Document bagvNKnpd57Mezk5aYywQ0g80

Exposure Indices for Epidemiological Surveillance of Carcinogenic Agents in an Industrial Chemical Environment Richard A. Greenberg, Ph.D., and Carlo H. Tamburro, M.D. A prospective system for establishing chemical expo sure indices was developed and implemented for 22 chemicals used at a Louisville chemical plant Validation of the indices was done statistically using industry-related cancer (liver angiosarcoma) and worker-matched controls. A rank ordered system for exposures was used to identify a relationship between the occurrence of disease and the presence of a suspect chemical used in the industrial en vironment A major difficulty in the epidemiology of occupa tional carcinogenesis is obtaining accurate exposure data, especially if the data must be procured after cancer develops. While epidemiological investigations of out breaks of disease, infectious or chronic, are always retro spective, the long latent period between exposure to a causative factor and the occurrence of cancer compli cates this problem. Routine continuous recording of expo sure to possible carcinogens is an ideal goal. Unfortunate ly, this is not practical for most chemicals in a modem in dustrial setting. What is eminently practical, however, is a system utilizing rank ordering of exposures for highly suspect chemicals. The B. F. Goodrich Louisville Chemical plant devel oped such a system in 1974 in response to the discovery of cases of hepatic angiosarcoma.1*5 This system was the basis of the initial reports. It was extensively modified during a prospective medical screening program estab lished by the University of Louisville under contract NOICN-55212 with the cancer control program of the Nation al Cancer Institute.*1 The authors are unaware of other similar existing data sets. Occupational studies are usual ly based on group, rather than individual, exposures. An excellent review of the literature is given by Camble et al.* A recent discussion of a computerized system is given by Kerr.' From the University of Louisville, Dept, of Community Health (Dr. Greenberg) and Dept of Medicine (Dr Tamburro), P O Bo* 35260. Louisville. KY 40232 The system for establishing exposure indices on an on going basis was developed through employee work histo ries and rank ordered job exposure categories. All jobs are classified uniquely by both area location and work de scription by means of area-description (A-D) codes which identify employment occurring in a particular building or area, independent of job; in a particular job, independent of building or area; or by both area and description. The exposure index combines two components, i.e., work history and job exposure category, by utilizing the A-D code. The chemical exposure rating is an ordered, six-category ranking assigned to each A-D number for each calendar year, as follows: Rating 0 1 2 3 4 5 6 Level of Exposure Absent from Environment (on leave, furlough, layoff, etc,) Lowest Exposure (includes exposure up to somewhere near one hour per day) Minimal Exposure to Low Levels (chemical in building -- not handled; low vapor pressure and dust level; individual prob ably works on different floor) Moderate Exposure (works around the chemical, but exposure is minimal; individ ual is frequently exposed to little spills or leaks and infrequently -- less than once per month -- to large spills or leaks) Works in Area Subject to High Occupation al Exposures (normally exposure is mini mal, but large spills or leaks occur once per month or more) Works in Areas Where Level is High (exposure levels in area are frequently high; might consider that some risk is involved if the chemical is very toxic) Intimate Contact -- Skin or High Inhala tion (includes individuals with daily and direct contact with the chemicals, such as poly cleaner in the old days and those who handled slurry) Reprinted train Journal of Occupational Medicine May, 1M1, Volume 23, No. S pp. 3S3-3M JOM 1M1 353 R&S151635 Table 1. -- Detailed Work and Exposure History. Tear 1944 1945 1945 1946 1947 1948 1948 1949 1949 1949 1949 1949 1957 1957 1957 1957 1972 1972 1973 1974 1974 1A Work History A0 Building Job 000 000 111 111 111 111 121 121 112 112 112 112 121 121 121 121 117 000 000 000 Terminated 576 576 194 194 194 194 192 192 253 253 253 253 235 235 235 235 573 574 574 574 'Other chemical No. Month* 6 5 7 12 12 8 4 4 8 8 8 8 12 12 12 12 5 7 12 8 IB Exposure Rank lor Each Chemical Vinyl Chloride 2 2 5 6 6 5 4 4 2 2* 3* 1 1 ..................................... 1 1 .................................... 1 1 .................................... 4 1 .................................... 4 3 .................................... 4 3 ..................................... 3 1 .................................... 3 1 ..................................... 6 3 ..................................... 4 3 1 ..................................... 1 2 1 .................................... 2 1 1 .................................... 2 1 1 ..................................... 2 1 1 .................................... 22* 4 1 1 1 1 1 1 2 1 1 4 4 4 Twenty-two chemicals from two distinct manufacturing processes, one related to synthetic rubber and the other to plastics, were selected for rating. The ratings were assigned by panels of chemical exposure judges. The full six-point scale was not utilized for all chemicals. Al though it would have been ideal to have a quantitative, continuous scale of actual exposure to parts per million, this was not available for all B. F. Goodrich employees for all chemicals (nor is it likely to be available in any other industry); In order to overcome the subjectivity inherent in this exposure rating system, knowledgeable people from each area of the plant were selected to serve as that area's chemcal exposure judges. The primary criterion for their selection was experience in the given area. As a rule, production foremen and technical people (i.e., chemical engineers) were in the majority among each area's panel of judges. The area production foremen were especially invaluable as they are experienced in and knowledgeable of all jobs performed in their areas. Most valuable, how ever, were those individuals who had been at the plant since its inception, or shortly thereafter. The chemical ex posure judges met as a group and agreement was by con sensus. Table 1A gives the work history for a hypothetical in dividual. Note that each job is identified for each year in dicating both the A-D number and the number of months worked during the year. This individual, therefore, began work in 1944 and worked for six months on A-D number 000576. He worked an additional five months on this number during 1945. He completed the last seven months in 1945 working at A-D number 111194. Table IB iden tifies the exposure rating assigned to each of the 22 monitored chemicals for each A-D number for each year. For instance, the first row indicates the exposure ratings assigned in 1944 to A-D number 000576. In 1945 these re mained unchanged. The third row indicates the exposure 354 ratings for A-D number 111194 during 1945. The informa tion in Tables 1A and 1B is combined to give cumulative exposure rank months (CERM) and average exposure ranks (AER). In this example, this individual in 1944 would have had 12 CERM for vinyl chloride in 1944, having worked six months at an exposure rank of 2 during the year. During 1945, this individual would have had CERM to vinyl chloride of (5 x 2) plus (7 x 5) or 45. He would have worked a total of 12 months and would have had an AER (45 + 12) equal to 3.75. The assumption is made that the CERM can be used as a rank order statistic. The assump tion is empirically validated later in this paper. The CERM cannot be considered as representing interval data and no such use of it is intended. The results given here are based on revised work histories and A-D codes developed dur ing the NCI contract period. These same codes are now applied to the individual monitoring of all employees. The data are accurately obtained in a prospective man ner. Work histories are obtained on an ongoing basis. Each time an employee changes jobs, the change is iden tified by a payroll change card indicating the last A-D number, the new A-D number and the date of change. In addition, the rating of exposures by A-D number is now done on an annual basis using auxiliary information avail able about the A-D number including individual and area monitoring pertinent to the particular A-D number. Thus far this is available for only three chemicals and only since 1974. The rating of exposures is now obtained from chemical exposure judges consisting of representatives of both labor and management This process takes about two weeks annually. In order to determine whether such a system could work retrospectively, work histories were abstracted by A-D numbers from payroll records for all employees who worked on or after January 1, 1974, from records dating back to the opening of the plant in 1942. Payroll records R&S151636 Table 2. -- Standard Normal Deviates Comparing Exposure Ranks of Angiosarcoma Cases with the Average Exposure of Matched* Controls (by Selected Chemicals). Chemical 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 10 20 21 22 1 (23 Controls) -1.43 -2.13 -3.54 -3.31 3.97 -2.01 -2.33 0.29 -3.92 -0.76 11.27 -1.69 -1.78' -2.41 -1.87 5.13 -0.77 1.55 2.56 9.11 -3.87 6.55 2. (29 Controls) 2.32 -0.93 -2.16 -1.79 0.84 -2.58 -2.59 -0.47 -0.27 3.64 7.59 -1.35 -1.41 -2.91 -1.63 3.12 -3.43 -1.45 -4.36 2.65 -2.39 2.34 Group 3 (36 Controls) -0.66 -2.00 -3.16 -0.06 5.08 -1.59 -1.93 1.56 6.44 0.17 0.93 0.45 5.95 -2.35 0.40 5.77 5.53 5.68 4.64 3.72 -2.91 2.36 `Matched by age, sex, race, year ol employment and survival as a B. F. Goodrich employee to January 1,1974 4 (4 Controls) -2.20 -0.53 -1.50 -1.76 .1.11 -- -1.50 -- -2.28 -0.71 5.08 -1.00 0.00 -0.85 -1.21 2.55 -1.50 1.07 -0.83 2.60 -1.31 2.85 did not have the A*D numbers that were subsequently developed; these older records referred to many jobs which no longer existed, and to some which were per formed in buildings long since tom down. However, a staff of knowledgeable employees matched the jobs list ed on these payroll records with current A-D numbers. Final determination for controversial work records was made, wherever possible, by an individual employee's review, .The chemical exposure judges, who were assign ing ordered rating exposures to A-D numbers for each year, faced these same problems. They used, in addition to their memory, whatever records of chemical processes and procedures that were available. (The company main tains a file on all products ever produced and all pro cesses ever used at the plant) This system of determining work exposure data was validated empirically in the following manner The in cidence of hepatic angiosarcoma at the B. F. Goodrich plant since January 1, 1974, presented in four subjects. Each of the four subjects was matched by exact year of birth, by sex and race (white or non-white), by exact year of employment and by continuing employment at the B. F. Goodrich plant through January 1,1974. All matches are included in the subsequent analysis. Each of the angiosarcoma subjects and the corresponding matched group were compared on CERM separately for each cal endar year. The results were then ranked within each cal endar year and the ranks were summed over the calendar years. If the work histories and the ordered rating expo sures were no better than random assignments, one would expect a uniform distribution of ranks within each year and independence from' year to year. Otherwise, rank order exposures should demonstrate higher exposure to vinyl chloride in those individuals with hepatic angio sarcoma. Table 2, which gives standard normal deviates, was cal culated from the observed sum of ranks, the expected sum of ranks, and the theoretical standard error (condi tional on the observed pattern of tied ranks) of the expect ed sum of ranks. Sums were over years. The results show a very dear pattern of high exposure to chemical number 16, vinyl chloride. This is to be expected a priori if the work history data and the exposure rank data are valid. A second finding of importance is that all four angio- Table 3. -- Comparison of Frequency of Observed and Expected Excoss Exposure to Selected Chemicals. Number of Angiosarcomas Showing Excoss Exposure 0 1 2 3 4 Total Expected* Proportion 1/16 4/16 6/16 4/16 1/16 `Calculated from the binomial distribution with n=4 and p= 1/2 Observed Frequency 6 6 2 1 5 20 Expected Frequency 1.25 5.00 7.50 5.00 1.25 20.00 Contribution to Chi Square 18.05 0.20 4.03 3.20 11.25 36.73 = XJ D < 0.001 355 R&S151637 IAnsiosarcoha RA N K 0F oT T A L E xr0 U3R E Ansiosarcoha Case II (23 controls) t *1 M H II M * ** * `I ** * " * Year of Employment Rg. i -- Observed and axpacted exposure rank to vinyl chloride tor each tun year oi employment matched by age, sex, race, year of employment and survival as a B. F. Good rich employee to January 1,1974. (29 CONTROLS) R A N K 0 F 7 0 T A L E 'EXPECTED RANK (15.5) X p 0 s u R E' $ si $t s si *e t *i t *i * ** * 7* 111Angiosarcoma Case Year of Employment (36 controls) Rg 2. -- Observed and expected exposure rank to vinyl chloride tor each full year of employment matched by age, sex, race, year of employment end survival as a B. F. Good rich employee to January 1,1974. R A N K 0 F T o T A L E x ^EXPECTED RANK (19.0) p 0 s u R E Rg 3. -- Observed and expected exposure rank to vinyl chloride tor oach hill year of employment matched by age, aox, race, year of employment and survival aa a B. F. Good rich employee to January 1,1974. R&S151638 356 iiTTsesTtoiT u *> % , ei *? ca t 70 n 7? *1 Year of Emplotment Angiosarcoma Case IV R A N K 0 F T o T A L E x p 0 s 0R E (Chemical Sumer) (H controls) tEXPECTED RANK (3.0) Fig 4. -- Observed and expected exposure rank to vinyl chloride (or each full year of employment matched by age, sax, race, year of employment and survival as a 8. F. Good* rich employee to January 1, 1974. I IE *7 t r t* It 71 7| >3 71 Year of Employment Fig S. -- Observed and expected cumulative exposure rank sums for four angiosarcoma cases compared with matched sots of controls. (Matched by age, sex, race, year of employment and survival as a B. F. Goodrich Chemical Company employee to January 1,1974. The numbers of controls for the four angiosarcoma patients were 4, 23, 29 and 36. The maximum possible rank sum Is 96 and the minimum is 4. Expected rank sum is 50. A rank sum of 80 or more would occur by chance 2.6% of the time. A rank sum of 20 or less would occur 2.6% of the time.) sarcoma subjects also showed significantly high expo sures to chemicals number 20 and 22. Chemical number 20 is a group of catalysts used only with vinyl chloride and chemical number 22 is hexane, used as a solvent for the group of vinyl chloride catalysts. In addition, the data also strongly suggest a high exposure to chemical number 11, diethyl maleate, also a specialized catalyst used only for a specialized process with vinyl chloride. A clear pattern of exposure is apparent in Table 2. All four angiosarcoma subjects showed excess exposure to 5 of 20 chemicals (numbers 5,11,16,20,22). By chance, this would be expected to occur for only 1.25 chemicals. All four angiosarcoma subjects also had less than expected exposure to six additional chemicals (numbers 2, 3, 4, 7, 14,21). The expectation again is only 1.25. When the bino mial distribution is inspected (i.e., the distribution of the number of successes in four trials with the probability of success at each trial being one-half), the results are statis tically significant (xi = 36.7: p < 0.001). These are summa rized in Table 3. Once more the null hypothesis of ran dom assignment of exposures is rejected. For chemicals number 6 and 8, the exposure was tied for all members of group 4 and they were not included in the analysis. Figs 1 to 4 show the observed to the expected rank exposure for each employee with angiosarcoma by each full year of exposure. The pattern is one of consistent high exposure as compared to the matched controls. It is evident that the system does reflect the autocorrelation in jobs over time as well as the exposure to vinyl chloride among the subjects with angiosarcoma. To remove the assumption of independence from year to year (as would be required in new field studies), the CERM were summed over years for each angiosarcoma subject and the associated controls. These then provided a single ranking for each matched group. The observed in dependent ranks of the four angiosarcoma subjects were then summed and compared to the expected rank sums in 5. The exact distribution of the rank sums was obtained by direct enumeration. The angiosarcoma subjects again" show significantly high exposure to chemicals number 16 (vinyl chloride), number 20 (catalysts), number 22 (hexane used as a catalyst solvent), and number 11 (diethyl maleate). Fig 6 displays the vinyl chloride rank of the angiosarcoma subject in each matched group; this is the chemical through which empirical validation of the pro cedure is achieved. Prior knowledge of the etiology of angiosarcoma was used by the authors to validate these exposure indices. It is questionable what the situation would have been if such prior knowledge had not existed. This study would have shown that employees who subsequently developed angiosarcoma had had high exposure as compared to 357 R&S151639 40 35 30 sc g 25 o 20 % 15 * 10 5 0 I 1! Ill IV ANGIOSARCOMAS v-- Fig 6. -- Vinyl chloride exposure rank ot matched Individuals with angiosarcoma. (Matched by year ot birth, sex, race, year of employ* ment and survival as a B. F. Goodrich Chemical Company employee to January 1, 1974.) matched controls for a set of four highly correlated chem icals selected from a larger set of 22 studied. These 22 had been selected because of potential toxicity or carcin ogenicity by the following criteria: known hepatotoxin, suspected carcinogen, degree of toxicity, degree of con tact and location relative to vinyl monomer polymeriza tion areas. The chemicals would certainly be suspect but because of correlated exposures, a determination of an exact cause-effect relationship would not be possible, it is certain that further studies would be initiated to inter,, pret the observed events. The total cost incurred in setting up this system retro spectively (i.e., the cost of going back into records cover ing plant operation from the beginning up to the present), including the standardization of 321 job classifications ($450), the application of A-D codes to work histories for 1400 employees ($4,760), the assignment of exposure rat ing to each job classification for each of 35 years for each of 22 chemicals ($880), the extraction of medical history data ($6,000), and computerization ($7,920) was $20,010, or $16 to $17 per employee. The prospective cost, i.e., the cost of work that is recorded regularly forward in time (not including work and exposure history from past times) averaged $3,281 per year, or $2 to $3 per emplpyee, with a breakdown as follows: job classification (1 to 3 new jobs per year), $60; work history (240 new employees per year), $136; exposure indices (20 new chemicals per year), $160; medical data $1,425; and computer costs $1,500. Costs were determined on the basis of actual man-hours worked using industrial hourly wages of the individuals who performed the work. . These data demonstrate empirically that a system of rank ordered individual exposure indices (CERM) for highly suspect chemicals can be implemented in an in dustrial environment and can identify a known causative relationship between exposure and the development of disease. This system can provide a means of monitoring industrial environments of any size at minimum costs. It can also allow for the introduction of sophisticated monitoring methods and prospective surveillance of an environment, taking into account co-factors which in fluence biological outcome. References 1. Creech )L and Johnson MN: Angiosarcoma of liver in the manu facture of polyvinyl chloride. / Occup Med 16:150,1974. 2. Creech JL and Makk L: Liver disease among polyvinyl chloride production workers. Ann NY Acad Sd 246:88-94,1975. 3. Greenberg RA, Tamburro CH, and Kupchella CE: Prospective medical surveillance program for detection and prevention of indus trially related cancer, in Prevention and Detection of Cancer, Parts 1 and 2, H. Nieburgs (Ed ). New York: Marcel. Dekker, 1978. 4. Tamburro CH: Texas Reports, Environmental Cancer A Report to the Public, 1978. 5. Tamburro CH: Chemical hepatitis, pathogenesis, detection and management. Med Clin North Am 63:545-566,1979. 6. Tamburro CH, Greenberg RA. Newby LC, and Turns DM: Im plementation and Assessment of a Demonstration Cancer Control Detection and Prevention Program in a Cohort of Industrial Workers, Final Report NCI Contract #N01-CN-55212, November, 1978. 7. Tamburro CH: Medical surveillance for chemical hepatotoxicity, in Guidelines for Detection of Hepatotoxicity Due to Drugs and Chemicals, C. S. Davidson, C. M. Leevy, and E. C. Chamberlayne (Eds.). (DHEW-NIH Publication No. 79-313, October, 1979), pp 60-80. 8. Camble 1 and Spirtas R: lob classification and utilization of com plete work histories in occupational epidemiology. I Occup Med 18:299-404,1976. 9. Kerr PS; Recording occupational health data for future analysis. I Occup Med 20:197-203,1978. R&S151640 358