Document MZLRyJGoR2rZGQw3gm9D8QKy

(Imw f -Wftifr<. cun 4I4-4U * c**. -ucfeo4. J 4.). postHeld, r of and one ppiy iling nd*. Occupations and Cigarette Smoking as Factors in Lung Cancar LESTER BRESLOW, M.D,, F-A^-HJk, UMAR HOA&UN, GLADYS RASMUSSEN, and HERBERT K. ABRAMS, MJ>,, F.AJM-LA. studies have implicated cigarette amok, Tkit fear fear ttadf t/in li--aI nUiM lending 10 eannaat kitrr tigdretta smoking with earncar. In addition. It UtU taaarei oteapetiemt that team la haaa tome atiolagital rriallonskin la the din ram. Thar need further ttadf, ee k it mggarned. lug tad certain occupations * -- es pecially those involving exposure to radioactive * and chromate ores 1 and asbestos *--in the causation of lung cancer. However, the number of per. sons known to have such occupational exposures in the United States is rela tively smalL + The sharp increase in long cancer mortality constitutes one of the Boat spectacular disease phenomena of the past two decades. In the United States the age-adjusted mortality rate for lung cancer climbed from 2.7 per 100,000 population in 1930 to 11.0 in 1948, a The present investigation was undertaEew four years ago: (1) to determine whether any additional occupations might be involved in lung cancer and thus deserve intensive investigation, and 751 to obtain data on the role of tobacco usage in lung cancer. more than fourfold increase. In at least one state, California, pulmonary cancer Methodological Considerations dow accounts for more deaths than does pulmonary tuberculosis, and among ">Im aged 4S-64 it causes 4 per cent of all deaths. ' Attempts to explain this rising mor tality as spurious, e-g, as due to im proved diagnosis of the disease, do not Hand up. The lung cancer death rate-- which unfortunately b still our bam Definitive study of whether as occu pation or cigarette smoking U a causa* tive factor in lung cancer requires that the incidence rate for the exposed group he determined. This rate is then com pared with that of t control population, the most acceptable control being all the rest of the population from which the approximation of incidence rate -- hat increased two and a hall times more Dr. Breslsw is chief, Burasa al Chreaie rapidly among men than among women. Phcstaa. Caltfaraia Suit Department el Pub One can hardly presume that physicians lic Health, Barhaitr, CallL: tad Mr. Hoaslla hare applied diagnostic improvements tad Mrs. lte--aeeea its site eieorined with this dapaitawet. Dr. Abums, whs ve so modi more completely to men than istmsriy chief, Bureaa of Adah Health. to women. Likewise, the increasing pro portion of lung cancer found as a cause Callfcrmia State Department of PabUe Health, is bow medical director, Health aad WcUem Fuad, Chlcefs Oflce, Theater tad Amenwsot of death among autopsies supports the Building Jtellers' Union, Local 23, Plicae*. .conclusion that we are dealing with a real increase in the disease.1 IB. Thu paper is hated oa a uodf supported hr a rwseres past (CS 9047) from the Na tional Caacv Inothel*. PHS, tad wee pew Environmental, changes appear to seated before the leduecrial Hypcae Section oBer the best explanation for the oi the Americas Publle Health Aseoetsrtea at the Eighty-first Aaaual Meetiag la New York, phenomenon. Previously published N. Y, November 1<X 1131 m SCF-FA-4320 IV 8004 0046 HFM -011708 Occupations and Cigarette Smoking as Factors in Lung Cancer LESTER BRESLOW, M.D., F.A.P.H.A., LeMAR HOAGLIN, GLADYS RASMUSSEN, and HERBERT K. ABRAMS, M.D., F.A.P.H.A. studies have implicated cigarette smok This jour-year study offers addi tional- evidence lending to connect heavy cigarette smoking with can cer. In addition, it lists several occupations that seem to have some etiological relationship to the dis ease. They need further study, as it is suggested. ing 2 5 and' certain occupations 0 -- es pecially those involving exposure to radioactive7 and chromate ores8 and asbestos9--in the causation of lung cancer. However, the number of per sons known to have such occupational exposures in the United States is rela tively small. f. The sharp increase in lung cancer mortality constitutes one of the most spectacular disease phenomena of the past two decades. In the United States the age-adjusted mortality rate for lung cancer climbed from 2.7 per 100,000 population in 1930 to 11.0 in 1948, a The present investigation was under taken four years ago: (1) to determine whether any additional occupations might be involved in lung cancer and thus deserve intensive investigation, and (2) to obtain data on the role of tobacco usage in lung cancer. more than fourfold increase. In at least one state, California, pulmonary cancer Methodological Considerations now accounts for more deaths than does pulmonary tuberculosis, and among males aged 45-64 it causes 4 per cent of all deaths. Attempts to explain this rising mor tality as spurious, e.g., as due to im proved diagnosis of the disease, do not stand up. The lung cancer death rate-- which unfortunately is still our best Definitive study of whether an occu pation or cigarette smoking is a causa tive factor in lung cancer requires that the incidence rate for the exposed group be determined. This rate is then com pared with that of a control population, the most acceptable control being all the rest of the population from which the approximation of incidence rate -- has increased two and a half times more rapidly among men than among women. One can hardly presume that physicians have applied diagnostic improvements so much more completely to men than to women. Likewise, the increasing pro portion of lung cancer found as a cause of death among autopsies supports the conclusion that we are dealing with a real increase in Lhe disease.1 Environmental changes appear to offer the best explanation for the phenomenon. Previously published Dr. Breslow is chief, Bureau of Chronic Diseases, California State Department of Pub lic Health, Berkeley, Calif.; and Mr. Hoaglin and Mrs. Rasmussen are also associated with this department. Dr. Abrams, who was formerly chief, Bureau of Adult Health, California Slate Department of Public Health, is now medical director, Health and Welfare Fund, Chicago Office, Theater and Amusement Building Janitors' Union, Local 25, Chicago, 111. This paper is based on a study supported by a research giant (CS 9047) from the Na tional Cancer Institute, PHS, and was pre sented before the Industrial Hygiene Section of the American Public Health Association at the Eighty-first Annual Meeting in New York, N. Y., November 10, 1953. 171 172 FEB., 1954 AMERICAN JOURNAL OF PUBLIC HEALTH exposed group came. Assuming the same diagnostic and reporting standards for both groups, one may thus deter mine how the chance that a member of an exposed group will develop lung cancer compares with the chance that such a person would develop the dis ease if he were not a member of the exposed group. If the likelihood of de veloping lung cancer is substantially in creased by being in the exposed group (e.g., a certain occupation) one may conclude that the exposure is a causative factor in the disease. The key to this analysis lies in determining comparative incidence rates for a suspect group and a control population. But first it is necessary to identify suspect exposed groups. That was the aim of the present study. The im portance of pin pointing these groups arises from the fact that incidence studies require assembling sizable popu lations and observing the occurrence of the disease in them--a large and ex pensive task. If, for example, data were obtained implicating occupations beyond those already known to be causative factors in lung cancer, one would be justified in assembling popu lations of the occupational groups and observing them for lung cancer inci dence rates--a study which would de velop more definite proof of the presence or absence of an etiological relationship. Several methods of investigation have advanced our knowledge in this sphere. First, is clinical observation of associa tion between a case or a few cases of lung cancer and some exposure, e.g., asbestos. One defect of this method is that such clinical observation depends largely on chance and often involves such small numbers of cases that the validity is questioned. Where made, of course, such observations are helpful. Another common method of identi fying suspect occupations in a highly fatal disease, such as lung cancer, is to note the frequency of the various occu pations on death certificates where the disease appears as the cause of death. The frequencies of the occupations on such death certificates may then .be com pared with those on all death certificates for the general population. If certain occupations appear with much greater frequency on death certificates for lung cancer than on certificates where death is due to other causes, one may conclude that the occupation may be involved and should be intensively investigated. Important achievements may be credited to the use of this method. However, major difficulties with this use of death certificate data arise from the following circumstances: only one occupation may be recorded; while this is supposed to be the "usual occupation," it is often not correctly recorded and duration in the occupation is not recorded. All these circumstances tend to blur the identifi cation of suspect occupations from death certificates. The method selected in the present study was to determine the lifelong occupational and tobacco-usage history of lung cancer patients (and a control group) by actual interview with the patients. The work history included the number of years engaged, name of com pany, industry, a description of duties, and listing of products and hazardous materials known to the patient. Data Collection Patients observed during 1949-1952 in 11 California hospitals (three county general hospitals, five veteran or mili tary, two university, and one private hospital) comprised the study group. An interviewer especially qualified in occupational analysis obtained the data from each patient in the cooperating hospitals whose condition permitted the interview. The session with each pa tient often lasted two hours or more, since many details about smoking were LUNG CANCER VOL. 44 173 obtained and the information about Method of Analysis every occupation was highly detailed. A total of 518 patients who finally had histopathologically proved diagnoses of Inns cancer were ultimately included in the analysis. # For every patient in the case series another patient admitted to the hospital about the same time--of the same age (within five years), sex, and race--for a condition other than cancer or a chest " disease was chosen in a random manner as a matched "control." Two inter viewers, one working in Northern Cali fornia and one in Southern California hospitals, obtained the data from the lung cancer patient and control patients using the same schedule in identical manner. That interviewer bias did not enter into the determination of tobacco- usage history seems likely, among other reasons, because the results closely parallel those obtained in other investi gations where such bias apparently was excluded,4'5 and because the results for the two interviewers were very similar. It is difficult to conceive how inter viewer bias might have affected the occupational data, since both lung cancer patients and control patients mentioned hundreds of occupations and none of these was suspected by the in Each record was reviewed by a physician, particularly with respect to the pathologic diagnosis. An industrial hygiene chemist completed the listing of materials usually involved in the occupations mentioned by the patients but not necessarily known to them. Thereafter, the data were coded and tabulations prepared. Attempts to analyze the data concern ing type of industry and exposure to materials already identified as toxic or carcinogenic did not prove fruitful. Hence the major analytical effort with respect to work history went into the data on occupation itself. This had been coded according to the Dictionary of Occupational Titles (2nd ed., 1949), prepared by the Division of Occupa tional Analysis, U. S. Employment Service, a code with 367 three-digit titles. Particular attention was devoted to the question of years of exposure be cause present knowledge of carcino genesis due to environmental agents in dicates that duration of exposure and a long latent period are critical factors. Data on tobacco usage were analyzed so as to reveal the type and degree of smoking, as well as the fact of smoking itself. terviewers. Preliminary findings were Finally, an attempt was made to sepa kept from the interviewers until the rate the effects of tobacco usage and end of the data collection. occupations which appeared implicated. Table 1--Age Distribution of Lung Cancer Patients and Controls Age Total Untlcr 40 40-49 50-59 60-69 70 and over Adeno carcinoma Cases Other Types of Carcinoma 46 3 6 19 14 4. 472 22 56 194 154 46 Total 518 25 62 213 168 50 Controls Total 518 26 59 213 169 51 174 FEB., 1954 AMERICAN JOURNAL OF PUBLIC HEALTH Table 2--Sex and Race Distribution of Lung Cancer . Patients and Controls Sex and Race Total Total male White Negro Other Total Female White Negro Other Adeno carcinoma 46 42 41 ,. 1 4 3 1 Cases Other Types of Carcinoma 472 451 420 20 11 21 15 3 3 Total 518 493 451 20 12 25 18 4 3 Controls Total 518 493 461 20 12 25 18 4 3 Findings on Smoking Tables 1 and 2 reveal the distribution of the lung cancer patients according to their biological characteristics -- age, sex, race, and the essential similarity of the matched control series in respect to these characteristics. Tobacco usage by lung cancer pa tients and the control group is shown in Table 3. Here one may note that 484 (93 per cent) of the cases gave his tories of having smoked cigarettes; whereas only 394 (76 per cent) of the controls gave a similar history. No such greater frequency of tobacco usage in other forms (pipe, cigar, chew, and snuff) occurs among the lung cancer patients when compared with the con trol group. In fact, 68 of the control group used tobacco in a form other than cigarettes exclusively; whereas only 15 of the lung cancer patients did so-- relative frequencies which are not much Table 3--Tobacco Usage During 20 Years Prior to Study Reported by Lung Cancer Patients and Controls Manner of Using Tobacco Total Total not smoking cigarettes No use ol tobacco Tobacco other than cigarettes, only Total smoking cigarettes Cigarettes, only Cigarettes and other Adeno carcinoma 46 6 4 2 40 31 9 Cases Other Types of Carcinoma 472 28 15 13 444 298 146 Total 518 34 19 15 484 329 155 Controls Total 518 124 56 68 394 240 154 'H ig Cancel* Controls ;al Total 518 . LUNG CANCER VOL. 44 175 Xal>*e 4__ Amount of Cigarette Smoking During 20 Years Prior to Study Reported by Lung Cancer Patients and Controls Package? of Cigarettes Smoked per Day Total Cases Number Per cent 518 100 34 7 19 4 71 14 304 59 80 15 10 2 Controls Number Per cent 518 100 124 24 50 10 105 20 199 38 22 4 18 3 ing smoked cigarettes;! 94 (76 per cent) of the! similar history. No suchj icy of tobacco usage iia pipe, cigar, chew, and among the lung cancej! compared with the com fact, 68 of the controjf icco in a form other tha| sively; whereas only lSl incer patients did so cles which are not muchS tudy Reported by Total 518 34 19 15 484 329 155 Controls j Total l 518 124 -I 56 j 68 i 394 i 24 J 154 ]t jlijfferent for the nonusers of tobacco in L> cMje,.*two seri*es. IpThis suggestion of an effect by cig arettes only is consistent with other in stigations in the field. Some support ftSi the hypothesis that cigarette smokr-affects the development of epithelial jcinoma of the lung more than adeSfearcinoma may be noted. Six out i4>6 (13 per cent) of the cases of ienocarcinoma did not smoke cigjettes; whereas only 28 out of 472 M per cent) of the patients with other Tj|ies of carcinoma, described variously squamous, epithelial, undifferen$ted, did not smoke cigarettes. pJne commonly used measure of the jt'Ount of exposure to cigarette smokso ing, the average number of packs smoked per day, appears in Table 4. Seventy-four per cent of all lung cancer patients reported smoking one or more packs of cigarettes on the average per day over the preceding 20 years, com pared with a frequency of only 42 per cent among the controls. "Excessive cigarette smoking" (two or more packs per day) occurred almost four times as commonly among the case series as among the control group. Other items which may indicate the amount of cigarette smoking include the age at which the individual began smoking, whether or not he smokes be fore breakfast, and whether or not he "inhales." Table 5 reveals that, by all 5--Ago of Beginning Cigarette Smoking, Smoking Habits, and "Cigarette Cough" Reported by Lung Cancer Patients and Controls id. Cigarette Smoking Characteristics Number u. al cigarette smokers Bogan 15 years of age egan 15-24 25 years ol a^e jjhale" jj^luke belore breakfast " paretic cough" 518 484 166 286 32 449 373 204 Cases Per cent Based on Cigarette Smokers Only 100 34 59 7 93 77 42 Per cent Based on Total 100 93 32 55 6 87 ' 72 39 Number Controls Per cent Based on Cigarette Smokers Only 518 394 100 116 29 243 62 35 9 327 83 229 58 112 28 Per cent Based on Total 100 76 22 47 7 63 44 22 176 FEB., 1954 AMERICAN JOURNAL OF PUBLIC HEALTH Table 6--Estimated Prevalence Rates * of Lung Cancer per 100,000 Males of Selected Ages and Various Cigarette Smoking Habits Number Estimated Prevalence Rates Quantity Smoked Daily t Lung Cancer Control All Quan Heavy * * tities None Relative Prevalence t Heavy All Quan tities Patients aged 50-59 years 1. Doll and Hill 276 275 94 42 7 13 6 2. Wynder and Graham 258 210 117 47 4 27 11 3. Sadowsky, Gilliam, and Cornfield 198 208 87 44 12 7 4 4. Present 213 213 102 46 6 17 7 Patients aged 60-69 years 1. Doll and Hill 197 199 137 63 0 ., ,, 2. Wynder and Graham 187 160 138 70 5 29 15 3. Sadowsky, Gilliam, and Cornfield 76 71 44 71 5 9 4 4. Present 168 169 433 78 17 25 5 * Estimated according to the method of Cornfield.11 Data for studies of Doll and Hill; Wynder and Graham; aod Sadowsky, Gilliam, and Cornfield from the latter paper.13 f Quantity smoked daily, except for present study, is all forms of smoking expressed as cigarettes, with pipes and cigars converted to their equivalents in cigarettes as follows: 1. Doll and Hill--1 os. of tobacco a week =: 4 cigarettes a day; 2. Wynder and Graham--1 cigar -- 5 cigarettes, 1 pipeful = 2% cigarettes; and 3. Sadoweky, Gilliam, and Cornfield---1 cigar = 10 cigarettes, 1 oz. pipo tobacco -- 20 cigArettes. 4. For present study, quantity smoked daily includes only cigarette smoking. + Relative prevalence is the.ratio: Heavy--rata among heavy smokers/rate among nonsmokera; all quantities--rats among 6niokera/rste among nonsmokers. ** Heavy smoking is defined as the following number of cigarettes or equivalents of cigarettes per day:--1. Doll and Hill--50 or more; 2. Wynder and Graham--35 or more; 3. Sadowsky, Gilliam, and Cornfield--41 or more; and 4. present study--40 or more. of these criteria, the lung cancer pa tients smoked cigarettes to a greater ex tent than did the control group. Also, approximately two-fifths of the lung cancer patients stated that they had had a "cigarette cough" usually antedating the onset of their illness by many years; 120 said that they had had a cough for five years or more. The frequency of such "cigarette cough" was only about one-half as great among the control group and only 39 of the control pa tients indicated that their cough had lasted five years or more. Significance of Findings on Cigarette Smoking The data clearly show cigarette smok ing to be more frequent and intense among lung cancer patients than among the control group. Our data parallel that of other recent, well controlled studies; all indicate a positive correla tion between cigarette smoking and lung cancer, a correlation which increases steadily with the amount of cigarette smoking. One can hardly escape the conclusion of Doll and Hill4 that "smoking is a factor and an important factor in the production of carcinoma of the lung," and the more recent con clusion of Doll,10 "The results amount, I believe, to proof that smoking is a cause of bronchial carcinoma." The cigarette smoker or prospective cigarette smoker may wish to know how much his chances of developing lung cancer are increased by such smoking. Acknowledging that a correlation exists, one may still inquire how many times does cigarette smoking increase the per 100,000 Males ig Habits ates lyt lone Relative Prevalence t ; Heavy All" Quantitiea : 7 13 6i 4 27 11 : 12 7 6 17 4 -i 7; 0 5 29 15 ^ 59 17 25 4j 5 ; Hill; Wynder and Graham; and! sed as cigarettes, with pipes atuh| >f tobacco a week = 4 cigaretleij and 3. Sadowsky, Gilliam, aa| nt study, quantity smoked daily! nonemokera; all quantitlea~ratej i of cigarettes per day:--1, Doirj and Cornfield--41 or more; aodl p. Our data parallel! ecent, well controlled! ate a positive correla-i rette smoking and lungI ation which increases! j amount of cigarette! ;an hardly escape thel Doll and Hill4 that? ctor and an important! jduction of carcinoma! I the more recent con-j "The results amount, I! f that smoking is aj il carcinoma." smoker or prospective! may wish to know hotfl ;s of developing lungl ised by such smoking! lat a correlation exists,! juire how many time*! smoking increase thej LUNG CANCER VOL. 44 177 Elihood of lung cancer? Further, "Sp j9 that likelihood in absolute s? To answer these questions lately would require that records of ikrette smoking for large numbers of aons in the lung cancer age be .Junbled. These persons would then l^laced in categories with respect to Ikrette smoking, e.g., nonsmokers, Jlerate smokers, and heavy smokers, iflaily they would be observed for a Bttibient length of time to determine the Se of lung cancer among the various Agones of cigarette smokers and the ppsmokers. (Such studies are now Jjader way.) ^However, statistical devices have been developed for data such as those in the Dfesent investigation which, if one is S&Ajm^g to make certain assumptions, an answer to the question, how Ipjany times does cigarette smoking in crease the likelihood of lung cancer? Application of Cornfield's formula l1,12 to the data from four studies, including the present one (Table 6), indicates that for men aged 50-59 the prevalence of lung cancer is from four to 11 times as great among smokers as among non smokers, and from seven to 27 times as great among heavy smokers as among nonsmokers. Similar ratios were found for the 60-69 age group. This formula involves the assumption that in so far as smoking history is concerned, the pa tients are a representative sample of all lung cancer cases and the controls are a representative sample of all other per sons in the general population. Findings on Occupations As noted earlier, the records for lung cancer patient and controls showed all Table 7--Twenty Per cent Sample from Complete List of Occupational Groups 1 in Which There Were at Least 5 Cases or 5 Controls Employed for 5 Years or More Occupation Persons Employed in Occupation at Least 5 Years Cases Controls : 006 Authors, editors, and reporters 'F 088 Ship captains, mates, pilots, and engineers ! : 105 Clerks, general office -' 155 Canvassers and solicitors ` 180 Salesmen to consumers 227 Waiters and waitresses, except private family V. 266 Policemen and detectives, public service - 307 Animal and livestock farmers 317 Farm hands, animal and livestock ... 433 General woodworking occupations, not elsewhere classified 516 Painters, except construction and maintenance vt-.v- 524 Brick and stonemasons and tile setters 532 Construction occupations, not elsewhere classified " '4i 542 Locomotive firemen 560 Attendants, filling stations and parking lots . . 583 Mechanics and repairmen, not elsewhere classified 7 8 2 3 7 14 6 6 10 8 3 5 45 3 10 15 9 8 8 11 4 10 3 5 10 7 6 1 25 6 9 16 Note: Semple rundomly selected from the 81 occupation groups which met the stoted frequencyduration requirements (at least 5 esses or 5 controls with 5 years or more employment). Occupa tions classified according to 3-digit categories in the Dictionary of Occupational Titles (2nd cd.). Washington, D. C.: U. S. Employment Service, 1949, 178 FEB., 1954 AMERICAN JOURNAL OF PUBLIC HEALTH Table 8--Occupation Groups * in Which the Frequency of 5 Year Employment Among Cases Was at Least Twice as Great as Among Controls Occupation Persons Employed in Occupation at Least 5 Years Cases Controls 016 Engineers, civil 138 Stock clerks 266 Policemen and detectives, public service 430 Lumbermen, raftsmen, and wood choppers 473 Machine shop and related occupations, not elsewhere classified 485 Welders and flame cutters 520 Occupations in production of petroleum 523 Construction machinery operators, not elsewhere classified 524 Brick and stonemasons and tile setters 538 Brakemen, railroad 570 Firemen, other than process firemen 573 Cranemen, derrickmen, hoistmen, and shovclmen 575 Drillers, extraction of minerals and construction 5 7 6 11 8 10 5 7 5 10 11 15 7' 1 3 3 5 3 1 2 3 1 5 2 6 2 * Occupations classified according to tlia Dictionary of Occupational Titles (2nd ed.). Washington, D. C.: U. S. Employment Service, 1949. Includes only occupations in which there were at least S cases with S years or more uf employment. occupations followed and their dura tions in years. A listing was prepared of the occupations pursued by at least five patients or controls for five years or more. This listing revealed 81 occu pations in this arbitrary frequency-andduration category. Table 7 presents a random sample of these occupations and the number of patients and controls who at some time during their work history had been en gaged in them for five years or more. It will be noted that among the 518 lung cancer patients seven had worked as authors, editors, or reporters (occu pation) for five years or more, and among the 518 controls, nine had been similarly engaged. Relative frequencies of occupation would appear to have the same logical significance as such fre quencies of cigarette smoking. The lat ter is, of course, much more common among both case and control groups than is any single occupation. Hence, the numbers pertaining to any one oc cupation would be less impressive than in the case of cigarette smoking. How ever, the frequency with which the lung cancer patients had pursued certain oc cupations was substantially greater than the frequency with which control pa tients had pursued these same occupa tions. The occupations for which the fre quency among cases was at least twice as great as among controls, using the same frequency-duration category as before, appear in Table 8. Such a tabu lation, of course, shows the occupations classified according to an arbitrary code which was not devised to reflect com mon occupational exposures that might he responsible for lung cancer. Further analysis therefore included the de-. tailed examination of records of in dividual cases and controls and re grouping of occupations according to possible common exposures. For ex- y of 5 Year e as ns Employed in pat ion at Least 5 Years s Controls 1 3 3 5 3 1 2 3 1 5 2 6 2 Titles (2nd cd.). n which there were ning to any one ocless impressive than ette smoking. Howwith which the lung pursued certain ocantially greater than which control pathese same occupa- for which the fres was at least twice controls, using the ration category as tble 8. Such a tabuiows the occupations to an arbitrary code ised to reflect comxposures that might ung cancer. Further included the deof records of inl controls and reations according to exposures. For ex- EH-' !>:-: ample, one occupational group, number HiIi?So (plumbers, gas fitters, and steam ^fitters) showed 16 cases and 10 controls 4,1 a five-year employment history, i Examination of these 26 individual Bt*ecords revealed that substantially all ^Idhe difference between case and control Series was due to the steam fitters who sjiad a heavy exposure to asbestos. It was also noted that other types of work gsar. Evolving asbestos occurred more fre quently in the employment history of peases than among the employment his- pftory of controls. Thus, a new category, M'including steam fitters, asbestos workers, * "Hand boilermakers--all involving expoSt'sure to asbestos--was found to occur 10 times among the cases and only one time among the controls. Table 9 presents the data for occupa"V'tions grouped according to possibly significant exposures. Although numijers in each of the categories are small, the differences between cases and con trols appear to implicate the occupations listed. It is of interest to note that few; several of those with the most striking _ differences involve exposure to metallic particles and fumes and products of metallic combustion. This observation tends to corroborate the suggestion of Wynder and Graham 13 that "hot metal" LUNG CANCER VOL. 44 179 occupations may be involved in lung cancer. In recent data of Doll10 the only occupation with an important excess in the cancer group was that directly con cerned with the production of gas. Little or no support was found indicating that other occupations suffer special risks. However, Doll's study was based upon experience in England where the em ployment pattern differs from that in the United States, especially California. Also, Doll used an occupation code of only 76 categories which may have been too coarse to reveal the differences found in the present investigation where a code of 367 categories was used and analysis carried to subgroupings of these. It may be noted that 77 persons with lung cancer had an exposure of more than five years to suspect occupations (unduplicated count in seven occupa tion groups) according to the data in Table 9. This number represents a not insignificant proportion of the total (518) lung cancer patients in the entirestudy. A further question upon which the present data might be expected to throw some light is the extent to which ciga rette smoking and occupations operate Table 9--Cases and Controls Employed for at Least 5 Years in Selected Occupation Groups * Occupation 1. Welders and slice! metal workers doing welding 2. Steam filters, boilermakers, and asbestos workers 3. Electric bridge crane operators--metal industry 4. Occupations in ibe extraction of lead, zinc, and copper ore 5. Marine engineers, firemen, oilers, and wipers 6. Construction and maintenance painters 7. Cooks, commercial (excluding cannery cooks) Persons Employed in Occupation at Least 5 Years Cases Controls 14 2 10 1 51 93 12 6 22 12 35 21 Occupation groups based on rogroupiug according to common occupational exposures after dciailcfl cxnniinalion of rcronlo. 180 FEB., 1954 AMERICAN JOURNAL OF PUBLIC HEALTH Table 10--Example of Adjustment for Smoking Applied to Welders and Sheet Metal Workers Doing Welding None Packages of Cigarettes Smoked per Day <y. %- 1- 2 or Not re more corded Total All cases and controls, male * 1. Cases 2. Controls 3. Total 4. Proportion of cases among total cases and controls Welders and sheet metal workers doing welding, male 5. Cases 6. Controls 7. Total 8. Expected number of welder cases i' 22 110 132 0.167 0 1 1 0.167 16 69 296 80 45 105 193 22 61 174 489 102 0.263 0.396 0.605 0.783 02 00 02 11 1 12 1 0 1 0.0 0.792 7.260 0.783 10 493 18 493 28 986 0.357 0 14 02 0 16 0.0 9.002 9.002 Expected proportion of cases among welders = -- = 0.563 16 14 Observed proportion of cases among welders =: -- =: 0.875 16 14 Ninety.five per cent confidence limits t for the proportion -- = (0.617, 0.984)--statistically significant at 95 per cent level. 16 * Females excluded from total group, since only males are included In groups to which test was applied. t Obtained by applying the proportion of cases among total cases and controls In each smoking category to the number of welders in that smoking category [line No. 4 X line No. 7j. J Hald, A. Statistical Tables and Formulas. New York: John Wiley, 1953. Note: This test also was applied to all other occupation groups shown in Table 9. The group 4<steam fitters, boilermakers, asbestos workers" was on the borderline of statistical significance at the 95 per cent level. None of the <9ther groups met the 95 per cent level of statistical significance. as separate factors in lung cancer. For example, both cigarette smoking and welding appear to be related to the disease, but do welders get lung cancer more frequently merely because as a group they are heavy cigarette smokers? Although data from the present study involve small numbers for many of the occupations, analysis does indicate that welding as an occupation operates as a factor distinct from cigarette smoking in relation to lung cancer (Table 10). The group of steam fitters, boilermakers, and asbestos workers lies on the border line of statistical significance when the ^effect of cigarette smoking is controlled. In the other occupational groups the numbers are such that the differences might arise from chance variation more than five times in a hundred. However, all the occupations listed in Table 9 deserve intensive study to determine their precise relationship to lung cancer. One important step is to assemble populations of individuals em ployed in these occupations, and with varied smoking histories, in order to observe their lung cancer incidence rates. These rates may then be com pared with the rate for the correspond ing age-sex group in the general population. One may thus determine plied to Welders ng ked per Day 2 or Not re more corded Total 80 10 493 22 18 493 102 28 986 0.783 0.357 1 0 1 0.783 0 14 02 0 16 0.0 9.002 -statistically significant at 95 pet which test was applied. in each smoking category to the le 9. The group Msteam fitters, 9$ per cent level. None of the ;cupational groups the ;h that the differences chance variation more i a hundred, the occupations listed rve intensive study to precise relationship to le important step is to \ ions of individuals em-; occupations, and with histories, in order to ung cancer incidence tes may then be com ate for the "correspondoup in the general e may thus determine LUNG CANCER VOL. 44 181 J.'- keing in a particular occupational PP up affects the chances of developing e cancer and likewise how cigarette looking modifies these chances. and mechanism responsible should be identified in order that protective meas ures may be introduced. REVERENCES Conclusion ffr data in this study constitute still ^1- other link in the chain of evidence ILnecting lung cancer with cigarette Jamoking, evidence which the New Eng|nd Journal of Medicine notes to be ^so strong as to be considered proof vlyithin the everyday meaning of the Jtword." 14 it is lime for those concerned Jjlftth health education and with the Ipfethics of advertising, at least in health iltiahd medical journals, to take note. JSiFurther investigation should seek the jppfexact component of cigarette smoke-- ftnftnr arsenic, or other--which is respon"Rfsible. .gs^The data also suggest that several ^^occupations (Table 9), in addition to . ^iithose. previously identified as having an -^etiological relationship, may be involved jn the development, of lung cancer. Hi'hese occupations should now be inijtensively studied to determine to what extent persons engaged in them suffer ^special risk of lung cancer. There after, if an etiological relationship is clearly established, the precise exposure 1. DalT, M. E., Doll, R., and Kennawny, E. L. Cancer of (he Long in Relation to Tobacco. Brit. J. Cancer 5:1, 1951. 2. Schrek, R., Baker, L. A., Ballard, C. P., and Dnlgoff, S. Tobacco Smoking as nu Etiological Fac tor in Disease. Cancer Research 10:49-58, 1950. 3. Wyndcr, E. L., and Graham, E. A. Tobacco Smok ing as Possible Etlologic Factor in Bronchiogcmc Carcinoma. A Study of 664 Proved Cases. J.A.M.A. 143:329-336, I960. 4. Doll, R., and Hill, A. B. A Study of the Aetiology of Carcinoma of the Lung. Brit. M. J. 2:1271-1286, 1952. 5. Levin, M. L,, Goldstein, H., and Cerltardt, P. R. Cancer and Tobacco Smoking. A Preliminary Re port. J.A.M.A. 143:336-338, 1950. 6. Hucpcr, W. C. Occupational Tumors and Allied Diseases. Springfield, III.: Thomas, 1942. 7. Rostoski, 0., Saupe, E., and Schmorl, G. Die Bergkronkhcit dor Erzbergleutc in Schneeberg la Sachsen ("Schneeberger Lungcnbrebs"). Ztsclir. f, Kicbsforsch 23:360-684, 1926. - 8. Macklc, W., and Gregorious, F. Cancer of the Respiratory System in the United States. Pub. Health Rep. 63:1114^1127, 1948. 9. Gloyne, S. R. Pneumoconiosis; Histological Survey of Necropsy Material in 1205 Cases. Lancet 1:810, 1951. 10. Dolt, R. Bronchial Carcinoma: Incidence and Aeti ology. Part I and Part II. Brit. M. J. 521-527, 585-590, 1953. 11. Cornfield, J. A Method of Estimating Comparative Ratos from Clinical Data. Application to Cancer of the Lung, Breast, and Cervix. J. Nat. Cancer Inst. 11:1269-1275, 1951. 12. Sadowsky, D., Gilliam, A., and Cornfield, J. The Statistical Association Between Smoking and Car cinoma of the Lung. Ibid. 13:1237-1258, 1953. 13. Wynder, G. L., and Graham, E. A. Etlologic Factors In Bronchiogenic Carcinoma with Special Reference to Industrial Exposures; Report of 857 Proved Cases. Arch. Inrlust. Hyg. & Occup. Med. 4:221, 1951. 14. Editorial. Cancer of the Lung. New England J. Med. 249:465-166, 1953. Journal 25 Years Ago Addendum to the January, 1954 "25 Years Ago." In addition to the articles fe(9n &e common cold mentioned last month, the 1929 Journal (p. 449) describes eight-page booklet "That Mean Cold," by present-day associate editor of the journal, Raymond S. Patterson. It was reportedly his first production as director health education for the John Hancock Mutual Life Insurance Company, the 'January issue (p. 94) having reported his appointment as director of health education, Life Conservation Service of that company. Dr. Patterson had already been the author of A.J.P.H.'s "Annotated Public Health Bibliography" for several [years, it having first appeared under his by-line in March, 1926.