Problems and perspective in epidemiological study of occupational health hazards in the rubber industry.

An epidemiological analysis of the problems in the study of companies engaged in the manufacture of rubber products in different countries and in different time periods is given. Selected findings on cancer of gallbladder and biliary system, cancer of the lung, and tumors of the central nervous system among rubber workers are presented.

I believe it most appropriate to acknowledge at this conference, the remarkable scientific contributions and vision of Dr. Wilhelm C. Hueper who, for so many years, pioneered the study of occupational cancer and established the basis for the international concern on occupational and environmental cancer. Scientists all over the world are deeply indebted to him for his dedicated service to society.
The problems and perspectives of the study of any industry relate to the attitude, interest, and sense of social responsibility of the management of that industry or plant; and to the degree of awareness and understanding of the workers who unfortunately have not been informed of the hazards to which they were exposed. Further, the problems relate to the degree of responsibiilty exercised by the government for the protection of the industrial population and to the support provided by the government for the necessary resources for the identification and control of industrial health problems.
Past experience has taught us that, for the rubber industry as well as virtually all other industries, the absence of industrial medical surveillance systems and the absence of any environmental monitoring systems, the absence of any early warning detection system for current and delayed health effects among the workers, has created problems of recognition of health hazards, which have accumulated for decades.
Further, the absence of and inadequacy of experimental testing of chemicals prior to their Department of Industrial Environmental Health Sciences, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania 15261. use in the rubber industry, relative to chemicals tested, experimental design, lengths of exposure, and methods of exposure or even of testing chemicals under conditions of production, subject to heat, and in combination with other chemicals, all have compounded and delayed the problems of recognition of the potential health hazards among the workers.
What has not been properly recognized, too, is the generation of very large numbers of new chemicals during the work process and the interaction of these chemicals with other clhemicals in the work environment, all of which have not been identified and, therefore, never studied individually or in combination for the toxic and carcinogenic effects. The permutations of numbers of chemicals that could evolve in the work environment, in so many combinations, is tremendous, and for these no toxicological or carcinogenic data are available.
In the present paper, I shall provide an overview relative to the epidemiological considerations, utilizing selections from a series of my own studies on the rubber industry to emphasize certain points for this conference. Virtually, every cancer site, as well as other major cancers, were studied and presented in the summary study (1). My analysis of the epidemiological problems was developed after reviewing the various types of studies conducted and the problems of interpretation of the findings of the studies in different countries and in the same country.
The future perspective relative to the rubber industry in the United States is more hopeful because of the constructive approach being developed between labor and management toward the resolution of industrial health problems, which began with the tripartite agreement between the International Union, the rubber companies, and the universities. I am pleased that the tripartite agreement which I helped to design has resulted in scientific advancements for the identification and control of occupational cancers and other health hazards in the rubber industry.

Epidemiological Considerations
It must be recognized that rubber companies have utilized several thousand chemical compounds in various combinations in association with the work processes during the past decades, and the continuous introduction of new chemicals, together with the markedly different working conditions at each plant, all contribute significantly to the diversity of each plant and the problems of epidemiological evaluation.
Further, the introduction of different chemicals in different time periods in different companies engaged in the manufacture and processing of rubber products with resultant different latent periods for the biological effects severely handicaps any simultaneous correlation analysis of processes of different companies.
The combination of these and other factors, operates to submerge the detection of the occupational cancers which may exist. Consequently, when similar observations of specific cancer sites or other biological effects are observed, utilizing the same methods and time periods among several rubber companies, these observations take on added significance. Within this context, however, there may be a tendency to discount observations from a single plant, or observations from different plants in which the excess may occur in different departments. This is unwise. The individual study may accurately reflect and indicate some of the potential problems, even though some may be missed.
For each individual plant there may be substantial differences inherent in the size and physical layout which bear on the uniqueness of the physical location of different processes and the handling of materials and chemicals. Further, the age and type of equipment and process and the presence or absence and degree of controls at different points in time provide for variation in each plant. Each company may vary further in the combination of use of the same and different chemicals in each of the processes or in different processes.
There is still the complex problem in intercompany comparisons of differences in production schedules, the volume of chemicals and how they are used and handled, of selection, retention, and separation of employees, and of retirement policies and requirements of life and sickness insurance coverage, etc., for each of the companies over the decades.
For these reasons, we believe that the evaluation of each individual plant, even with its own epidemiological limitations, has real significance. In contrast, strong caution is expressed relative to the averaging effect of the mortality experience of a series of rubber plants which when combined and viewed as a total may submerge the recognition of real excess risks which were detected in some of the individual plants in specific departments and processes.
We are particularly concerned about three fundamental aspects in the study of any industry which has been in existence for 60 years or so and has used such a substantial number of complex chemicals and undergone so many successive major changes in succeeding years.
The first concern in any epidemiological study is the year, or years, in which the cohort study is established-the prospective longitudinal study done retrospectively in time. The latent period for occupational concern may be very great indeed, as demonstrated in the case of the association of asbestos and mesothelioma, which may extend to 40 or 50 years. Consequently, recent cohorts which allow only for comparatively short latent periods have built-in limitations and may provide false negative findings.
Within this context, studies of employee cohorts of the 1930's, 1940's and 1950's, or in succeeding decades, in effect reflect the manufacturing processes and chemicals of prior years. It must be reemphasized that major changes have occurred in the rubber industry, in the chemicals and in the processes used, since the industry's inception. Consequently, carcinogenic exposures, even though not recognized, may have been lessened or discontinued by chance, not because of recognition of the hazard but as a result of evolution of the manufacturing process and changes in chemicals used. Conversely, there may be unrecognized carcinogenic chemicals introduced which cannot be detected epidemiologically until the number of years has elapsed to meet the requirements of the latent period for theze chemicals. Therefore, epidemiological studies of cohort populations in different decades may not be comparable; not only for the same company, but even more so for different companies. At least in studies of the same company, investigators can determine whether the carcinogenic risk persisted over different time periods, provided the employee populations are observed for the same number of years required for occupational cancer. This principle has been effectively applied by some investigators in determining whether risk of bladder cancer had been diminished in different time periods in the rubber industry.
Cohorts of current employees designated at a fixed point in time, reflect a mixture of those hired that year and the few years immediately prior to that date, and the survivor group who may have been employed 20, 30, or 35 years before.
For cancers which have long latent periods, the cohorts identified in the 1950's and 1960's would not necessarily detect the particular cancer site or excess risk unless there was, inherent in the cohort, a sufficiently large subcohort of employees who had been hired 20, 30 years or more before in that plant, in that particular department or process where the detection of risk was observed-in essence, if the subcohort of the early hired workers was very large and happened to be in departments or processes where the carcinogenic risk existed.
False negatives or misleading results may occur if investigators utilize total plant mortality rate as an estimate of risk, even though the dilution effect of the total population is well recognized.
A second concern is in the definition of the cohort, the composition, accuracy, and completeness of the data source, and the characterization of the population specified for inclusion in the study. This has a major bearing on the evaluation of findings conducted by different investigators of the same type of industry, in this case, the companies engaged in the manufacture of rubber products. It is already apparent that cohorts initiated in two different time decades, reflecting different prior occupational exposures, would, therefore, be observed for different lengths of years and would not be comparable, nor are cohorts identified by different criteria comparable (e.g., department process designation versus total employees or retired employees or those covered solely by life insurance).
There is a tendency in the use of life insurance data to assume a completeness of coverage of all employees under all conditions and all periods of time, which may not in fact exist. The insurance criteria for inclusion of employees into the system may change over time as may the applicability of such criteria to employees which may leave the company. It would not be appropriate, as occurred in the study of the chromate industry (2), to combine data from a series of rubber plants, all of which entered into insurance coverage at different years, had different numbers of years of actual coverage, and different years of observation since insurance coverage, and therefore different specific years and length of years in which death certificates would be available. Under this combination of limiting factors, comparative evaluation of the findings of specific causes of death for each of the companies would tend to limit and obscure the full extent of the findings.
Third is the problem of utilization of appropriate controls. The general U.S. population is frequently used to derive the expected rates, even though investigators know that the industrial working population (the total plant actuarial) life experience may be 40% better than that of the general population (3) which includes the sick, the disabled, and those unable to work.
The resulting higher expected rate derived from the general population narrows the difference between the observed and the expected, and therefore limits the degree of recognition which may occur for a particular cancer site. In some instances the detection or recognition may be completely submerged. The reason given for the use of the U.S. population as a control is one of convenience. This does little, however, to correct the impression which evolves from such studies-that the observed rate was only twice that of the general population or even less than the general population, which may be misleading in terms of determining the degree of real biological effects.
We have utilized, wherever possible, internal controls-the comparison of employee groups within the same plant. In this method all the employees are subjected to the same selective influences and procedures and events at the company of preemployment and periodic medical examinations, of employment and separa-tion practices, of insurance policy requirements and policy changes, of shifts and requirements in production, of reasons for termination, of medical care in the community, and of residence in the area, all of which may at least be brought within the boundaries of comparability.
For this purpose the internal cohort comparisons are made of the employees of each department group, compared with the balance of the company, i.e., all employees in all other departments minus the observed department.
I should like to emphasize this approach, the utilization of the balance of the company to derive the expected rates. My series of tables on lung cancer will illustrate this major point. The use of this approach will detect higher risk groups that would not be noted utilizing the total population as comparison.
Internal comparisons of employee groups alone, although effective, still require a further refinement-the characterization of the employee cohorts of the respective departments by date of hire and duration of employment. This has considerable bearing on the sensitivity of detection of the excess in specific cancer sites. Duration of employment data lessens still further the dilution factor. Duration resource data were not available to us.
Another major consideration is the necessity of review of the pathology reports and microscopic slides of the various cancer sites that were observed in an epidemiological study. This will be illustrated with the findings relative to cancer of the gall blader and biliary system. It is not enough to cite the cause of death, e.g., a particular cancer; there should be verification of the specific cancer site.
The problem has been the reluctance to do this, because it could not be done for the controls when the general population was used for this purpose. However, if as was done in our studies, the follow-up on the histological diagnosis is carried out uniformly for the entire cohort, then, both the study population and the internal controls would have the same criteria of histological identification.
In the ultimate analysis, the objective, once the excess risk process or department has been established for each cancer site, is to identify the specific chemicals or combinations of chemicals, in each department or process which may be the etiological factors.
In previous studies (4,5) representing the first epidemiological considerations of the rub-ber industry, we identified a number of cancer sites for which excess mortality rates occurred, and the present study is a continuation of the study of biological effects among workers in the rubber industry. Table 1 shows the continuation of the study of cancer of the gallbladder and bile ducts among rubber workers. There were 30 deaths due to cancer of gallbladder and bile ducts for males and females.

Malignant Neoplasm of Gallbladder and Bile Ducts
The findings have emphasized the importance of repeated efforts to obtain hospital and pathology reports which have changed the diagnosis as to precise site of origin of the cancer in some instances. However, such data are not always available. As shown in the footnote of Table 1, three cases were excluded because pathology information on follow-up revealed, for example, cancer of pancreas or of liver.
For comparative purposes, the estimated frequency of the anatomical designation for this rare cancer site was derived from the total of autopsies cited by Edmondson (6, 7) utilizing the Los Angeles County Hospital autopsy series. We had calculated M5) the ratio for cancer of the ampulla of Vater to total autopsies as 1:2333; for the common bile duct as 1:4600; and for the gallbladder 1:275 (there is considerable difference of opinion among pathologists to the effect that cancer of the gallbladder may occur in 1:500 or 1:1000 autopsies); for cholangiomna, 1: 1293 autopsies. Regardless of the eventual agreed-upon ratios, there is no doubt that these cancers are very rare, and this designation is sufficient for our purpose. In viewing Table 1, it must be emphasized that the published ratio for these anatomically designated parts of the biliary system relate to the total autopsy and not to total deaths. For each of the comparisons in both the cohorts, and from the supplementary union deaths, the specific type of cancer is shown.
In the summary by anatomical site for the cases in which this could be determined in the total series, there were three cancers of ampulla of Vater, six of the common bile duct, 13 cancers of the gallbladder, and two designated cholangioma and six bile ducts.
This repeated consistency of the observation in prospective studies and in retrospective stud-Environmental Health Perspectives to I" 1-ies in different rubber companies of rare cancer sites with such frequency, together with the determination that chemicals capable of causing cancer of gallbladder and biliary system were used in the processes of the rubber industry in the early years (5), establishes an occupational relationship.
Malignant Neoplasm of Lung Company 2 Table 2 shows that within the major departments pertaining to manufacturing processes, a higher mortality rate occurred for the curing department for each of the age groups when compared with the balance of the company. For age group 35+, the mortality rate was 167 vs. 94.
Two other departments (nonprocess) ( Table   2) also showed a marked excess in mortality rates over the balance of the company. For the machine maintenance and tire shop, the excess rate occurred in each age group and was highest in ages 35+ (176 vs. 93). For the craftsmen the rates were highest in ages 65+ (603 vs. 284) for the balance of the company. Table 3 shows the mortality rate by combination of risk departments. When the curing department is compared against the balance of the company in which the high risk nonprocess departments are removed (machine maintenance and tire shop, machine shop, and craftsmen) then the excess for the curing department becomes even more marked. Company 3 Table 4 shows that an excess in mortality rates occurs for the tire and tube building departments and for the curing department. In the curing department, the rate for ages 25-64 was over four times the balance of the company (149 vs. 34). (For this age group the curing department in company 2 had a mortality rate twice the balance of the company). Table 5 shows an excess occurs in the tire and tube building departments for ages 65+ (305 vs. 159) and a consistent excess occurs in the curing department for all age groups. Table 6 shows the rates by subcohorts. For the tire and tube building departments, two of     In a demographic study of cancer and other tumors of the brain and central nervous system in which all deaths (1944)(1945)(1946)(1947)(1948)(1949)(1950)(1951)(1952) for the state of Ohio were classified by degree of confirmation, it was determined that white male resi- b Data limited to industries with 10 or more employees in the first quarter, 1939 in the group of Ohio male residents dying of tumors of the central nervous system, 1944-1952.

Company 4
Rates were computed on the basis of the 1940 Census population and age adjusted on the basis of the total population of continental United States in 1950. Separate population data were not available by duration of employment for each industry. d Approximately 15% of the total was employed in miscellaneous or unreported industries and no data were computed on the length of employment of this group. On the other hand, only about 1%g of the population used as a basis for computing rates was in the miscellaneous or unreported industry group. dents of Summit County (the center of the major rubber companies) had a significantly higher rate than expected. A similar comparison of Summit County rubber workers confined to histologically confirmed cancers of the central nervous system showed a significantly higher rate than expected. Table 7 shows the age-adjusted mortality rate for 934 deaths due to cancer and tumors of the central nervous system which occurred during the period 1944-1952 according to industry classification and duration of employment. Data on prior employment in industry were obtained from multiple sources.
In the comparison for those employed 4 yr or more in the designated industry in 1939, the highest death rate occurred in the electrical machinery category (64.59 per 100,000), then rubber (58.69) and paper and allied products (55.05). When the duration is 8 yr or more in that industry, then the rate for the rubber industry is highest (58.69), then electrical machinery (47.28), and paper and allied products (46.79). Table 7 also illustrates that death rates by industry for a single point in time may change extensively when further refined by duration of employment.
For company 3 (Table 8), three (50%) of the six cancers of the brain (193) for the entire company occurred in the tire and tube building departments with a resultant excess over the balance of the company (25 vs. 4 for ages 25-64 and 66 vs. 20 for age 35+). The two deaths which occurred in the miscellaneous departments were classified as control laboratory and pipefitter.
In the intercompany comparisons for categories 19, 223, and 237 of the International Classification of Diseases (8), there was an excess in the tire and tube building departments for companies 2, 3, and 4 and for the curing department in companies 2 and 4. Stock preparation and machine maintenance and tire shop in company 2 showed marked excess over the balance.