Tobacco Documents Online

Y l Smoking and Other Risk Factors for Lung Cancer In Women %1' Anna H. Wu, Ph.D.,3' Brian E. Henderson, M:D.,3 Malcolm C. Pike, Ph.D.,3' and Mimi C, Yu, Ph:D:3•s ABSTRACT-A case-control study, aamong white women in Los Angeles County was conducted to investigate the role of smoking ;nd other factors in the etiology of lung cancer in women. A total of 149 patients with~ adenocarcinoma (ADC) and 71 patients with squamous cell carcinoma (SCC) of the lung and their age- and sex-matched controls were interviewed. Personal cigarette smoking accounted for almost a11 of SCC and'about hall of ADC in this study population. Among nonsmokers, slightly elevated relative risk(s) (RR) for ADC were observed for passive smoke exposure from spouse(s) 1RR=1.2: 95% confidence interval (CI)=0.5, 3.3),and at work (RR=1.3; 95%' C1=0.5, 3:3). Childhoodpneumonia (RR=2.7; 95% CI=1.1, 6:7) and childhood exposure to coal buminp (RR=2.3; 95% CI=1.0, 5.5) were additional',risk,fseton f'or ADC. For both ADC and SCC, increased risks were associated with decreased intake of fi-carotene foods but not for tota), preformetl vitamin A foods and vitamin supplements.-JNCI 1985; 74:7a7-75t Lung cancer is now the fourth most common cancer in women (1) land has been projected to be the leading cause of cancer mortality among women by the mid-1980's (2). Causes of lung cancer, other than cigarette smoking (3);, have not been clearly identified, but associations with exposure to passive smoking (4-6); exposure to combus- tion products of heating and cooking fuels (7), and occupational exposures (8=1'0) have been suggested. In addition, lung "scarring" (11) and a low dietary intake of ,6•carotene (12-14) and preformed vitamin A(15-17) may increase the risk of lung cancer. This paper reports a case-control study of' ADC and SCC of the lung in white females in Los Angeles County. Each of the above-mentioned factors was investigated. METHODS Female patients diagnosed with primary ADC or SCC of the lung were prospectively identified by the CSP, the population based tumor registry for Los Angeles County. (18), between April 1, 1981,, and August 31, 1982. On the basis of inforrnation collected routinely by the CSP, we limited eligibility to white Los Angeles County residents, with no history of cancer (other than non-melanoma skin cancer) and under age 76 at diagnosis; we verified these variables at interview. We also excluded cases if they were born outside the United States, Canada, or Europe; were not English-speaking; or were not residents of Los Angeles County at the date of diagnosis. A total of 490 eligible cases were identified. Of these patienu, 190 had died or were too ill to participate by the time we contacted their attending,physician. Permission was granted to contare272 of the remaining 300 patients. Eight patients were not located, and 44 refused to be interviewed so that we obtained completed question- naires on 220- On the basis of information on the CSP abstract, no significant differences were noted' betaveen. 747 those interviewed and those not interviewe& in terms of age, marital status, religion; and smoki~ng status recorded on medical: records. However, those who were not interviewed were more likely to have distant metastases at the time of diagnosis (58%) compared to those who were interviewed (11%): Comparable percentages of eligible SCC (43%) and ADC (46%) patients were interviewed. We selected one individually matche& neighborhood control for each interviewed case. The control had to fulfill all the criteria given above for cases (with reference date taken to be the same as that of the matching case) and', in addition, was matched with the case onidate of birth (±5 yr of birth date). Our control selection algorithm defined a specified sequence of houses to be visited in the neighborhood where the case lived atdate of diagnosis. Our goal was to interview the first eligible resident in this sequence. If no one was home at the time of the visiti we left an explanatory letter and, made a follow-up visit after several days. For any patient, 80 housing units were 'vi'sited' and' 3 return visits: were made before failure to secure a matched control was conceded. In 150 instances the first eligible person agreed to pa. rticipate, in 55 instances the second eligible control in the sequence was interviewed, and in 15 instances the third eligible control was interviewed. Cases and controls were interviewed on the telephone with the use of a structured questionnaire designed to elicit information on personal smoking habits, exposure to passive tobacco smoke, lung diseases, dietary intake of vitamin A, types of heating and cooking fuels ever used, and reproductive history. We also obtained a lifetime history of all jobs (job title„activities, and exposure) of at least 6 rrtonahs' duration. For childhood passive smoking exposure, we asked about the smoking habits (i.e., amount and years of smoking) of father, mother, or other household members AaaREV7ATON5 L'SED:'.AIK=adenoarcinoma: Cl -confidence intcn-a l; CSP=Univenity of Southern California/Los Angeles t:ounry, Cancer SurveillSnee Program; RR=relriive risk(s); SCC=squamous cell car- cinoma. 'Received June 11, 1984; revised November 28; 1984; accepted December 11, , 1984. YSupponed by grant S163 from the American Cancer Society. sDepartmem of Family and Prevenuve Medicine, Univerxity, of , Southern Cilifornia School of Medicine. Parkview Mediol Building B, 2025 Zonal Ave., I_os Angelh, CA 90033., •Prrsent add:esr Imperial Cancer Research Fund's Cancer Epide- mio]ogy Unit, Radcliffe InGimary;, Oxford Univenity„ Oxford OX2 6T•!E. Eng)and. 5We thank the word-processing pool for preparauon of the manu- script. JNC. VOL 71. NO..4. APRIL 1985

748 Wu, H.nderson, Pike, and Yu when they lived with the respondent during, her child- hood and teenage years. For passive smoke exposure during adult life, we asked about the smoking habits of spouse(s) i and'other household members when they lived with the respondent. Passive smoke exposure at work was assessed only in terms of the average number of hours per day to which the respondent believed she was exposed at each job. The questions on vitamin A intake specifically asked about average frequencies of consumption of 21 vegeta- bles and' fruits that are high in fil-carotene and 7 foods that contained preformed vitamin A during the calendar year 3 years before diagnosis of the case (19), Pattern of use of vitamin supplements was also assessed for the same period. On the basis of U.S. Department of Agriculture tables of food values for standard portion size (common household measure) of each item (20), we estimated average daily intake of P-carotenc (or, vitamin A) by summing the product of the 6-carotene (or vitamin A) content of each food item and its reported frequency of consumption. Quarailes of consumption were constructed on the basis of the intake pattern of the 220 controls. All cases were diagnosed microscopically. Their rou- tine pathology reports were reviewed for mention of lung scarring. Statistical analysis was conducted with the use of multivariate logistic regression methods for individually matched case-control studies (21). RR were estimated by odds ratios. A case-control pair was excluded from any given analysis if the information for either the case or the control was not known for the relevant variable(s). Since personal smoking will often; if not always, confoun& other associations, RR for other factors were always given after adjustment was made for personal smoking. For ADC, RR for certain factors were given separately for nonsmokers, ex-smokers, and' current smokers; this was not done for SCC because the numbers of non- smokers and ex-smokers were too few. RESULTS We interviewed 149 ADC and 71 SCC cases and their matched controls. The mean age at diagnosis was 59.7 years for ADC cases and' 61.4 years for SCC cases. The mean ages (at date of diagnosis of the index case) for the respective control groups were 59.5 and' 61.1 years. Personal cigarette smoking.-For both ADC and;SCC, there was a significant trend in risk associated with increasing number of cigarettes smoked per day and with decreasing age at which smoking began (table l1): Both aspects of smoking remained significant after adjustmentt was made for the other. Passive smoking.-FamiTies tended to share similar smoking behavior. Controls whose father, mother. or spouse(s) smoked were more likely,to smoke„to be heavy smokers, and to start at a younger age than controls whose family members did not smoke. For ADC and SCC, after adjustment was made for personal smoking habits, there were no significantly increased risks for having a mother, a father, or spouse(s) who smoked or for being exposed at work (table 2). For nonsmoking ADC cases, we did not observe anN elevated risk associated with passive smoke exposurt from either parents (RR=0.6; 95% C1=0:2, L7), fron- spouse(s) (RR=l.2; 95% C1=0.5, 3.3), or at worl' (RR=l.3; 95% CI=0.5, 3:3): Increasing RR (RR=1.0 1.2. 2.0) were found.-with increasing years (0, 1-30, ?31 of~passive,;3smokt, exposure . during adult • life ° frotr spouse(s) and at work, but the results were not sta tistically significant. Since the exposures may have occurred'concurrently, the years of exposure represen units rather than chronologic time of exposure. Childhood exposures.-For both ADC and SCC, n( significant association was found with history of luns, diseases (specifically;, asthma, bronchitis, pneumonia tuberculosis, fungal diseases, emphysema, and lun! abscess) diagnose& by a physician at least 5 years befor diagnosis of the case. When the analysis was restricted v lung diseases that occurred before age 16 (childhood), :, significantly elevated'. RR for pneumonia was observe for ADC after adjvstment was made for personal smokini, habits (RR=2.7; 95% CI=1.1, 6.7), and the RR for SC( (RR=2.9; 95% CI=0.5, 17.4)!was in the same directior Parental smoking did not explain this effect. Table shows that for ADC, the effecu of childhood pneumoni! was most apparent among nonsmokers: Of the 2 TABLE 1-Personal amoking,Aabite oJewex awd controla Smoking status ADC SCC RR 95% CI Caae/control RR 95% CI Caseicontrol Nonsmoker 1.0 29/62 1.0 2/30 Ex-smoker" 1.2 0.6. 2.3 21/37 7.7 0:8:70.3 8/18 ,Cwrrent smoker 4.1" 2.3.7.5 99150 35.3' 4.7.,267.3 61/23 Current smoker: No. cigarettes/day. 1-20 2.7 1.4. 5.4 38F 28 17.7 2.3; 138.2 19/ 9 a ?211 6:5' 3.1.13,9 61/22 94.4' 9:9:904.6 42,'9 Current smoker: age started to smoke. yr ?25 1.1 0.4. 3,2 8/14 7.8 0:8.717 6/5 19-24 2.5 1.0. 5,8'. 22/15 47.1 4.4. 498.5 18. 7 <_ 18 8.0" 3.6. 17.9. 69/21 115:7" 9.8. 1371.2 37/11 " Had stopped smoking atJeut 3 yr before diagnosis year of case. 'P (linear trend) <.001. I . ,NQL VOL 74:..NO. 4. APRIL 1985, 2023512726

TABLE 2.-Eaposxrr to passive smokiny in cases and controtr Lung Cancer In MVotrt*n 749 Smokin status ADC SCC g Adjusted RR' 95% CI Adjusted RR° 95% CIi Mother smoked 1.7 0.8, 3.5 02 0;0: 1,5 Father smoked 1.3 0.7,2.3 0.9 0:3,2.9 Spouse(s) smokedb 1.2 0.6.2.5 1.0 0,1„7.6 Exposure at the workplace 1.2 0.8.2.2 2.3 0:7„ 7.9 " Adjuated for number of cigarettes smoked per day and age at, starting to smoke. We eliminated from the analysis 15 pairs of ADC and 4 pairs of SCC in which either the use or the control'..caa never married: nonsmoking ADC cases, 8(2896) gave a history of childhood pneumonia. Elevated RR, adjusted for personal smoking habits, were observed for exposure to burning coal used for heating or cooking in a stove or fireplace during the majority of childhoo&and teenage years (ADC: RR=2.3; 95% C1=1.0, 5.5. SCC: RR=1.9; 95% CI=0.5, 6.5). For ADC, elevated RR were observed in each personal smoking habit category (table 3). TABLE 3.-RR and 95% confidence internala of ADC of the lunp aecrordiny to childhood pneumonia and eool burning by personal smoking Aabtita Exposure RR (95% CI) among: Nonsmoker Ex-smoker Current amoker Childhood pneumonia' No 1.0 1.4 (0.6, 2.4) 5.1 (2.5, 10.3) Yes 3,1(1.0; 9:9) 1.5 (0.2, 10.8) 10.9 (2.1, 57.9) Childhood coal burningb No 1.0 1.6 (0.6, 3.5) 6.3 (3.0, 13.3). Yes 3.2 (0.9, 11.8) 4.3 (1.0. 17,8) 9.5 (2,1, 41.9). ° Before age 16. The analysis was based on 149 case-control pairs of ADC 'Includes heating or cooking with coal burned in a stove or fireplace during childhood and teenage years. The analysis was based on 143 caae-control pairs of ADC. Dietary vitamin A.-Table 4 presents RR for ADC, adjusted for personal smoking habits, by quartiles of indices of vitamin A consumption. Because of the smaller sample size of SCC cases, the indices were dichotomized. For ADC, a significantly increased risk was observed only for those in the lowest quartile of' S-carotene consump- tion (<2,000 IU/day) compared to those in the highest quartile (>4,000 IU/day);, but no appreciably increased risks were observed for those in the intermediate groups. For SCC, an elevated, but not statistically significant, RR was observed for women with S-carotene intake below the median: When those in the lowest quartile of $- carotene consumption, i.e., Ikss than 2,000 lU/day, were compared to those consuming more than 2,0001IU/day, the unadjusted RR was increased to 1.7' (from 1.3); but after adjustment the RR was not greater than compari- sons above and below the median (both RR=1.5): There was no association with an index of total preformed vitamin A(i.e., dairy products, eggs, liver, and vitamin supplements) for either cell type. However, for ADC and SCC, an association was observed for dairy products and eggs (table 4). Other factors.-We could find no association between any occupation or occupational category and risk of ADC or SCC, but there was an excess number of cooks (4 cases and 2 controls) and beauticians (8 eases and 5 controls) among nses; both occupations have been suggested in previous studies. Elevated RR adjusted for personal TABLE 4..-Dutary intake of 0•carotens, tataW prefornud rtitamin A, and dairy products and epps amoeq caau and controls Quartile S-Carotene° Total preformed vitamin A6~` Dairy products and eggs` Adjurted RRd 96% CI Adjusted RR' 95% CI Adjusted RR" 95% C1 ADC 1 (high) 1.0 1.0' 1.0 2 0.8 0.8.2.0 0:6 0.3.1.4 1.7 0.813.9 3 1.3 0.6,2.7 1.1 0:5.2.5 22 1.0i4.8 4 2.5 1.1,5.7 1.2 0:5,2.8 2.7 1.2,6.8 ~ SCC 1 and 2 1.0 1.0 1.0 8and4 1.5 0.6,3.8 1.0 0A,2:4 1.6 0.7,8:9 'Includes 21 vegetables and fruits: leafy lett.uce, other leafy green, broccolu carrots„tomatoes„green peas, green beans, lima beans, asparagua, summer squash, winter squash, sweet potatoes and/or yams, green pepper, red pepper, hot red' chili pepper, cantaloupe. ~ watermelon, peachee„apricota, nectarines, and tomato and/or V8 juice. Analysil was based on 147 paiiY of ADC and 69 pairs of SCC:, "Ihcludes eggs, cheese, butter and/or margarine. cream, milk, beef andlor calfdiver, chicken andlor turkey liver, and'vitamin suppkments. ;Analyais was based on 147 pairs of' ADC and 71 pain of BCC. -~t ~ Adjusted for number of cigarettes smoked'per day. ,NQ. VOL 74. NO. 4, APRIL.19d5

750 Wu, H.nd.rson, Pik., and Yu smoking habits were observed for a history of hysterec- tomy (RR=1.7; 95% CI=0:9, 3.2) and nullipar~ity (RR=1.7; 95% CI=O!8, 3.7) among ADC cases and a history of miscarriage (RR=1.5; 95% CI=0.5, 4.9) among SCC cases. Multiple logistic regression analysis was condutted to assess the possible confounding effects of personal' smoking habits, childhood pneumonia, childhood coat burning,,and'S-carotene intake. The results were similar to those when each factor was adjusted for personal smoking habits alone. DISCUSSION This case-control study examined risk factors for the two main cell types of lung cancer in women-ADC and SCC. Although histologic typing was done by the individual pathologist at each participating, hospital, studies comparing interobserver and intraobserver varia- bility in classification of lung cell types reported a high concordance rate for cell types other than large cell carcinoma, which was excluded in this study (22, 23). In this study population, about half of ADCand almost all of SCC can be attributed to personal smoking habiu;, the amount smoked and the age at which smoking began were strong determinants of risk of disease. However, there are marked differences in the strength of associatiom between smoking and cell type of lung cancer, as has been noted previously (24, 23): The role of passive smoking in the etiology of ADC among nonsmokers is not clear. Our data are not consistent with the findings with regard to nonsmokers obtained by Hirayama (4)~ and Trichopoulos et at. (5) who reported a twofold to threefold increased risk due to passive smoking. H'owever, the histology of the cases in these studies is not clear, and their data suggest that any; effectof passive smoking is larger for SCC cases (3, 6): Of our 29 nonsmoking ADCcases, 12 were bronchoalveolar cell carcinomas, and this cell type is specifically, men- tioned by Correa et al. (6) to have a weaker, association wizh passive smoking. The effect of passive smoking by cell type of lung cancer needs to be investigated further in studies with much larger numbers of nonsmokers. Childhood lung disease may have a role in lung cancer etiology. Certain features of the lung of a child (e.g., susceptibility to airway closure and high peripheral resistance) might make it more vulnerable to residual abnormalities from respiratory illness (26). This notion is supported by observations that both smokers and nonsmokers with childhood respiratory diseases have impaired lung function capacity, that their rate of decline in ventilatory function capacity with age is more rapid than that in individuals without childhood respiratory problems, and that they have higher rates of clinical diagnosis of chronic obstructive pultnonary disease (27, 28). Women with childhood respiratory problems may have incurred epithelial damage to the airway resulting in airway hyperreactivity and are more susceptible to: other insults to the lung_ We cannot rule out the possibility of a chance finding or of preferential recall of JNCt. VOL 74. NO:.L APRIL 1985 childhood pneumonia by cases. However, our data appear to be internally consistent, since we found a significantly higher frequency of . lung scarring men- tione6 in the pathology reports among cases withh previous childhood pneumonia (12/30=40%) compared to those without (39/189=21%). The association of l'ung cancer risk with exposure to coal heating or cooking warrants further investigation. Although coal was identified as the major heating or cooking fuelI used during childhood and'teenage years of a significantly higher proportion of cases, we did not have detailed information on the years of use. Excess risks of lung cancer have been reported for coke oven workers (29, 30) and British gas workers (31), who were heavily exposed to products of coal carbonization. Studies of men suggest that their lung cancer risk is lowered by greater dietary 0-carotene (12-14, 32, 33) and vitamin A intake (15, 17; 32, 33), but the evidence for women is less clear (1'2, 13, 32, 33). We observed'a significantly increase&risk for ADC with the lowest level of S-carotene consumption and a similar association for SCC. These results are consistent with findings for females in Singapore (12) and in Japan (13), but they are not supportive of data for females in Hawaii (32) and England (33), Our observation of no association with ar. index of total' preformed vitamin A (i.e., dairy products eggs, liver, an&vitamin supplements) and no associatior with total vitamin A intake (preformed vitamin A ant S-carotene-data not shown due to domination b% preformed vitamin A) is consistent with findings fos females in Hawaii (32). Conflicting findings have beer reported for subgroups of preformed vitamin A foods anc supplements. A higher consumption of liver and vitamir supplements has been reported previously for femal<t cases as compared to controls, but the opposite result- have been observed for males (33, 34). Our data shoh no case-control difference in the intake pattern o vitamin supplements and a higher consumption of live among cases. Our, finding of an elevated lung cancer ris; associated with low levels of intake of dairy products ha not been reported for females, although similar resultt have been observed for males (15-17 ). Our results on th- role of 0-carotene and preformed vitamin A were simila for ADC and SCC, despite suggestions that vitamimA (0 fl-carotene) is more strongly protective against SCC tha; against ADC (17). InitiaLreports of an inverse relationship between bloa retinol levels and subsequent risk of cancer at all sites (3` 36) have not been supported~by recent studies (37, 38 This situation emphasizes the need to reexamine even th consistently observed association of vitamin A (or ~ carotene) intake with male lung cancer. Possible sources of bias in our data musc:be consi,derec Both lung cancer cases and controls were derived fror population-based samples. However, because this disea, is debilitating and rapidly fatal, 190 patients had died 4: were too ill to participate by the time of initial contac We did not conduct proxy interviews because questior on childhood exposures and dietarry history could not tlt assessed' adequately:, As expected, the group who was n 20235127128 Ellils.~..

Lung Cancr+r In Women 751 inten•iewed was tnute likely, to have metastatic disease at diagnosis but was similar in all demographic variables measured. In addition„ information abstracted from med'ical records showed'similar smoking status for those interviewed and those not interviewedl If cases who were not inteniewed because of poor survival differe& from those who survivedlonger and were interviewed in terms of'the other risk factors under study, this could have biased our results. However, this appears unlikely since our data showed that histories of childhood pneumonia and exposure to coal fires were similar among cases regardless of stage of disease at diagnosis. There is also no evidence that cancer survival is associated with dietary vitamin A intake. The etiology of SCC can be explained almost entirely by cigarette smoking. Cigarette smoking, however, ex- plains only about haU of the ADC cases. On the basis of this study, childhood lung disease and exposure to coal fires in childhood explain at least another 22% of ADC cases. Passive smoking and vitamin A may be involved, but more research, is needed to clarify t:heir roles in lung cancer etiology. REFERENCES (1) StLVERItERG E. Cancer statistics, 1982. CA 1982; 32r15-31. (2) HENDEASON BE. Descriptive epidemiology and geographic pathology. In: BurehenalL JH. Oettgen HF, eds. CGncer achievements, challenges and prospects for the 1980's. Vol! 1. New York: Grune, 1981151-69. (3) U.S. Public HealthServici. The health consequences of smoking: Cancer. Washington DC: U.S. Govt Print Off: 1982 (DHEW publication No. (PHS)82-50179]. (4): HrRArAMA T. Non.smoking wives of heavy smokers have a higher risk of lung cancer: A study from Japan. Br Med J 1981; 282:183-185. (1)iTRICHO-OULOS D, KALAfiDIDI A. SPARROS L. MACMAHONB: Lung cancer and passive smoking. Int J Cancer 1981; 27:IH. (6) CoRREA P,PICRt.E LW, FoN-rHAME, LIN Y, HAEwszEL W. Passive smoking and lung cancer. Lancet 1983;2:595-597. (7) Xe ZY, XtAO HP. Li G. Air pollution and lung cancer in Liaoning Province. Natl Cancer Inst Monogr. In press. (8) BREStiow L. HoAGwn L, RASMt'SSEN G, AaRAMS HK. Occupa- tions and cigarette smoking as risk factors in )ung cancer. Am J Public Health 1954; 44:171-181. (9), WyNDER EL, BERG JW. Cancer otthe lung among rson-smoken- special reference to histologic patterns. Cancer 1967; 90:1161-1172; IO0) MENCR HR. PtRE MC, HENDERSOH BE, Jtwe JS. Lung cancer risk I among beauticiarss and other femak workers:BneJcommunxa- tion. J Nad Cancer Inst 1977r59:1423-1425. (11)AUERIACH 0,GA'Rr7NREL LPARRs VR.Sar cancer of the lung- inaease over a 21 year period Cancer 1979;43:636-642., t12) MACLENNAN R.,DACosTA J, DAr NE.,LAw CH, Nc YK. SHAN- MUOARATNAM K. RlskfaQors for lung, cancer in Singapore Chinese, a population with high fetnalt incidence rates. Ini J Cancer 1977; 20:854-860. 1131 HtRArAMA T. Diet and cancer. NutncGncer 1979; 1:67-81. (141 SHERELLE RB. LEPPER M. Liu S, et al. Dietary vitamin A and risk of cancer in the Western Elecuic study. Lancrt 1981„ 2 l i185-1190: (15) B)ELAE E. Dteunyviumin Aand human lung cancer. ]ht JCAncer 1975; 15:561-565. (16), METTLINC, GRAHAMS,SN'ANSONW Vitamin A and lungcancer JM1Ch 1979; 62:1435-1438 (17) KVALE G. B)ELttE E. GART, JJ. Dieury habits and lung cancer risk. lnt,J Cancer 1983; 31:397,405. (181 MAt'.tt TM- Cancer Surveillance Program in Los Angeles Counn Nail Cancer Inst Monogr 1977; 47:99-101. (19) GRAHAM S. MErrLrN C. Fiber and other constituentrof vegeublts in cancer epidemiology In: Newell GR, Elltson, NM', eds Nutrition and oncereuolog) and treatment. New York: Raven Press. 1981:189-215. (20) ADAMS C. Nutritive value of American foods in common units U.S. Department oflAgriculture Handbook No. 456. Washtng- ton, DC U.S. Department of Agriculture, 1975.. (21) BRESLOw NE. DAV NE. Statistical method's imcancer research Vol I-The analysis of case-control'studies. IARC Sct Pubi 1980, 32:5-338: (22) YESt:ER, R. GERST B. At'ER6ACH 0. Apphotion of the World Health Organization classification of lung carcinoma to biops) material. AnnThorac Surg 1965: 1:33-49. (23)YESNER R. CARTERD.Pathology'.Of carcinoma of the lung. ClYn Chesr Med 1982; :3:257-289: (21) DoLL R. HILL AB, KREteERC L The significance of cell type in relatiom to the aetiology of lung cancer. Br J Cancer 1957; 11:43-48. (25) WvnnER EL. CoVEy LS, MAStteHrK. Lung cancer in women: Present and futurrtrends. J NattCaneer Inst 1973;,51:391-401. (26) KATTAN M. Long-term sequelae of rrspiratory illness in infancy and childhood. Ped'utr Clin North Am 1979; 26:525-535. (27) Bt::RROws B. KNUOSON RJ, LEUowrrz MD. The relationships of'. childhood respiratory illness to adult obstructive airway disease. Am Rev Respir Dis 1977; 115151-759: (28)I SAMET JM„ TAGER IB.SPEIZER FE. The relJrtionship between respiratory illness in childhood'andchronic airflow obstruction in adulthood. Am Rev Respir Dis 1983; 127:508-523. (29) REDMOND CK, CIOCGO A. LLOYD JW, RUSH HW. Long-term mortality study of steel workers. VI. Mortality from malignant neoplasms among coke oven worken., J Occup Med' 1972; 14:621-629. (30) LLOrD JW. Long-term mortality study of steelworkers. V. Respiratory cancer in coke plannwork'en. J Occup Med 1971t 13:53-68. (31) DOLL R. FtSHER EJ.,GAMMON W. etal. Mortality of gasworkers with special reference to cancers of the lung and bladder, chronic bronchitis. and pneumoconiosis. Br J Ind Med 1965- 22t1-12. (32) Hwnx MW. Kot.OrvEL LN. HANRtr: JH: Dieurq viumin A. orotene, vitamin C and risk of lung cancer in Hawaii. Am J Epidemiol 1984't 119i227-237. (33) GREGOR A. LEE PN, RoE FJ, WttsoN Mj, MELTON A. Compari- son of dietary histories in lung cancer cases and controls with special reference to vitamin A. Nutr Cancer 1980;2:93-97. (34) SMrrH PG, JreR H. Cancers among users of preparations contain- ing vitamin A. Cancer 1978; 42:808-811. (35) KARR JD. SMtTH AH, SwrrzER BR, HAMES CG. Serum vitamin A (retinol) and cancer incidence in Evans County. Georgia. JNCI 11981; 66:7-16. (36) WALD N, IDLE M, BOREHAMJiLowserum vitamin A and subse-quent risk of cancer. Lancrr 1980; 2:813-815. (37) WtLLETT WC, Pout BF, Ui+DERM•ooD BA, et at. Relation of serum vitamin A and E and'orotenoids to the risk of cancer. NEng1iJ Med 1984;310!430-434. (38) STAHELIK HE, BUESs E, ROSEL F, WIDMER LK, BRARACHERBVitamim A, cardiovascular risk fanors., and mortaLty. Lancet -9fi2; 1:394,395. ! JNCI. VOL 74~. NO 4;.APRIL 1985