimation in case-referent studies. Am j Epidemiol 103:226-235, 1976. 24. Mantel N, Haensael W: Statistical aspects of the analysis of data from retrospective studies of disease. JNCi 22:719-748 1959. 25. Miettinen OS: Principles of epidemiologic research. Depart ment of Epidemiology and Biostatistics, School of Public Health, Harvard University (unpublished course text), 1 975. 26. Rothman K, Boice J: Epidemiologic analysis with a pro grammable calculator, Washington, D.C.: National Institutes of Health Publication No. 79-1649, 1 979. 27. Fisher RA: Statistical Methods and Scientific Inference. London: Oliver A Boyd, 1956. 28. Decoufle P: Further analysis of cancer mortality patterns among workers exposed to cutting oil mists. JNCi 61:1025-1030, 1978. 29. |arvholm 8, Lillienberg L, Sallsten G, et al: Cancer mor bidity among men exposed to oil mist in the metal industry. JOM 23:333-337, 1981. 30. Mjlham 5: Cancer mortality patterns associated with expo sure to meuls. Ann NY Acad Sci 271:243-249, 1976. 31. Ott MG, Holder BB, Langner RR: Determinants of mor tality in an industrial population. JOM 18:171-177, 1976. 32. Greenwald P, Friedlander BR, Lawrence CE, et al: Diagnos tic sensitivity bias: An epidemiologic explanation for an apparent brain tumor excess. JOM 23:690-694, 1981. R&S 041174 386 Brain Tumor in a Chemical Plant/Bond et al been ascribed to occupational exposures, and the influence of this diagnostic access bias has been largely ignored. Recent evidence indicates that industrial populations with such benefit programs may receive a more sophisticated diagnostic workup when brain tumor is suspected than does the general population.32 One would expect this diagnostic access bias would have the effect of overstating the risk of brain tumor mortality among industrial cohorts relative to the general population. Case-control studies are particularly prone to a number of potential biases, which can inadvertently lead to incor rect results. The biases of most concern are selection bias, information bias and confounding bias.25 Selection bias refers to a distortion of the estimate of effect (odds ratio) resulting from the manner in which cases and controls are selected into the study. A selection bias may have occurred in this study, although considerable effort was expended to minimize its effect. Since the cases studied are, in all probability, a sample of the brain tumor deaths among former employees (due to incomplete case ascertainment), it was important that the ways in which the sample was unrepresentative of the larger group be known, so that controls could be selected in the same manner. The cases studied were strongly suspected to have been employed for a longer duration with the com pany than were unascertained cases. To compensate for this, one of the control groups, group B, was matched to the cases on length of employment. Control group A, like the cases, was somewhat biased toward employees with longer service anyway, since it was selected from among deaths known to the company. Information bias refers to the distortion of the estimate of effect resulting from misclassification of subjects on either the disease or exposure variables. If the case defini tion was somewhat loose and individuals who did not truly have primary intracranial neoplasms were included as cases, then differences in exposure history between cases and controls could be diluted and obscured. The case definition employed in this study attempted to incorporate the best available medical information, although, for some cases this consisted of only the death certificate. The employees' company work history record served as the primary source of exposure information in this study. While it is believed to be the best available documented source of information for such a purpose, there may be inadequacies in the record. Both job classification title and department name and the relevant calendar dates are de tailed in the record. In most instances, these were quite descriptive and subjects could be grouped by similar expo sure history quite easily for analytical purposes. In some instances, job title and department were rather nonspecific and the industrial hygienist had to infer exposure history from interviews with employees with historical knowledge, or from other records available. Industrial hygiene survey data were not available for the early years of operations and since there was no chemical or physical agent suspected as an etiologic agent a priori, no attempt was made to rank exposure levels. Some misclassification of exposure was, thus, possible; however, every effort was made to classify individuals correctly and since the industrial hygienist was unaware of the case or control status of the subjects, it is unlikely a systematic misclassification occurred. Journal of Occupational Medicine/Vol. 25, No. 5/May 1983 Probably of more concern, the relatively small number of cases available for study and the relatively low back ground worker population exposure rate to each of a wide variety of agents combine to limit the statistical power of this case-control study. For example, assuming an a-level of 0.05, the study would have had a 90% chance of de tecting, as significant, a 3ft-fold risk for having been em ployed in a magnesium department. The study would have had less power to detect associations of similar strength with other exposure variables to which the background worker population exposure rate was less. Finally, bias due to confounding variables may also distort the estimate of effect. Confounding variables in this study would be those factors associated with the risk for brain tumor and independently associated with any of the exposure variables. Matching in the design of the study and stratification in the analysis were employed to control all factors thought to have been potential confounders. There is always the possibility that some variables not measured could have confounded the results. In conclusion, this plant-based case-control study of brain tumor mortality failed to identify any significant associations with workplace exposures, although an ele vated risk was suggested for employment in the Machine Shop prior to 1945 based on three exposed cases. Acknowledgment Or, William Fishbeck, Bernard Lasich, Patricia Schultz, Randy Mikel and others within Dow Chemical U.S.A. helped with this study. Gordon R. Reeve (NIOSH) and Dr. |. William Lloyd Occu pational Safety and Health Administration helped in case-ascer tainment and provided helpful comments. References 1. Reeve GR, Bond GG, Lloyd JW, et al: An investigation of brain tumors among chemical plant employees using a sample-based cohort method. }OM 25:387-393, 1983. 2. Bond GG, Cook RR, Shellenberger R), et al: Case-control study of brain tumor deaths among employees at a chemical manu facturing plant, in Proceedings of the New York Academy of Sciences, New York City, 381:73-82, 1982. 3. Gold E8: Epidemiology of brain tumors, in Lilienfeld AY (Ed.): Reviews in Cancer Epidemiology. New York: New York Elsevier North-Holland, Inc., 1 980, pp 245-292. 4. Choi NW, Schuman LM, Gullen WH: Epidemiology o primary central nervous system neoplasms: I. Mortality from pri mary central nervous system neoplasms in Minnesota. Am ) Epi demiol 91:238-259, 1970. 5. Culler S|, Young JL: Third National Cancer Survey, Inc dence Data. Washington, D.C.: DHEW Publication (NIH) 75-78/, 1975. 6. Swenberg JA: Chemical-and-virus-induced brain tumors. Natl Cancer Inn Monogr 46:3-10, 1977. 7. Quasi JF, Wade CE, Humiston CG, et al: A Two-Year Toxicity and Oncogenicity Study With Acrylonitrile Incorporated in the Drinking Water of Rats. Midland, Mich.: Toxicology Research Laboratory, Health and Environmental Sciences, Dow Chemical U.S.A., I 980 (for the Chemical Manufacturing Association). 8. Maltoni C, Lefemine G, Ciliberti A: Carcinogenicity bio assays of vinyl chloride monomer: A model of risk assessment on an experimental basis. Environ Health Peripect 41:3-29, 1981. 9. Alexander V, Leffingwell SS, Lloyd JW, et al: Brain cancer in petrochemical workers: A case series report. Am j In Med 1:115123,1980. 10. Thomas TL, Decoufie P, Moure-Eraso R: Mortality among workers employed in petroleum refining and petrochemical plants, JOM 22:97-103, 1980. 385 R&S 041173