Document 5bpRxv48Ekdax7e56Eaz2dGaJ

I Reprinted by the U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Healch Service From PUBLIC HEALTH REPORTS Q Vol. 79, No. 11, November 1964 o o The Estimation of Expected Rates in Occupational Disease Epidemiology PHILIP E. ENTERLINE, Ph.D. .> t\ PROBABLY no environment impinging on vation factor is an important area for future J' the health of man has so much effect and work; in occupational health and that the great is so malleable as the environment in which hemajority of health problems of industrial work V works. No doubt, because of this, there has ers probably are not caused directly by condi been great interest in relating working condi tions of work but may be increased in severity K(r- tions to health, an interest that predates more and progression by insult3 at the work place general epidemiologic studied of disease and U)- which has occupied many outstanding epidem The first step in the identification of many i. iologists and medical statisticians. of the harmful occupational exposures which Bor some diseases evidence of an unfavorable remain or which are likely to occur in the future effect of the working environment has been is to identify the separate effects of the occupa strong, and control measures have been suffi tional environment on disease and disability. ciently obvious so that these diseases have ceased One method of doing this will be to compare the to be important occupational hazards. Such rate at which disease and disability develop in conditions as lead intoxication, mercury poison a population exposed to fairly well-defined ing, and benzol poisoning have largely disap working environments with an expected rate peared because they were easily identified as derived from a population not so exposed. In associated with the occupational environment, many instances this will call for considerably and industrial hygiene measures were relatively more sophistication in the application of epi simple and easily applied. In some instances, as demiologic methods than has been required in in bladder cancer in the dye industry and phossy the past. Any population is unique, and this jaw in the match industry, the use of certain seems particularly true of persons pursuing a substances was discontinued to protect the given occupation at a particular place and time. health of workers. No other experience of an7 .other population is It is certain that many important occupa precisely applicable for the estimation >>fin tional exposures remain which affect health and expected' rate. Nevertheless, estimates wiirbe elevate morbidity and mortality rates. These needed. This is a brief discussion of some prob exposures may either be primary in the etiology lems encountered in deriving such estimates. of disease or may merely serve to aggravate existing conditions. Hatch feels that the aggra Infrequent Diseases Dr. Enterline is chief. Biometrics and Social Studies Branch, Division of Occupational Health, Public Health Service. At the outset it should be noted that some dis eases occur so infrequently that only a few cases can produce information of epidemiologic sig- 973 j i r 752463 0015 ^ r'A*'.-'- AO4'* .. 1 h> Ltr^f nificance, find elaborate attempts to estimate expected rates are unnecessary. For example, the early discovery by Sir Percival Potts that scrotal cancer occurred with unusual frequency among chimney sweeps, and the conclusion that some agent in their working environment was responsible, was made relatively easy by the fact- that scrotal cancer is extremely rare, and only a few cases with a common occupational exposure are strong evidence of cause and effect. In a more recent example, Wagner and co-work ers reported on the occurrence of mesothelioma in South Africa (2). They accumulated a series of 47 cases, a remarkably large number in view of the total reported in the world's literature at the time of their report, and related nearly all of these cases to asbestos exposure. cpa Overwhelming Effects of Job ExpItoMsUurfeMs ' Some environmental exposures produce such an overwhelming response that elaborate at tempts to estimate expected rates are unneces sary. The effects of working in the production of chromates on lung cancer rates provide an excellent example. Machle and Gregorius studp, ied insurance records for seven chromate pro'-C.. ducing plants and compared the percentage of total deaths among chromate workers attributed to lung cancer with the percentage_in_a_group v of industrial policyholders in the Metropolitan Life Insurance Company (S). Among chro mate workers, 42 of a total 193 deaths were caused by lung cancer; among the industrial policyholders only 10 of a total 733 were at tributed to lung cancer. Annual mortality rates per 100,000 for cancer of the lungs and bronchi were also compared with rates for male employ ees of an oil refining company. necessary. Frequently,J^hetheoreticalquestion to be answered is whether illness occurs at the same rate in a cohort entering a particular occupation as it would Have occurred if this cohort had chosen another occupation similar in all respects to the occupation iinder" study except that the "extra" hnzard did not exist This type of question can probably best Be answered by dividing the subjects randomly into groups, with some groups exposed to the agent or agents under investigation, some not so exposed, and all groups followed to deter mine the incidence of disease and death. Al though this experimental method is clearly preferable, most investigations of the etiology of disease in human populations, for ethical or other reasons, usually must be based on the ob servation of naturally occurring events. How ever, the results of observational investigations can approach the results of experimental studies if reasonable care is exercised in the derivation of an expected rate. The advantage of the experimental investiga tion in studies of disease is that it permits con trol of all variables which might affect observ able disease so that the separate effects of the agent under investigation can be measured. Since many of these variables can be readily identified, such as age, sex, diagnostic criteria, economic level, and general living conditions, reasonable care in the selection of a control group can greatly improve the usefulness of observational studies. As Hill puts it, in con ducting such studies, "one must have the experi mental approach firmly in mind" (4) That is, in conducting observational studies one must attempt to "hold constant'^all relevant, varT ables other than tKe" variable under investiga tion! ' Occu pation Chromate workers___________ Oil rcfi ning company workers 50 years of age or less 197 5 Over 50 yearn of age AS0 22 Deriving Expected Rates For relatively frequent conditions with a variety of likely etiological agents and for which there are no overwhelming occupational effects, explicit comparisons with an expected rate are Measuring Nonoccupational Effects One of the reasons the rate of disease deviates in a given occupational group from some ex pected rate is the influence of nonoccupational factors. The problem is illustrated by data shown in table 1 (5, 6). Tuberculosis is usuallyassociated with work, in dusty trades, such as coal mining, and it is generally believed that dust inhalation is probably a potentiating factor in this disease. By comparison with workers 974 Public Health Reports 752463 0016 Table 1. Tuberculosis death rates per 100,000 among coal miners and other groups of workers, selected age groups, United States, 1950 Age group (years) Cool min ers All occupa tions Profes sional work ers Inter mediate occupa tions Labor ers 20-24.......... 23. 5 25-34.......... 35.7 35-44.......... 75. 6 45-54.......... 156.9 12. 3 18. 8 34.0 56. 8 2. 4 3.3 7.9 25.3 7. 2 13.4 2S. 5 49. 1 32. 8 58.5 100.2 143.3 Sources: References 5 and 6. in all occupational groups, coal miners have a high death rate, but compared with laborers, the death rate among coal miners is relatively low (table 1). Which is the proper compari son for an evaluation of the importance of the work environment? Certainly more data would be needed to answer the question than are available in table 1. In view of the rele vance of general living conditions to the ^ development of tuberculosis, it would seem im^ portant that the group selected for comparison ) should have living conditions roughly similar to those of coal miners. A method for detecting an effect of the non[ occupational environment on variations in ) disease rates among groups of workers is ilI lustrated by Higgins' data (7). He calculated 1 standardized mortality ratios for bronchitis among men in certain occupations in England and Wales and their wives (table 2). Without the similar ratios for wives one might readily conclude that bronchitis is associated with the working environment. study. Table 3 shows the actual number of deaths among miners and in the entire Prov ince for each of the 6 years of the study. For the Province, the number of deaths increased sharply during the period. The years 1954-55 were selected for computing expected rates on the grounds that at least part of the increase from 1950-55 was the result of improved recog nition and reporting of lung cancer and the years 1954-55 more nearly represented diag nostic levels as they related to miners. The outcome of the study hinged upon which years were used in computing the expected lung can cer death rate. Using 1954 and 1955 on which 'to base the expected number provides a contrast of 6 as compared with 12 actually observed, a difference not statistically significant (P>0.05). On the other hand, if death rates for the 6-year period had been averaged, as is usually done in this type of study, the contrast would have been 4 expected as compared with 12 actually ob- Toble 2. Standardized mortality ratios for bron chitis among agricultural workers, coal miners, and foundrymen, aged 20--64 years, England and Wales Occupation 1931 1951 Men Wives Men Wives Farmers 36 49 31 Agricultural workers 52 66 53 Hewers and geltcis__ ___ 170 182 200 Others underground_____ 126 102 93 Surface workers 172 225 131 Iron and steel foundry workers 179 238 177 52 82 190 177 139 217 Source: Reference 7. Comparable Diagnostic Criteria Another problem in developing expected dis Table 3. Lung cancer dearhs among oil male* ond among asbestos miners, Quebec Province, ease rates is illustrated by a study of mortality Canada from lung cancer among asbestos miners in Quebec, Canada (8). In this study a cohort of miners, identified in 1950, was followed for the Year All males Asbestos miners 6-year period 1950-55. Twelve deaths from cancer of the lung were noted. To determine whether this number was excessive, average an nual age-specific male death rates were com puted for the Province of Quebec for the years 1930........ ....................... - .. 1951......................................... 1952................................ ........ 195.1.............................. .......... 1954............ ............................ 1955..................... ................... 190 22(1 245 303 303 357 3 2 0 3 i 3 1954 and 1955 and applied to the cohort under Source: References. 66SS Jrr-trtp. ^=2) Vol. 79, No. 11, November 1964 u-d-t. ^y P. 60-55 ---P O 975 ^ !<-C- 752463 0017 served, a statistically significant difference (P<0.05). Since a simple test of statistical significance was decided on as a method for accepting or rejecting an association of asbestos mining and lung cancer, the question of diagnostic compa rability was crucial to the study. Unfortu nately no data were presented to support the authors' belief that diagnostic criteria differed in the manner specified between the population studied and the population from which ex pected rates were derived. As a result, doubt is cast on the conclusion that asbestos mining and lung cancer are not associated. This study illustrates the difficulty in epidemiologic studies of diseases for which diagnostic criteria may be in a state of flux. The sharp rise in the number of lung cancer deaths in Quebec Province is indeed suggestive of some change in diagnostic criteria or nomenclature during a 6-year period. It would be a difficult task, however, to demonstrate that this change did not apply equally to asbestos miners, partic ularly in view of the small numbers of lung cancer deaths noted in this group. Thus, if comparability of diagnostic criteria can be ques tioned, studies to detect modest environmental effects may be of limited value. Selective Factors Observable disease rates among employed persons are no doubt affected by nonoccupational environmental factors and by diagnostic criteria. In addition, selective factors act to modify observable disease. In general, work- Table 4. Lung cancer deaths and rates among du Pont males, 1956--60, and United States white males, 1958 Age Rates per 100,000 du Pont Ratio: deaths U.S. to du Pont U.S. du Pont rates rates All ages___ 25-44................... 45-54................... 55-64................... 88 23. 2 ' 30. 3 10 4. 0 5. 3 34 38. 0 50. 1 44 114. 0 141. S 1. 3 1. 3 1. 3 1. 2 1 Adjusted to the age distribution of the du Pont Co. SouncE: P.eference 9. Table 5. Observed and expected number of deaths from specified causes among em ployees of the cigarette division,. American Tobacco Co., October 1946--December 1952 Cause Ratio: Expected Observed expected number number to observed All causes 500 338 1. 48 Cancer, all forms 60 44 1. 36 Cancer of respiratory system 7 6 1.17 Cardiovascular diseases... 239 161 1.48 Coronary disease 81 65 1.25 Source: Reference 10. ing populations tend to be healthier than the total population from which they are drawn. Unhealthy persons frequently do not attempt to enter the labor market. Also, Many companies have a preplacement examination which further tends to msuxea~Kealthy work force. Mortality rates tor a working population in nonhazardous employments should vary, therefore, from rates for a general population because of selec tive factors. For example, for conditions which have a sudden onset and are rapidly fatal, rates should be similar ~for both groups, but the workers should have a lower rate than the geneial population for conditions with_predisposing iUhessFand long' periodsof incapacity,'such as rheumatic Keartjjeease. The influence of selective factors may be re flected in a study by Pell and Fleming (9). They compared the death rate for cancer of the lung among male employees of E. I. du Pont de Nemours and Company during the period 195660 with the U.S. white male death rate for 1958 (table 4). At all ages the U.S. rate was 30 percent higher than the rate among du Pont employees. One cannot help being impressed with the possible influence of selective factors in this instance since lung cancer is of sufficiently short duration to be only rarely cause for exclu sion on the basis of a pre-employment examina tion, and, since it usually occurs at older ages, is an unlikely reason for persons not entering the labor force. Alternative explanations for the low rate among du Pont employees jvould, of course, include more favorable environment al factors or a difference _in_diagnostic criteria. 976 Public Health Reports I 752463 0018 I: A study by Dorn and Baum of mortality among workers in cigarette factories in Virginia, North Carolina, andv Kentucky ex tends these observations to causes of death in addition to lung cancer (10). Mortality for selected causes of death was compared with expected mortality based on the experience of the general population of North Carolina and Virginia, the two States where most employees worked (table 5). Deatli rates for cardiovascu lar disease were a third lower among employees of these cigarette companies when compared with the general population, and for all forms of cancer, were roughly 5:5 percent lower. Both studies illustrate the possible influence of selective factors. Moreover, thev suggest that occupational hazards could be easily over looked when a rate for employees in a particular industry is compared with an expected rate based on a more broadly defined population. In a population with a lung cancer death rate So percent below that for the TT pnpnlpfjon, for example, some agent could l>e introduced V- ^ which would double the lungr cancer death rate \ -i without causing an alarming deviation above an expected rate ba5e9~~on_tlie ~total_lKS._e- perience. v ft Discussion In planning epidemiologic investigations of tlie contributions of occupational environments to relatively frequent diseases with a variety of likely etiological agents, and without over- Vol. 79, No. II, November 1964 977 752463 0019 whelming occupational effects, expected rates should be derived wherever possible as part of the original study design. Some of the work currently in progress at the Division of Occu pational Health, Public Health Service, illus trates the kinds of measures which will prob ably be increasingly important in the future if etiological agents in the occupational environ ment are to be identified. A long-term study of the development of lung cancer among uranium miners fortunately con tains sufficient detail regarding the work en vironment of these men to permit grouping by extent of their occupational exposure to radon, the likely etiological factor being studied. Thus, lung cancer mortality is being compared between uranium miners with high radon ex posures and miners with low exposures. A large-scale study of pulmonary diseases among bituminous coal miners includes examinations of other manual workers living in or near coal r mining groups. communities and A study of the of the wives of both mortality experience of men working in the asbestos textile industry includes for comparison a large cohort of men ; _ I working in about the same geographic area in the cotton textile industry. The question to be answered is whether asbestos is carcinogenic, particularly to the lungs. The similarity be tween the industrial processing of raw material which produces asbestos textiles and the indus trial process which produces cotton textiles pro vides an opportunity for a comparison which is likely to produce meaningful information. It is apparent that in these studies the com parisons proposed will be imperfect. Neverthe- ess, they represent attempts to refine expected rates and may ultimately point the way toward improvements in occupational disease epidemi ology. If a suitable provision for estimating expected rates cannot be incorporated in future epidemiologic studies of occupational environ ments, the desirability of elaborate data collec tion will be, in many instances, questionable. The credibility of epidemiologic associations for many important diseases will hinge on the extent to which the characterisllcs'df'tlie popu" Iation used for comparison approach the char acteristics of the population under study. REFERENCES (7) Hatch, T.: Major accomplishments In occupa tional health In the past fifty years. Amer Industr Hyg Assoc J 23: 1OS-113. MorchApril 1984. (2) Wagner, J. G., Sleggs, C. A, and Marchand, P.: Diffuse pleural mesothelioma and asbestos ex posure In the North Western Cape Province. Brit J Industr Med 17: 260-271 (1060). (3) Machle, W., and Gregorius. F.: Cancer of the respiratory system In the U.S. chromate-pro ducing industry. Public Health Rep 63: 11141127, Aug. 27, 1048. (4) Hill, A. B.: Observation and experiment New Eng J Med 248: 093-1001 (1053). (5) Enterline. P. E.: Mortality rates among coal miners: Amer J Public Health 54 : 758-768, May 1064. (6) Guralnick, L.: Socioeconomic differences in mor tality by cause of death: United States. 1930 and England and Wales, 1949-53. Presented at the meetings of the International Union for the Scientific Study of Population. August 1963. (7) Higgins. I. T. T.: An approach to the problem of bronchitis in industry: Studies in agricultural, miniug and foundry communities. In Indus trial pulmonary diseases, edited by E. King and C. M. Fletcher. Little. Brown A Co., Boston. 1960. (8) Braun. D. C.. and Trnan. T. D.: An epidemiologi cal study of lung cancer in asbestos miners. AMA Arch Industr Health 17: G34-653, June 1933. (9) Pell, S.. and Fleming. A. J.: Cancer of the lung in an industrial population. Arch Environ Health 4 : 135-13S. February 19C2. (10) Dorn, H. F.. and Baum, W. S.: Mortality among workers in 'cigarette factories. Industr Med Snrg 24 : 239-241. June 1935. I 978 Public Health Reports 752463 0020