Document nmQy5J8V5VpRBVy2v6JMYb8L2

Occupational Epidemiology: Detection of Cancer in the Workplace JOHN M. PETERS, MD; WILLIAM E. WRIGHT, MD, and DAVID H. GARABRANT, MD, Los Angeles There are fundamental approaches that can be used by health professionals to study the contribution of occupational exposures to cancer occurrence. It would be helpful for clinicians who see patients with cancer to be aware of the methods used to identify those types of cancer that are related to work. Such physicians may be asked by companies or unions to collaborate on the design of epidemiologic studies in work settings or to assist in interpreting results, or they may be consulted by patients from such work envir onments. Understanding these approaches is essential in order to anticipate, detect and prevent cancer caused by occupational exposures. 'hat certain types of occupational exposure cause limitations of epidemiologic techniques and to demystify Tcancer has been known for years. The best known some of the epidemiologic jargon. examples involve rare tumors or epidemics so severe The debate about how much cancer in humans is that they are easy to recognize; for example, scrotal caused by occupational exposures usually pits the pro cancer in chimney sweeps and bladder cancer affecting totype industrialist, who claims that occupational 60 percent to 70 percent of dye workers. In recent years exposures contribute in a minor way, against the other unusual tumors have been associated with occu prototype environmentalist, who claims that a large pational exposures; for example, mesothelioma in as proportion of cancer is caused by environmental or bestos workers, oat-cell lung cancer in chloromethyl- occupational exposure. In our opinion, this debate is methyl ether workers and angiosarcoma of the liver in fueled largely by ignorance. No one knows. The studies vinyl chloride workers. The clustering of these rare have not been done; the data are not available. System tumors did not escape the attention of alert clinicians. atic and comprehensive studies, with the involvement These examples represent situations in which the risks, of industry, organized labor, government and univer are extremely high. sities, would begin the work of determining how much A more difficult problem is to relate occupational cancer is caused by occupational factors. Each discovery exposures to common tumors or to uncommon tumors of a connection between occupational exposure and when the risk is not enormous. These approaches in tumor production provides basic information that can volve quantitative techniques rather than clinical vigil be applied to reduce the incidence of cancer. This ance. For example, showing the relationship between article describes processes and approaches that can radon daughter exposure in uranium mining and an be used toward that purpose. The reader who wishes excess risk of lung cancer required the study of a to go beyond the limited scope of this paper is referred relatively large population of uranium miners and the to the book Occupational Epidemiology.1 use of epidemiologic and biostatistical techniques. This paper will focus on quantitative techniques that can Conceptual Approach be applied to the study of cancer in the workplace and Epidemiology is a word that many people misunder to observations of associations of tumors with exposures stand. In this paper, we are simply talking about trying previously thought not to be hazardous. It is written to relate an occupational exposure to a measure of to inform physicians of the uses, applications and disease, namely cancer incidence or mortality. Does Refer to: Peters JM, Wright WE, Garabrant DH: Occupational epidemiology: Detection of cancer in the workplace, In Occupational disease--New vistas for medicine. West J Med 1982 Dec; 137:555-559. From the Division of Occupational Health, Department of Family and Preventive Medicine, University of Southern California School of Medicine, Los Angeles. fe<luests t: John M. Peters, MD, Division of Occupational Health, Dept, of Family and Preventive Medicine, USC School of Medicine, 2025 Zonal Ave., Los Angeles, CA 90033. THE WESTERN JOURNAL OF MEDICINE 555 DETECTING CANCER IN WORKPLACE chemical X cause tumor Y? To answer that question we need a measure of exposure and we need a measure of cancer incidence or mortality. Optimally we would like information that would allow us to establish a dose-response relationship: How much X causes how much Y? Exposure or Dose In most studies of the relationship of a working population to cancer, the exposure or dose side of the dose-response relationship has been neglected or under emphasized. Too little effort has been expended toward categorizing relevant exposure by type, intensity and duration. The exposure that is relevant to cancer is that which occurred years before the onset of the cancer. However, in most work environments few retrospective data on exposure exist. If we walk into a rubber factory, it is difficult to ascertain what the exposure levels have been for the last 20 or 30 years. However, we can estimate the type and intensity of exposures if we know how the industrial processes have changed. Sometimes an epidemiologist can find factories that have not changed or can simulate previous work conditions so that exposure measurements can be made. By character izing present and future exposures, we can provide valuable information for medical surveillance and pros pective studies. Often the best an epidemiologist can do is to determine whether exposure took place--that is, classify workers into exposed and nonexposed groups. The duration of exposure is usually known, so this index of exposure can represent an estimate of lifetime dose. Tumor Outcome or Response A simpler part of the equation is counting the cases of cancer. Sources of data on cancer occurrence in populations include local cancer registries, death cer tificates for a group of workers or death benefit forms in company or union files. While some death certificates are inaccurate and there can be some misclassification of cancer by histologic type, these sources of cancer incidence and cancer mortality information are a solid enough basis to make the associations that are impor tant. Inherent Difficulties in Population Approach The importance of assessing exposure has already been stressed, but there are difficulties that must be considered. For example, the induction period for tumor development is usually years long. This makes the relevant exposure remote from the incidence or onset of the tumor. Most malignant lesions caused by occupa tional exposures have induction periods ranging from 12 to 50 years. Because the causative exposure can have occurred during a long period, there may have been many changes in the working environment. In addition, records are often lost, or the worker may have forgotten the materials and processes used in the work. Job mobility also complicates estimation of exposure; that is, the worker may have had several jobs in several locations during the induction period. Frequently, to protect trade secrets, companies are reluctant to release a list of ingredients in materials. In many instances, information on industrial hygiene has not been routinely collected over the last 30 or 40 years, making quantifica tion of exposures difficult. When quantitative data are not available, we often rely on qualitative information; that is, someone was exposed to asbestos at unknown levels for a known period of time. Other complications in relating exposures to disease result from confounders and competing risks. For ex ample, to ascertain the relationship between an occupa tional exposure and the incidence of lung cancer, epidemiologists must consider smoking habits. In retro spective studies, detailed information on smoking habits is rarely available although more generalized studies, such as population surveys ascertaining smoking habits by occupation and by socioeconomic status, allow esti mation of the effect of smoking. By knowing the smok ing patterns of specific occupations and by knowing the relationship between smoking levels and lung cancer incidence, adjustment for smoking factors can be made in any population. There are several other potential confounders; for example, diet as it relates to the incidence of stomach cancer or other tumors of the alimentary tract. Avocational exposures are frequently important and can confound any apparent result between occupational exposure and tumor incidence. For example, a person who has a hobby of restoring furniture may be exposed to solvents such as benzene that could result in the development of leukemia. Other possible confounders include air pollution, physical activity, stress and drugs. Another important issue is threshold versus nonthresh old relationships. Policymakers are currently debating about how to deal with carcinogens. If a certain amount of exposure is needed to increase the incidence of cancer, exposure might be permitted at levels below the point at which risk is increased. On the other hand if any exposure increases the risk, no exposure should be allowed. Rarely do data exist that allow scientists to determine whether there is a threshold or nonthreshold situation. Uses of the Quantitative or Epidemiologic Approach As noted earlier in this paper, rare tumors occurring at great excess can frequently be detected in clusters; however, common tumors cannot. By following the population in a certain occupational setting, tumor incidence and mortality can be determined. Given the right population and information on exposure, relatively small risks can be detected by the population approach and researchers can ascertain whether occupational exposures are causing common tumors. Although a large population with possible exposure to some sub stance or substances might have an excess risk, knowl edge of specific exposure within the population might allow the identification of a subpopulation with much higher risks. For example, when a large cohort of rubber workers was studied, results showed no striking excess in cancer frequency for any tumor site. However, when the population was broken down by department in 556 DECEMBER 1982 137 * 6 DETECTING CANCER IN WORKPLACE which they worked (where exposure becomes more discrete), excesses of stomach and colon cancer cases were found in those working in the rubber-compounding area, and lung cancer was excessive in workers from the tire-curing area. Well-done epidemiologic studies frequently provide information on dose-response relationships, such as the studies relating work in uranium mines and with asbestos to increases in lung cancer. Population approaches allow epidemiologists to examine latency or induction periods. Types of Population Studies There are three common approaches to studying populations that might be occupationally exposed to elements that possibly increase tumor risk. 1. A population exposed to a specific substance or substances can be followed to determine the number of tumors that develop. Such a cohort study allows an epidemiologist to relate all possible health outcomes to a specific exposure or set of exposures. 2. On the other hand, the starting point can be persons with a specific type of tumor. Here the epidemi ologist would attempt to determine what exposures have occurred in that group of persons that might explain the cancer. In this instance, the patients with cancer (cases) would be compared with similar persons without cancer (controls) and past exposures would be ascer tained. This approach is called a case-control study. 3. In some instances records may exist of the deaths of all workers from a certain company, union or other group. By using the death certificates, cause of death can be compared with the expected proportions of mortality based on age, race, sex and time of death. This is called a proportional mortality study. There are advantages and limitations to these three basic approaches that will be discussed in the next section. Cohort Studies Most studies of working populations aimed at iden tifying the relationship between occupational exposure and cancer use the cohort approach. A cohort study has the advantage of providing information on which cancer rates (new cases or deaths) can be computed. By starting with a denominator population with no cancer and following it, a researcher can determine the number of cancer cases that occur and how many deaths result from cancer within that population, thus providing a rate of cancer incidence or mortality. The occurrence of nonmalignant disease can also be examined by this approach. The major disadvantages are that cohort studies can be very time consuming and expensive be cause many persons must be traced and mortality infor mation collected. Most cohort studies are done by epidemiologists attempting to identify a cohort exposed to some substance of interest 20 or 30 or 40 years earlier, and to follow that cohort forward in time to the present. In order to do the cohort study a researcher needs a TABLE 1.--Example of Cohort Study* Observed and Expectedt Deaths Among a Cohort of 13,571 Rubber Workers% Cause of death 1CD No.% Observed Expected Deaths Deaths All causes.................. All cancer ..................... 140-205 Digestive .................. 150-159 Stomach .............. 151 Large intestine . . 153 Respiratory .............. 160-164 Genitourinary......... 177-181 Bladder ................ 181 Lymphatic and hematopoietic . . . 200-205 Leukemia ........... 204 Vascular diseases of central nervous system . . . . 330-334 Circulatory diseases . . 400-468 Respiratory diseases . . 470-527 Digestive diseases . . . . 530-587 Genitourinary diseases 590-637 Accidents, poisoning, violence..................... 800-999 Residual ....................... 5,079 986 368 98 104 255 155 48 105 55 493 2,448 218 217 75 279 363 6,090.6 1,064.9 364.1 90.4 103.4 293.4 159.6 39.5 94.4 42.7 524.0 2,871.8 327.0 279.0 135.8 440.1 448.0 SMR 83 93 101 108 101 87 97 122 111 129 94 85 67 78 55 63 81 The results are presented on a large cohort of rubber workers.2 While none of the standard mortality ratios (SMR's) for the entire group are very remarkable, further analysis by type of job showed excesses of cancer that could be explained by exposures in those jobs. Note that the starting point is a cohort of 13,571 rubber workers of whom 5,079 died over about 25 years. tExpected deaths computed on basis of age-time-cause specific mortality rates for US white men. t International Classification of Disease, 7th Revision. Standardized mortality ratio = 100 X observed number of cases/ expected number of cases. cohort defined in time. For example, all of the shipyard workers employed from 1942 through 1945 might be followed to determine the incidence of mesothelioma that could be related to asbestos exposure in the ship yards during World War II. Once the cohort is identified, the basic procedure is to follow all of those workers to determine their vital status and to collect death certifi cates or other tumor incidence information for com parison with a reference population. The reference population might be the entire United States, the state in which the cohort lived or, in some instances, the county of work or residence when county data are available for comparison. A cohort study is normally done when there is an exposure of interest and the suspected outcome is a common tumor. The more precisely exposure can be identified, the better the chance of relating exposure to disease. The example given in the previous section on the "Uses of the Quantitative or Epidemiologic Approach" illustrates this point. Table 1 shows the results of a large cohort study that classified exposure by (1) work in a company and (2) work in specific departments of the company. The mortality (standard mortality ratio) is expressed as a simple ratio of the observed number (X100) of deaths in cohort to the expected number of deaths in white men in the United States. Case-Control Studies While cohort studies are based on company (and sometimes union) records, case-control studies are THE WESTERN JOURNAL OF MEDICINE 557 DETECTING CANCER IN WORKPLACE TABLE 2.--Example of Case-Control Study* Factor Concordant Pairs: Both Exposed Mother Got chemicals on skin.......... .... 0 Inhaled chemicals or fumes . .... 1 One or both of the above ... .... 1 Father Exposed to chemical solvents .... 3 Exposed to paints ................ .... 0 Worked in aircraft industry . .... 2 P - probability of result occurring by chance. Discordant Pairs Cases Exposed Controls Exposed Relative Risk 10 3 3.3 12 4 3.0 14 5 ' 2.8 17 6 2.8 7 1 7.0 10 0 One sided Test (P) .05 .04 .03 .02 .04 .001 *A total of 92 mothers of children who were under 10 years of age and had brain tumors were interviewed along with 92 matched control mothers. The exposures of interest are the occupational exposure of both parents. The table shows that exposure of both parents results in a risk of central nervous system tumors in the offspring (no excess risk would be a relative risk of 1.0). Note that the starting point was 92 cases recorded in the Los Angeles County Tumor Registry.3 Matched-pair comparison of parental occupational exposure of cases and controls. The mother was considered to have been exposed if exposed at any time from 1 year before conception through lactation. The father was considered to have been exposed if exposed during that period or at the time of diagnosis of a case. usually not. The case-control study examines an un common tumor. If a cohort study were done following 1,000 workers in whom leukemia might develop, the ascertainment of an excess of leukemia would be ex tremely difficult because the rate of leukemia is low and the number of cases expected in a population of 1,000 is small. However, a sufficient number of leukemia cases can be collected from a tumor registry, which can sometimes supply information on where the patients with leukemia worked and what their job was. Aside from serving as a useful tool in the study of uncommon tumors, the case-control study can assess multiple ex posures (more than one cause). The principal disadvan tages relate primarily to the dependence on interview information and the fact that only one disease is being examined. In order to do case-control studies there must be a way of identifying cases from the population of interest. Tumor registries can be used for this purpose. For example, the University of Southern California School of Medicine operates a tumor registry for Los Angeles County. This registry collects all new cases of cancer occurring among the more than 7 million persons residing in the county, approximately 28,000 new cases per year. One can use the tumor registry to identify all of the cases of leukemia, for example, find a suitable control group and interview the patients with leukemia and the controls for past exposure that might account for the leukemia incidence. Case-control studies should be done when the cause of a specific tumor is of interest. The example in Table 2 provides further details of the case-control approach. Proportional Mortality Studies In many instances information is not available to establish a cohort for follow-up. One may, however, have access to the death certificates on a group of workers with known exposure. Analyzing the death cer tificates without cohort information can very often provide valuable information. The principal advantages are those of speed and cost. Computer programs allow TABLE 3.--Example of a Proportional Mortality Study* Observed and Expected Deaths Among Vinyl Chloride Workers in a Proportional Mortality Studyi Cause of Death Observed Expected SPMRt All causes ................................. 161 All cancer.............................. . . Digestive ........................... . . Liver and biliary . . . . Lung ................................... . . Brain ................................... . . Other ................................... . . 41 13 8 13 5 10 Central nervous system vascular.............................. . . 8 Circulatory ........................... . . 66 External ................................. . . 22 All other causes................ . . 24 161.0 27.9 8.3 0.7 7.9 1.2 10.5 9.5 68.6 24.5 30.5 100 150 160 1,100 160 420 95 80 96 90 79 Death certificates of 161 workers who worked in a vinyl chloride manufacturing facility were analyzed. The table shows an excess in liver, lung and central nervous system tumors.4 Note that the starting point was an accumulation of death certificates. tExpected numbers based on proportional mortality ratios for white men in the United States. SPMR = standardized proportional mortality ratio = 100 X observed/ expected. very rapid analysis of a collection of death certificates. While some epidemiologists deprecate the value of this approach, the answers provided by this type of study almost always parallel the results of the more compli cated cohort studies. An example of this form of study is shown in Table 3. Note that mortality is expressed as a simple ratio (X 100) of observed deaths to ex pected deaths. Application of These Techniques The approaches discussed have a variety of applica tions, including etiologic research and medical surveil lance. The ultimate goal of all professionals in occupa tional medicine should be to prevent disease caused by occupational exposures. This requires detection of the adverse effects, reducing or eliminating the exposure responsible for the problem and monitoring the popula tion for early identification of other problems. The approaches described in this paper can assist physicians, 558 DECEMBER 1982 137 6 DETECTING CANCER IN WORKPLACE nurses and other health professionals immensely in identifying hazardous work environments. With training and consultation with epidemiologists, these techniques can be and should be used by most companies to detect problems (if they exist) or to demonstrate a healthy working population and to assure (through continuous monitoring) that no deterioration of health related to work occurs. Expertise in occupational epidemiology exists in several large universities in the West. Most of these academically based epidemiologists are more than willing to help interested companies or unions develop and implement these programs. It is through this kind of cooperation that problems can be anticipated and prevented. If the techniques and epidemiologic methods that currently exist are applied in many work settings, we should be able to identify causes of cancer, reduce exposures to . prevent many cases of cancer from occur ring and avert future epidemics of cancer caused by occupational exposure. REFERENCES 1. Monson RR: Occupational Epidemiology. Boca Raton, Fla, CRC Press, 1980 2. Monson RR, Nakano KK: Mortality among rubber workers--I. White male union employees in Akron, Ohio. Am J Epidemiol 1976; 103: 284-296 3. Peters JM, Preston-Martin S, Yu MC: Brain tumors in children and occupational exposure of parents. Science 1981 Jul; 213:235-237 4. Monson RR, Peters JM, Johnson MN: Proportional mortality among vinyl-chloride workers. Lancet 1974 Aug 17; 2 (No. 7877):397-398 Medical Practice Questions EDITOR'S NOTE: From time to time medical practice questions from organizations with a legitimate interest in the information are referred to the Scientific Board by the Quality Care Review Commission of the California Medical Association. The opinions offered are based on training, experience and literature reviewed by specialists. These opinions are, however, informational only and should not be interpreted as directives, instructions or policy state ments. Human Tumor Stem Cell Assay QUESTION: Is a human tumor stem cell assay considered an effective technique in the treatment of cancer or is it considered investigational? OPINION: In the opinion of the Advisory Panels on Chest Diseases, General Surgery, Internal Medicine, Neurosurgery and Pathology, human tumor stem cell assays must be considered investigational at this time, although they appear to offer promise as a valuable in vitro test of tumor sensitivity to chemotherapeutic agents. At present, there are problems in methodology, in standardization and in interpretation of the data derived. These assays effectively predict drug resistance, and thereby can safeguard patients from the toxic effects of drugs known to be ineffective. However, the test is limited in its ability to predict sensitivity reliably. Prospective correlative data remain scant and limited to a few tumors and drugs. Most of these data have been accumulated for ovarian carcinoma, which grows relatively well in soft agar systems, tends to be moderately responsive to chemotherapy and often has relatively accessible tumor cells in ascites fluid. The correlations may not apply to other types of tumors and may differ from drug to drug. Further, many of the published reports are retrospec tive, or, when prospective, are not true correlations for individual drugs since patients may be treated with combination therapy after in vitro testing of single agents. There is as yet no controlled study that indicates that treatment selected by assay is superior to a clinician's choice or that patients' survival is significantly increased by use of the assays. Long-term clinical studies, carried out in controlled research settings, are required to validate the reliability and clinical application of this technique. Once the method ological differences among laboratories are resolved, and better cell growth for various tumor types and more reliable end points for measuring the effects of drugs are achieved, the clinical studies needed to assess the role of this type of assay will be possible. Until such time as its clinical utility fulfills its theoretical expectations, human tumor stem cell assays merit continued investigation. THE WESTERN JOURNAL OF MEDICINE 559