Document ZBMX5K1ZjwK800nMeLwDm2BLZ

AMERICAN JOURNAL OF INDUSTRIAL MEDICINE 3 1 :2 1-27 (1 997) A Case-Control Study of Hematopoietic and Lymphoid Neoplasms: The Role of Work in the Chemical Industry Barbara L. Massoudi, PhD, MPH,' Evelyn 0. Talbott, DrPH,'* Richard D. Day, PhD, MSHyg,2 Steven H. Swerdlow, M D , ~Gary M. Marsh, PhD,' and Lewis H. Kuller, MD, DrPH' The present case-control study was conducted in an effort to determine if work in the chemical indust? is related to excesses of certain hematopoietic and lymphoid neoplasms. Cases who died from non-Hodgkin 's lymphoma, multiple myeloma, and leukemia were matched by race, gender, age, year of death, and county of residence to controls who died from cardiovascular disease. A total of 618 (309 matched pairs) white male residents of Kanawha Counc, W ,aged 23-96, who had died between 1965 and I990 were identijied. Conditional logistic regression was conducted and Fielded an associution between chemical industy work and death due to non-Hodgkin 's lymphoma, multiple myeloma, and lymphoid leukemia among subjects who died at age 4 5 . These results are consistent with thefindings of previous studies linking work in chemical manilfacturing to hematopoietic and lymphoid neoplasms, and indicate that the excesses may be related to the occupational exposures in men who died ut vounger ages. Am. J. Ind. Med. 31.21-27 0 I997 Wdey-Liss, [ne. KEY WORDS: chemical industry; leukemia; malignant lymphoma; mortality; multiple myeloma, occupational diseases INTRODUCTION Concerns about cancer risk to residents living near hemica1 manufacturing facilities in Kanawha County, .i\;\t Virginia spawned an epidemiologic study of trends in -er mortality in that area [Talbott et al., 1992;Day et a]., 1: Kanawha County death rates were compared to WV death rates, and to a control county with little chemical rninufacturing. There were no significant differences by mmt! for cancer death rates except for leukemia/aleukemia :%ament of Epidemiology, Graduate School of Public Health, University of = w m e n t of Biostatistics, Graduate School of Public Health, University of 2E:anment of Pathology, Divlslon of Hematopathology, Unlverslty of PI*School of Medicine. h n s o r e d by the EnvironmentalProtectAgency, grant number CR811173; the 'merCan Cancer Society, grant number 58-33. .Wspondence to: Or Evelyn Talbott, Department of Epidemiology, 130 m St Pittsburgh, PA 15261. X l e d for publication 24 April 1996. (OR = 1.27, 95% CI = 1.03-1.60) and lymphoreticulosarcoma (OR = 1.66, 95% CI = 1.24-2.07) among white males. The age-adjusted cancer death rates by period were higher for Kanawha County residents for all cancers and lung cancer when compared to West Virginia as a whole. The authors noted four limitations of the study. It was ecologic in nature and lacked diagnostic confirmation on the death certificate, as well as specificity. There was also no occupational or environmental information available to link to disease endpoints except for county of last residence. The present study was undertaken to examine the relationship between occupation in the chemical manufacturing industry and certain hematopoietic and lymphoid neoplasms in Kanawha County,w v . This study addressed two of the PreViOUS ecOlOgiC study limitations by utilizing individual rather than aggregate data for selected diseases, and by using occupational data as reported on the death certificate. In addition. specific diseases were examined as Opposed to broad classes. The population-based methodology allowed for an increased sample size and an evaluation of individuals who may have worked at any number of chemical manufacturing facilities in Kanawha County. To determine 3- Jiley-Lis, Inc. -"- 22 Massoudi et at. TABLE 1. Disease Analysis Groups Used in the Kanawha County Case-Control Study, 1965-1990. by ICD-9' list an( Disease analysis group No. ICD-9 cause of death an( we Non-Hodgkin's lymphoma 140 200.0, 200.1, 200.8, 202.0, 202.3. 202.8, 202.9 an( I Multiple myeloma 83 203.0 Acute leukemia 69 204.0, 205.0, 205.2, 206.0, 207.0. 208.0 tha Acute nonlymphocytic leukemia 45 205.0, 205.2, 206.0, 207.0 us1 Chronic lymphoid leukemia 61 202.4, 204.1 dol Chronic myeloid leukemia 24 205.1 wa Lymphoid leukemia 85 202.4, 204.0, 204.1, 204.8, 204.9 co1 Nonlymphoid leukemia 81 205.0, 205.1, 205.2, 205.9, 206.0, 206.1, 206.9, 207.0 93' Non-Hodgkin's lymphoma, multiple 309 200.0, 200.1, 200.8, 202.0, 202.3, 202.4, 202.8, 202.9, 203.0, his myeloma, and lymphoid leukemia' 203.8, 204.0, 204.1, 204.8, 204.9 SUI 'ICD-9, International Classification of Diseases 9th Revision. 'One case of an other immunoproliferative neoplasm, ICO-9 203.8, has been included in this overall group, but was not included in the multiple myeloma group above. if work in the chemical industry was related to hematopoietic and lymphoid neoplasms (ICD-9 200,202-208), deaths among white males due to the selected neoplasms were identified. Controls were chosen from a pool of eligible deaths due to cardiovascular disease (ICD-9 410-429) and were matched to the neoplasm deaths on age, race, sex, and year of death. MATERIALS AND METHODS The case group consisted of all deceased white male residents of Kanawha County whose underlying cause of death was listed as hematopoietic or lymphatic neoplasms (ICD-9 200,202-208), and who died in-state between 1965 and 1990. Death certificates for the cases were obtained from the State of West Virginia Department of Health and Human Resources Vital Registration Office. A total of 469 death certificates were identified as potential cases. The controls consisted of deaths due to cardiovascular disease (ICD-9 410-429) that most closely matched the case's year of death, county of residence, and age within 5 years. Cases and controis were required to have an occupation or industry that could be identified from the death certificate. In the event that a case had missing data on both the kind of industry and usual occupation fields, it was dropped. In the event that a control was missing these data, a new control was drawn. In the present study, deceased cardiovascular disease controls were selected because they were a practical choice as opposed to the elusive representative cross section of the population. The choice of this control group should not have introduced serious bias. Wang and Miettinen [19841 discussed the use of cardiovascular controls in occupational cancer studies, and such controls were used in a study of cancer among lens manufacturers [Wang et al., 19831. In a cohort mortality analysis of Kanawha County Union Carbide Corporation chemical workers [Rin- sky et al., 19881,the standardized mortality ratio (SMR) for arteriosclerotic heart disease was not statistically differe I from that of the general population (SMR = 96. 95% Cl = 93-101). Therefore, no association between nonexposure and death from cardiovascular disease would be expected For many of the younger cases, controls who died of cardiovascular diseases were not available, so accident deaths (ICD-9 E800-E999) were selected as a secondary control group. There is no evidence that the exposures of interesr are associated with accidental deaths. All death certificat:, were coded by a qualified nosologist according to t,i: ICD-9. The individual four-digit ICD-9 codes were divided into nine groups by the project pathologist (S.H.S.). These groups are based on disease, cell of origin, and. for leukemias, whether they were acute or chronic. The leukemia. were grouped based on acute vs. chronic, cell type, and. for the categories without enough cases, a combination of b0.h The neoplasms of lymphoid origin (NHL, MM. and 1 ~ 1 1 phocytic leukemias) were combined and analyzed as one group since the different cell types involved are more closely related to one another than they are to the cells in neoplasms of myeloid origin. Some oncogenic events are more associated with lymphoid neoplasms and others n ith myeloid neoplasms. Because of the varied criteria used. t h t nine groups are not mutually exclusive. For example, a c: \c of acute myeloid leukemia would be included in thr-: groups: acute leukemia, acute nonlymphocytic leukern:-c and nonlymphocytic leukemia. The diseases, sample size. and codes included in each group are shown in Table I. As the key question to be addressed by the stud) 1' occupational exposure of the cases relative to their age- matched controls, a sample of cases and controls were <-: lected from the entire study period (n = 100). Next of I n PO' si5 wc mr! sis mi 'Y I a' r( Irii eit thc oc (S Te tis va sic in( eg Pn mi Wl ge us ys1 le! 65 R fn W( an lai to be Si( Neoplasms and Work in the Chemical Industry 2 3 .` .`he death certificate were contacted by telephone ,s.tionnaire was administered to validate residence ~ -1 .. . of the decedent. While several attempts , . ..-Ji`to contact the entire sample of 100, due to time .._.-.-?[aqconstraints, only 50 subjects were reached. ~. -,`.+tigator-rated classification system was devised _-.<da \core to each case and control based on their LLlpatmnand kind of industry. The scoring was ,<)utknowledge of the status of case or control, and .._. ;:~,,med independently by two observers, and later ..,.-Gr2d.Csing this preliminary rating, an agreement of I ;,.:. rexhed. For the remaining subjects, the work ..,. . .\ah reviewed and a consensus was reached. The . L t r t placed into one of three groups; not exposed. .. txposed, and exposed. The exposed group con- .. . ..:jects whose occupation or industry data specified . ., .: :he chemical industry, or at a known chemical ..:;..;..;wing facility. The potentially exposed group con- . .:.s 1 1 :hose whose data did not specify work in a chemical -_I:aiiituring facility, but involved exposures to industrial- .r: -,?.stnicalssuch as solvents. clinical laboratory re- .:-'.:'. ::<. in the course of their daily work. The unexposed _ e . : i h i s t s of workers not routinely exposed to indus- :;hemicals during their work. An individual with -' .. .rated or known chemical manufacturer listed as --.:: : : a s t r y . or a chemical worker job title listed in their - - -piion. was classified as exposed. The Statistical Package for the Social Sciences-PC 'i'q 5- PC)and the Epidemiologic Graphics, Estimation and .,.- -:Puckage (EGRET) were used for data analysis. Sta- .:nLiI! sis included chi-square tests for categorical ' \ For the analysis using conditional logistic regres- id\ ratios were determined using two variables, the .-!:dent (exposure)and the dependent (the disease cat- `:~'n+ qxcitied in Table 111). Review of the results of the " l ~ \study [Day et al., 19901indicated that the concern I t k L V i t h men who died at younger ages. To explore ,?i"";'r the increase existed only in younger men as sug- -. -. .. , '? the previous work. an analysis was performed -'Jse dichotomization. Therefore, three sets of anal- .' '- -.., . .Y? ..im Ferformed: one on all 65 at the time of their subjects; a death; and second a third on on those those .'\ A i r %)l&r. RESULTS -. " 2 ari_cinalsample of cases consisted of 469 deaths .- '2:natopoietic or lymphoid neoplasms. Of these, 2 1 ' . . : a ~ iWs est Virginia residents who died out of state. -?,: .,<rt[herefore excluded. This reduced the study popu4i:orl to U 5 cases. After the death certificates were coded - ` *he ninth revision, 10 additional cases were excluded -f-,ill\e the underlying cause of death was no longer con``<?-:J to be a hematopoietic and lymphoid neoplasm (ICD-9 200, 202-208), leaving a total of 435 cases. The results presented in this paper are restricted to the 309 matched pairs where the case died of non-Hodgkin's lymphoma, myeloma, or lymphoid leukemia. The majority of controls (94%) died of heart disease (ICD-9 410-429), but for 6% of the younger deaths it was necessary to select from among the accident deaths. After coding underlying cause of death in ICD-9, the cause of death for seven of the original controls was not classified as heart disease in ICD-9 (410-429). These controls were not replaced because it was felt that no material difference in calculated risk would result. In no instance did a control have a hematopoietic or lymphoid neoplasm listed as a contributing or other cause of death. Cases and controls were compared on the basis of demographic characteristics, and found to be similar. Slightly more cases were married at the time of their death. Controls were more likely to have died somewhere other than the hospital, reflecting the nature of their disease, and were therefore slightly less likely to have had an autopsy performed. The median age at time of death for both cases and controls (matched on age) was 69, with a range of 23-96 for cases and 23-92 for controls. There were differences in age by hematopoietic and lymphoid disease group, however, with a median age of NHL cases of 66; MM cases of 71;and LL cases of 72. The industry category as defined by the Census Bureau system [Bureau of the Census, 19801 is shown for the study population in Table 11. The proportion of cases and controls within each major group appears similar. The largest single proportion of both cases and controls is employed in the manufacturing of nondurable goods, of which chemical manufacturing accounts for the majority. The hypothesis testing was conducted using conditional (matched) logistic regression. Table I11 shows the results for non-Hodgkin's lymphoma, multiple myeloma, and lymphoid leukemia in the total group and using the age-stratified groupings. Analysis of the potentially exposed group combined with the chemical workers or combined with the not exposed group made no material difference in the results. The analyses presented here compare the chemical workers to the baseline, or not exposed, group. Therefore. pairs where either member was potentially exposed were excluded from the analysis. A statistically significantly elevated odds ratio was found for chemical workers who died of nomHodgkin's lymphoma and who were less than 65 years of age at death (OR = 3. I 1, p = 0.003), while no association was noted in the unstratified data (OR = 1.52, p = 0.168). Among those who died of multiple myeloma, a statistically significant effect is seen with chemical industry work for both age groups combined (OR = 2.39, p = 0.039). No statistically significant associations were seen in those who died of lymphoid leukemia; however, those who died under age 65 had 24 Massoudi et al. TABLE 11. Industry Classification System for Kanawha County Case-Control Study, 1965-1990: Major Groups by Case-Control Status Industry category" Agriculture, forestry, and fisheries Mining Construction Manufacturing-nondurable (Chemical manufacturing) Manufacturing-durable Transportation, communication, and other public utilities Wholesale--durable Wholesale-nondurable Retail trade Finance, insurance, and real estate Business and repair services Personal services Entertainment and recreation services Professional and related services Public administration Armed forces Industry not reported Cases (no. = 309) No. % 4 1.3 28 9.1 24 7.8 90 29.1 (811 (26.2) 26 8.4 28 9.1 2 0.6 7 2.3 24 7.8 15 4.9 4 1.3 4 1.3 1 0.3 24 7.8 10 3.2 0- 30 9.7 Controls (no. = 309) No. YO 6 1.9 39 12.6 27 8.7 70 22.7 (58) (18.8) 20 6.3 39 12.6 5 1.6 9 2.9 30 9.7 6 1.9 5 1.6 3 1.o 2 0.6 13 4.2 13 4.2 1 0.3 21 6.8 'Based on Classified Index of Industries and Occupations. US. Bureau of the Census. odds ninefold higher for chemical exposure (OR = 9.13, p = 0.09). The most statistically significant association is seen in the combined non-Hodgkin's lymphoma, multiple myeloma, arid lymphoid leukemia group, where those who died under age 65 are over three times as likely to have worked in the chemical industry (OR = 3.31, p = 0.001). A total of 50 interviews with next of kin were completed in order to evaluate the validity of death certificate industry and occupation data. Agreement was defined as both sources indicating the exact same rating of the three exposure classes for a subject. The result was a 95.9% agreement between the two sources, with a kappa statistic of 0.92. There was very little movement from one exposure group to another, and in no case did an individual change from an exposed to a nonexposed group. DISCUSSION Chemical manufacturing in the Kanawha Valley began prior to World War 11, around the Kanawha river which served as both a water supply and a source of transportation [Talbott et al., 19921. The initial processes made use of abundant natural gas supplies in the area. During the war, and shortly thereafter, the industry began a rapid growth phase, as natural rubber sources were cut off [Rinsky et al., 19881,and the need for varied chemical products expanded. In a 1986 report, the National Institute for Chemical Studies reported that the Kanawha Valley chemical complex rankd fifth in the U.S. in terms of its size and overall complexi-!. Over the past 20 years, the plants have become more sp:cialized, and some have limited their production to a feu specific end products. Several studies of health effects among chemical worhers have included Kanawha County chemical workers in their populations. The mortality experience of a cohort of 29,139 male chemical workers employed at any of t h i x Union Carbide Corporation facilities in the Kanawha Val r\ was assessed by Rinsky [1988]. With 19.9% of the cohnn deceased, a statistically significant excess of deaths u noted for lympho- and reticulosarcoma. Elevations of the lymphatic tissue cancers were noted across most emplo! - ment duration categories and latency strata, but no duration-response trends were seen. A nested case-control study was conducted on 52 CB+ of non-Hodgkin's lymphoma (NHL) (International Cla ,Ification of Diseases 9th Edition [DHEW, 19771 ICD-9 2 O. 202), 20 cases of multiple myeloma (MM) (ICD-9 203 1. -3' cases of nonlymphoid leukemia (NLL) (ICD-9 205. 2061. and 18 cases of lymphoid leukemia (LL) (ICD-9 204) from the cohort study [Ott et al., 19891. Controls consisted of workers randomly selected from the total employee porulation. Elevated odds ratios were seen for NHL in the ,'n- 1 1 1 I I 1 1 Neoplasms and Work in the Chemical Industry 25 TABLE 111. Summary of Regression Results for Kanawha County Case-Control Study, 1965-1990: Chemical Workers vs. Not Exposed Disease group Matched pairs Odds ratio p value 95% CI` `jon-Hodgkin's lymphoma c65 265 vjeloma <65 265 ,\cute leukemia <65 265 Acute nonlymphoid leukemia c65 265 Chronic lymphoid leukemia (65 265 Chronic myeloid leukemia <65 265 Lvmphoid leukemia <65 265 Nonlymphoid leukemia (65 265 Non-Hodgkin's lymphoma, myeloma, & lymphoid leukemia c65 265 106 50 56 63 27 36 51 28 23 36 22 14 45 9 36 19 a 11 60 14 46 64 35 29 230 91 139 1.52 0.168 0.84-2.76 3.1 1 0.033 1.10-a.a2 0.97 0.938 0.45-2.08 2.39 0.039 1.04-5.48 2.50 0.121 0.078-7.97 2.14 0.21 1 0.65-7.01 0.69 0.412 0.28-1.69 0.69 0.500 0.23-2.03 0.62 0.590 0.1 1-3.52 1.09 0.865 0.40-2.96 1.oo 1.oo 0.32-3.18 1.26 0.824 0.1 6-9.66 1.79 0.258 0.65-4.93 b i:ia 0.777 0.39-3.58 b b b 1.10 0.821 0.48-2.51 9.1 3 0.092 0.70-1 19.40 0.68 0.433 0.26-1.77 0.82 0.606 0.38-1.76 1.27 0.628 0.49-3.30 0.29 0.117 0.06-1.37 i5 8 0.030 i.05-2.38 3.31 0.001 1.58-6.91 1.02 0.948 0.61-1.71 `CI. confidence interval `Did not converge-model not fit. .:no1 unit. among foremen, and among instrument men in the 'xintenanceand construction work area. When latency was .\widered. the relationship grew stronger among the instru- -..:it men. No apparent association with duration of em. . m i n t was seen in the ethanol unit. Several significantly itiw associations were found for suspect chemical :-`ups. including alkyl sulfates, and metal salts of high :k)\icity. Another large-scale cohort study was conducted on ``289 Union Carbide Corporation workers, which included '\ r'rliers in the Kanawha Valley, as well as elsewhere [Teta .:I.. 1990).A significant excess of deaths due to lymphoi rericulosarcoma was noted among hourly employees. -urnination of specific locations led to the discovery of -.iu>terof cases at the South Charleston Chemicals and ac\tics location in Kanawha County. No other locations -??eared to have an excess of these cancers. The most striking finding of this study is the association -:[wen work in the chemical industry and non-Hodgkin's lymphoma in the men who died at less than 65 years of age. An association was also seen for myeloma and work in the chemical industry, although statistical significance was achieved for myeloma only with the increased sample size of the overall group. Lymphoid leukemia, largely composed of cases of chronic lymphoid leukemia, was strongly associated with chemical industry work in men who died under age 65, although not statistically significant. A grouping of the three diseases, non-Hodgkin's lymphoma, myeloma. and lymphoid leukemia, showed the strongest association, with a tripling of the risk in the younger age group. In the overall group, the association was statistically significant. although no effect was observed in the older age group alone. A reanalysis of the data using the true median age at death as a cutpoint revealed no appreciable differences in the results. There are several possible reasons why the increased risk of these diseases was noted only in men who died under the age of 65. The first is that chemical workers of all ages j" 26 Massoudi et al. I have an increased risk of these diseases, but the increased Hodgkin's lymphomas could be differentiated as to those of RI II risk could only be demonstrated in the younger age group. low vs. intermediate and high grade. The incidence and prevalence of these diseases increase With regard to selection bias, it is theoretically possihl: dramatically at older ages. and it is likely that these diseases that the WV residents who died out of state were in sonic Da! ical are multicausal. Controlling for specific risk factors in way different from those who died in-state: however. ti I, .future studies could provide much more information about number was small and was spread across all disease grow i the true risk to older men. A second explanation might Observation bias was also not likely to have played a role .n Me Go1 rc j be that the disease is occumng more in the younger age this study because any exposure ratings were done in L; SUII I group because the chemical industry exposure is either blinded fashion. One limitation of this study was the fa<[ on I I causing more disease in younger men, or shortening the that potential confounders such as previous occupationai Can latency of a disease that would have normally occurred exposures and smoking could not be controlled for. 64: at older ages. An examination of the type, grade, and dura- The results of this study were consistent with the find- tion of lymphomas would provide more information about ings of previous ecologic research in detecting excesse. of this. If lymphomas that typically occur at older ages NHL, MM, and possibly LL in Kanawha County ut R PCl call 71: were found in younger men, this would support the short- males using a population-based case-control methodolo;! ened latency theory. Thirdly, since the cases were selected It is also consistent with the studies carried out by the 1 2 based on the underlying cause of death, there could have tional Institute for Occupational Safety and Health. and the 6r Of Vi! been hematopoietic and lymphoid neoplasms that were chemical manufacturing industry. This methodology aI- not included because they were listed as a contributing lowed for complete ascertainment of deaths due to the e- cause, or as another si,~ficant condition. Lastly, occupa- lected cancers through 1990, and an increased sample 4zr' tion and industry data as recorded on the death certificate compared to the cohort studies and the case-control stud:. 31 may be more accurate for individuals dying at a younger Union Carbide Corporation. In addition. the chern :d age, as they are more likely to be employed at the time of worker population was not limited to Union Carbide C,v- their death. poration employees, but included all residents of Kanau hl: Several limitations of this study are worthy of mention. County employed in chemical manufacturing. Future stud- A shortcoming of this study was the loss of power due to ies of hematopoietic and lymphoid neoplasms should in- resmcting the study population to those with appropriate clude an examination of age at death or preferably age a\ underlying causes of death. Also, the measure of exposure is exposure, as the present study showed an effect onl! ir, indirect and uses secondary data sources. A validity study those who died under age 65. Had the data not been stm- was conducted and showed death certificate and interview ified on age at death. only a weakly positive effect u i > u l d occupation and industry data were in agreement. Other ex- have been detected in the overall group. posure characterizations that would be useful include the Further research efforts in this area should includ? an duration of exposure. the time since exposure, specific independent review of the diagnostic material and medical agents and processes involved, and- any exposures aside records if at all possible. This could provide more reliablf from work, including hobbies, and the environment. and detailed information about the type, grade. and durarioe The measures of disease were not highly specific. As of the hematopoietic and lymphoid neoplasms. which migh! with any mortality study. the type, grade, and stage of the offer clues to the etiology. In addition, it would be useful IC non-Hodgkin' s lymphoma or multiple myeloma and the cn- review employment records of the major chemical m m - teria used for classification of the acute leukemias were facturers in the region. to determine if a cluster exists ir -in! unknown. The accuracy of cause of death coding for hema- single or group of plants. topoietic and lymphoid neoplasms appears to be relatively high, with 98.1% agreement in coding of MM, 88.44 of ACKNOWLEDGMENTS NHL, and a range of 34.3-86.3% of various types of leu- kemia [Percy et al.. 19811. Overall, the hematopoietic and This work was supported by research grant CR811 l-' lymphoid neoplasms were found to have specific concor- from the Environmental Protection Agency to Dr. Tal boli dance between 60.8 and 64.8% of the time [Gobbato et al., and Institutional Research Grant 58-33 from the Arne-l;ar 19921. Most coding errors involved assigning the code for Cancer Society to Dr. Day. another cancer within the hematopoietic and lymphoid neo- We are grateful to Drs. Bruce Case and M. Jane plasms grouping (ICD-9 200, 202-208). Although the lit- for their helpful insight in the development of this rese.Irir erature indicates that death certificates appear to be rela- effort. We thank Mr. Charles Bailey and Mrs. Doris Jon:. tively accurate for these types of cancer, the revieu. of for supplying and coding death certificates, and Dr. R O W histologic and other material would reduce the misclassifi- Rinsky for his review of earlier drafts of the manuqcrP cation of disease. It is possible that an even stronger asso- This work was performed at the Graduate School of Publl. ciation would have been found if, for example, the non- Health, University of Pittsburgh. Neoplasms and Work in the Chemical Industry 2 7 Safeguards Report (1986): "A Community Hazard Assessment for the Kanawha Valley of West Virginia." National Institute for Chemical Studies. Charleston. WV. .j ;(.bct F. Barbierat0 D. Melato M. Manconi R (1992): In- 2c.l[t1 crrtit'icatediagnoses in malignancy: An analysis of 1.405 .J.L.,. Hurn Pathol 13:1036-1038. \IJ. Greenberg H ( 1989):Lymphatic and hematopoietic tissue r: ;hc.lnical manufacturing environment. Am J Ind Med 16:631- Teta MJ. Schnatter R, Ott MG, Pel1 S (1990): Mortality surveillance in a large chemical company: ne Union Carbide Corporation 1974-1983. Am Ind Med 17:435-.147, U.S. Bureau of the Census (1980): "Census of U.S. Population: Alphabetical Index of Industries and Occupations." Washington. DC: U.S. Government Printing Office. U.S. Department of Health Education and Welfare (1977): "International Classification of Diseases, Adapted for Use in the United States." Wash- (>[( (;. W:lrd E. Greenberz H. Halperin W, Leet T (1988): Smdy .:! Jinong chemical workers in the Kanawha Valley of West . \in J Ind Med 13:129-138. Wang JD, Miettinen OS (1984): The mortality odds ratio (MOR) in wcupational mortality studies-Selection of reference occupation(s) and ref- erence cause(s) of death. Ann Acad Med 13:312-316. Wang JD, Wegman DH, Smith TJ (1983): Cancer risk in the optical manufacturing industry. Br J Ind Med 40:177-181.