Tobacco Documents Online

AMERICAt: JOURKAL or EPIDEMIOLOGYVol. 125. No. 1. CopynitM C 1987, bg The John. Hopkins Univeniity School of Hy`iene and Public HultA Pnnud in U.S.A. AII'.rishts rnerved RISK FACTORS FOR ADENOCARCINOMA OF T8E LUNG ROSS C. BROV1"tiSO4 " JOHN S. REIF,' THOMAS J. ItEEFE,' STANLEY W. FERGUSON' AND JANE A. PRITZLx drownson, R. C. (Dept. of Microbiology and Envkonnl.ntal' H.alfh, Colorado Stat4 U'., Fort Collins, CO 00523), J. f3. RaH, T. J. Keefa, S. W. F.rpuaon, and J.A Pftf. Risk factors for adenocarcinoma of fhe kaip. Am J Epidsnllol 1fM7;125:25- 34. The relation between various risk factors and adenocancinoma of tAe lung was wdwtad In a case-control study. Subj!.ats wen selected from tM Colorado Central Cancer Rpistry from 1979-1982 In ehe Denver metropolitan anu. A total of 102 (50 males and 52 females) adenocarcinoma case Interviews and 131165 males and 66 f.males) control int.crriews were completad. The oontrol' group consisted of persons witb eanc.n of the colon and bone marrow. The risk estinsates assoclatad with cigarette smoking were significantly elevated aniatp males (odds ratio (OR) : 4.49) and females (OR s 3.95) and were found to inerease sipnif'Kantly (p c 0.01) with incnasin9 levels of ciganttft smoking for both males and f.nlsles. For adenocarcinoma fn, fentales. the ape- and smoking- adjusted odds ratios at differ.nt levels of passive sntoke exposure followed an irlueasing overall trend (p = 0.05). After add3tiond adjustmerM for potanWl' eontow+ders, prior cFpantte use remained the most significant prsdictac of risk of adenoearcinoma among males and females. Analysis restricted to nonsmoking females revealed a risk of ad.nocareeinoma of 1.1511 (95X confidenCe Mt't.rvd (Cl) s 0.39-2.97) for passive smoke exposure of four or more hours per day. N.Mher sex showed siqnifieanthr elevated risk for occupational exposures, aMtltouptt ntatas bordered on significance (OR s 2.23, 95x Cl a 0.97-5.12). The results suggest tM need to develop cell tlrpe-spec#fic effoloqk:Arpatltesesc air pollutlon; lung n.oplasms; tobacco smoke poputlon Recent evidence indicates that lung can- w~elF=esiabllshed: but the relatio>ai ~: een sdenoc"a~clnoma aad `clgirettie'}tlmo cer may encompass several morphologically and clinically distinct dieeasea (1, 2). In .~u~~ clea~'(3, 5; 6)° ildustrialized western nations, incidence -Adenocarcmome is tbe moat frec~tlently rates are highest for squamous cell carci'- diagnosed form of lung cancer in the United Qomk_followed by adenocarcinoma (3, 4). States among women and nonsmokers (3, ig~tm~ 'iCfrl i>id~ ?). In a series of marly 30,000 cases of primary lung cancer, 22 per cent were spec- Recxiwd for publication March 28.1986. HeahColorado State Microbiology and Environmental CO. ' Colondo Department of Health, Denver, CO. ' Reprint requests to Dr. Rots C. Brownson at Ntrent addreu: Cancer Epidemiolo=y and Control PloR+Im. Division of Environmental Health and Ep- of Health, P:~O ooBox p1 ~t~Miswuri 65205. This muyFac liy offaCold S~uu U~~i yt't~tl PWnial fulfillment of the requiremenu for the degree of Doctor of Philosophy for Ron C. Brownson. This work was supported in part by National Insti- tutes of Health Biomedical Research Support Grant 2 507 RR-05t56-20 and a grant from the American Lunt A.ssociation of Colorado. The authors acknowledge the aristance of the Col- orado Central Cancer Registry, Colorado Department of Health and the staff of the Department of Micro- biology and Environmental Health, Colorado State University. They also thank Dn. Richard F. Hamman and Mowafak D. Salman for their helpful comments. 25 NOTI'CE This materiai may be OroteetEd by' CopY'Ighr law (Title 17 U.S. Codea, `.A

26 BROR'NSON' ET AL. ified as adenocarcinoma among males com- pared with 37 per cent among females (8). The role of occupational exposures in the etiology of adenocarcinoma remains incon- clusive (9, 10). Recently„a disproportionate increase in the incidence of adenocarci- noma has been noted in the United States (5). The changing histologic patterns of lung cancer incidence may be due to a change in diagnostic practices and classifi- cation or to increasing exposure to environ- mental carcinogens. The present investigation was designed to evaluate the role of smoking, passive smoking, occupation, community air pol- lution, and socioeconomic status in the etiology of adenocarcinoma of the lung. A case-control study was conducted to pro- vide additional data concerning the relation between exposure variables and this infre- quently studied and poorly understood form of lung cancer. MATERIALS AND METHODS Cases and controls were identified through the population-based Colorado Central Cancaer Registry maintained by the Colorado Department of Health. For the years and counties included, reporting was essentially complete. All diagnoses were microscopically confirmed and classified according to histologic type. Study partici- pants were required to have resided in the Denver metropolitan area for at least six months prior to cancer diagnosis in order to reduce migration bias. Case selection A total of 149 eligible cases of adenocar- cinoma (International Classification of Dis- eases (IiGD) code 163) were identified in the five-county Denver metropolitan area from 1979-1982. Selection was restricted to white males and white females. These ad- enocarcinoma cases were stratified by age and sex. Of the 149 eligible cases, 31 could not be located, 15 refused to be interviewed, and one did not qualify. A total of 102 case interviews (50 males and 52 females) were completed. The mean ages for male and female cases were 64.9 and 663 years, re- spectively. Control selection Controls were chosen from persons in the Colorado Central Cancer Registry who had cancer of sites considered to be unrelated to cigarette smoking. Specifically, persons with cancers of the colon (ICD code 153) and bone marrow (ICD code 169) diagnosed from 1979-1982 were chosen as controls and group-matched to adenocarcinoma cases according to age and sex. Matching was done at the group level so that the maximum number of cases and controls could be used in the analyses. Only whites were included in the study, and at least one control was required for each case within each age an& sex stratum. A total of 169 eligible controls were iden- tified. Of these, 24 could not be located, 13 refused to be interviewed, and one did not qualify. A total of 131 usable interviews (65 males and 66 females) were completed. Among controls, 80 were colon cancer pa- tients, and 51 were diagnosed with leuke- mia. The mean ages for male and female controls were 65.2 and 68.2 years, respec- tively. Data collection and analyses Epidemiologic data were collected by per- sonal interview. The interviewer was un- aware of whether the patient was a case or a control. A higher percentage of the inter- views in the case group (68.6 per cent) than in the control group (38.9 per cent) were completed by a relative or a friend. Among the 70 nonsurviving cases, 56 interviews were completed with a spouse, seven inter- views with a child, six with a sibling, and one with a close friend. For the 51 deceased controls, information was obtained from 42 spousea, six childten, two siblings, and one close friend. Socioeconomic status was assessed by ex- amining two variables, education and in- come. Educational' level was characterized by the highest grade of formal' education completed. Gross income was ascertained rJ ~ ~ CJ ~ ~ Cj 0' W

RISK FACTORS FOR ADENOCARCINOMA OF THE LUNG 27 for the previous year, or in ease of retirees, for the year prior to retirement. Smoking history was characterized for cigarettes, cigars, or pipefuls in terms of pack-years of exposure. Passive smoking data were analyzed as a dichotomous vari- able based on the smoking status of the patient's spouse and as a stratified variable based on the hours per day that the subject was in the presence of persons who were smoking. Occupational data were analyzed accord- ing to industrial category, occupational cat- egory,, and a self'-assessment of the expo- sure of the respondent to known lung carcinogens in the workplace. Those indus- tries and occupations known to be associ- ated with an elevated risk for lung cancer were coded and multiplied by the number of years in each category to estimate e:po- sure over time (11-13). In addition, each subject was shown a list of 12 groups of materials known to be lung carcinogens and was asked whether he or she had been exposed to the substances during a partic- ular occupation. Pulmonary carcinogens included materials such as asbestos, chro- mium, nickel, uranium ore, and mustard gas. Positive responses were coded as inte- gers and summed. An index of exposure to community air pollution was developed based' on estimated levels of total suspended particulates per census tract and the years of residence in each census tract (14). Total suspended particulate air pollution, which contains a benzene soluble fractiony was used as an indicator of polycyclic hydrocarbon (e.g., benzoIaJPyi'ene) levels. The total sus- pended particulate data were stratified into 10 air pollution exposure subgroups, and each census tract within the Denver area was assigned to a subgroup. The residence code consisted of years at each residence multiplied by the corresponding total sus- pended particulate exposure subgroup. In the first set of analyses, stratified con- tingency tables were constructed to adjust for age and smoking for the primary risk factors (15-17). Odds ratios for each level of exposure were calculated by Miettinen's standardized rate technique which controls for confounding factors (18). All analyses included adjustment for age based on the categories 30-49, 50-59, 60-69, 70-79, and 80-99 years. An extension of the Mantel- Haenszel procedure was used to statisti- cally evaluate overall trends in the propor- tion of cases according to level of exposure to risk factors (19, 20). Multiple logistic regression, was used to obtain maximum likelihood point and in- terval estimates of the odds ratio, as well as to control for the effects of various con- founding risk factors (21-23). The most significant predictors, based on the Mantel- Haenszel results, were included in the lo- gistic model. The dependent variable in these analyses was lung adenocarcinoma (case (coded as 1) or control (coded as 0)). Independent variables were entered in in- tervals, as recommended by Schlesselman (24). In order to identify the potential con- founding effect of the induction period of cancer, the exposure of each case or control to ambient air pollutants and industrial carcinogens was analyzed in two ways: 1) the entire residence and work history of each person was included; and' 2) only es- posures that took place 10 or more years prior to the time of diagnosis were consid- ered, The analyses were completed both for all subjects and for primary respondents only, to assess the validity of the surrogate interview data. A multiple logistic regres- sion model was also constructed for non- smoking female cases and controls. R.asvs.'rs **--;~-r~ -..' ;i ., ~.. F r 1~; ;7rr r„Si~*rt1aC`~s.cd~Y~1a; ~ f~,r;,.~ tableLl). The age-adjusted odds ratio for prior ciga- rette use among males was 4.49 (95 per cent confidence interval (CI) - 1.44-13,98). Among females, the risk due to cigarette smoking was 3.95 (95 per cent CI - 1.7C- 8.80). For adenocarcinoma in females, the age- and smoking-adjusted odds ratios at N N W w N W N ~

28 BROW NSON' ET AL. TABLE I Adjusted odds ratios (OR) and trend tests for odenocarcinoma of the lung according to level oj cigarette use and passive smoke e:posun, metropolitan DenGYr. CO, 1979-1982 Males Fetn.ies Factor No. of ases No. of contro6s OR' No. of cases No. of aontrots OR Prior ciprette use (pack-years) 0 4 19 1.00 19 47 1.00. 1-39 14 19 4.06 10 13 1.68 t40 32 27 7.68 23 6 14.80 Trend (p valueY, (<0.01) (<0.01) Passive smoke exposure (hours/dry) 0-3 16 28 1.00 29 53 1.00 4-7 19 24 1.76 11 8 3.06 Z8 15 13 2.68 12 5 2.33 Trend (p value) (0.46) (0.05) ' Odds retio for prior ci8arette use adjusted for age; odds ratio for passive smoke exposure adjusted for si{e and smoking. different levels of passive smoke exposure followed an overall trend, statistically sig- nificant at the 0.05 level. The age- and smoking-adjusted odds ratio for passive smoke exposure (using 0-3 hours per day as the referencalevel) was 1.01 (95 per cent CI - 0.42-2.41) among males. The corre- sponding risk for females was 2.42 (95 per cent CI - 0.94-6.22). Odds ratios for pas- sive smoke exposure were also calculated on a yes/no basis for the regular smoking history of the patient's spouse. The aden- ocarcinoma risk from smoking by the spouse was not significant for males (odds ratio (OR) = 1.40, 95: per cent CI = 0.66- 2.14) or females (OR = 1.54, 95 per cent CI - 0.72-2.35). The odds ratios and their 95 per cent confidence intervals for education level, in- come, community air pollution exposure history, and occupational exposures are presented in table 2. The lowest level of each variable was used as the reference category. Both education and income showed inverse trends with adenocarci- noma risk. Ameng males, annual income approached statistical significance with an odds ratio of 0.47 (95 per cent CI - 0.19- 1.19). No significant risks in the age- and smoking-adjusted odds ratios were shown for males or females according to their air pollution exposure history. No difference was noted regardless of whether the entire residence history of the patient or only the residence history 10 or more years prior to cancer diagnosis was used in the analysis. Of the occupational variables (ind'ustriai category, occupational category, or self-re- ported exposure to lung carcinogens), only occupational exposures for males bordered on significance (OR - 2.23, 95 per cent CI = 0.97-5.12). The multiple logistic regression risk es- tirnates for income, occupation, pack-years of cigarette use, and passive smoke expo- sure are shown in table 3.. For both sexes combined, annual income showed an in- i verse association with adenocarcinoma risk after adjustment for other risk factors (OR ~ = 0.85, 95 per cent CI - 0.72-0.98). A positive association between pack-years of ' cigarette use and cancer risk was found for ! males, females, and both sexes combined- 'I'he largest risk for adenocarcinoma asso- ciated with passive smoking was shown for females at the exposure level of 4-7 hours per day (OR - 1.91, 95 per cent CI = 0:78- 3.03). The fust-order interaction of pack- years of smoking and passive smoking was examined and found to be nonsignificant. ~,. N C N W - W Cl.~N C..W N I! <=:=.~

RISK FACTORS FOR ADENOCARCINOMA OF THE LUNG TAal.e 2 Adjusted odds ratios (OR) and 95% confidence intervals (Cl) for adenocarcinoma of the liuig according to education, ineome, air pollution rrsidence history, and occupation, metropolitan Denuer. C0, 1979-1982 29 •laks Females Factor n OR` 95% C1 n OR 95n°r C1 Education level (highest grade) 0-8 25 1.00 17 1.00 9-17 90 0.59 0.23-1.54 101 0.73 0.23-2.31 Annual income (tlioosands of dollan)t ct15.000 25 1.00 37 1.00 L315,000 86 0.47, 0.19-1.19 78 0.71 0.28-1.85 Residence history (tsposure-years)3 0-99 26 1.00 31 ' 1.00 t100 89' 1.66 0.66-4.19 87 1.51 0.58-3.96 Occupation (ezposure-years)g 0 76 1.00 112 1.00 i1 39 2.23 0.97-5.12 6 0.59 0.09-3.51 ' Odds ratio adjusted for age and smoking. , • Misning values. I The product of years at each residence sad the corresponding total surpended particulatc exposure svlgrouP. { CkwPations at high risk for lung cancer multiylied by the number of years in each category. TASLE 3 A(ultiple logistir regression odds ratios (0R) and 95% confidence intervals (CI) for odenocorcinoma of the lung according to income, occuyotion, cigarette use„and passive smoke ez,oosure, metropditan Denuer, C0. 1979-1982 Fann AIl sub)eets Males Femaks r n OR• 95% CI n OR 95% Cl n OR 95% Cl laeome 233 0.85 0.72-0.98 115 0.85 0.66-1.03 118 0.84 0.64-1.03 Oceupation 233 1.00 0.96-1.04 115 1.00 0.97-1.04 118 0.94 0.51-1.37 Pack-yean 0 89 1.00 23 1.00 66 1.00 1-39 156 2.62 1.82-3.41 33 3.74 2.37-5.12 23 1.93 0:88-2.99 t40 88 5.81 5.01-6.61 59 b.42 4.13-6.71 29 9.58 8.31-10.86 Passive amokint (6otrs/eLy) 0-3 126 1.00 44 1.00 82 1.00 4-7 62 1.24 0.b3-1.9b 43 0.84 0.00-1.80 19 1.91 0.78-3.03 t8 45 1.37 0.b4-2.20 28 1.17 0.10-2.24 17 1.21 0.00-2.68 ' Odds tatio adjuated for qe,.pooential confounding fictoa, and aez when eqpropriste. Logistic regression was conducted by us- ing only Primary respondents. These re- allts were similar to those found when all respondents were included. Active smoking was the only risk factor significant at the 0,051eve1 based on the analysis of primary respondents. The odds ratios for pack-years of smoking were consistently smaller for Primary respondents, whereas those for Passive smoke exposure were larger when Primary respondents were analyzed. The risk of adenocarcinoma due to pas- sive smoke exposure was examined among female nonsmokers (table 4). Nineteen fe- male nonsmoking cases were identified (36.5 per cent). Due to size limitations, passive smoking was divided into two cat- egories: 0-3 and four or more hours per day. An odds ratio of 1.68 (95 per cent CI - 0.39-2.97) was computed for the larger ex- posure category after adjustment for age, income, and' occupation.

30 BROWtiSON' ET AL TABLE 4 Multipie logistic regression odds ratios (QR) 'and 95% confidhrce interralt (Cll /or adenocarcinoma of the G+ng according to income. occupation, and paasiue smokt exposure among female nonsmokers. metropolitan, Deuvr. CO. 1979-1982 Factor n oR• 95% Cl fnco,ne 66 0.95 o.6o-iai V~11~il1Vil o...:......,,.ti:.a tbours/d+rr a-3 w 56 V.w+ 1.00 x4 io 1.68 0.39-2.97 • Odds ratio adjusced for age and pountial con- less frequently among females than among males (10, 30-32 ). The current study found significant risk estimates for adenocarcinoma associated with smoking of 4.49 for males and 3.95 for females. The age-standardized risk esti- mates at different levels of cigarette use showed significant trends ( p< 0.01) for males and females, indicating that _ - - , AamvcatCwoma was vtesent~`The risk esti- mates based on multiple logistic regression analyses for smoking were generally lower founding futorsm than the odds ratios calculated by the DtscussioN Numerous case-control studies of lung cancer have been conducted over the past 30 years. Few, however, have examined the data according to histologic type. There appears to be a general consensus that the various histologic types of lung cancer have a multifactorial etiology which includes cig- arette smoking and occupational and other environmental factors. ~5mo M ,t ss Ill ist ci tss~.i;-,f ti L' l ikl. 0 9 0 d!lnce e , .~:........ t25). Several reports have suggested that smoking may not be the major risk factor for adenocarcinoma in certain populations (26-28). Among white males, the age- standardized relative risk estimates for lung adenocarcinoma according to prior cigarette use have ranged from less than one at low levels of smoking to about six at high leveb of smoking (3, 29). Risk esti- mates of adenocarcinoma from smoking for females are commonly lower and vary widely among racial groups; for example, the relative risk estimates range from about one in Chinese women to four in Japanese women, and five in Hawaiian women (26, 30). The risk of smoking and adenocarci- noma for white females is usually between one and three, although the risk of lung cancer by histologic type has been studied methods of Mantel and Haenszel (15) and Miettinen ('17), since logistic regression al- lowed for adjustment for multiple factors. The risk estimates for smoking and ade- nocarcinoma found in this study and the presence of a dose-response relation were consistent with other studies (29, 31, 33). The effect of involuntary inhalation of sidestream smoke (passive smoking) on lung cancer etiology is a controversial cur- rent public health issue (34). Hirayema (35J reported a significant relative risk for lung cancer of 2.08 among wives of heavy smok- ers. A study conducted among Greek women found relative risks of 2.4 and 3.4 for wives of light and heavy smokers, re- spectively (36). A case-control study in Louisiana identified an increased risk for lung cancer among nonsmokers married to heavy smokers and for subjects whose mothers smoked (37).. Garfinkel et al. (38) found an increased lung cancer risk for women whose husbands smoked 20 or more cigarettes per day. A recent study in Los Angeles found a slight increase in risk of adenocarcinoma among nonsmoking women exposed to passive smoke (39). Sev- eral other studies have failed to link passive smoke exposure to an increased risk of lung cancer (40-42). Prior studies that have evaluated passive smoking and lung cancer have differed in the indez of passive smoke exposure, cell type, and degree of histologic verification (34). In the present study,indeses of passive smoke exposure were obtained in two ways: N O N ci Ca N W N Ca ~smog aii ~~ 1

RISK' FACTORS FOR ADENOCARCINOMA OF THE LUNG 31 O 1) by ascertaining the regular smoking his- tory of the spouse of each subject on a yes/ no basis; and 2) by determining the average hours per day that the subject was exposed to smoking (at bome and at work). No significant risk estimates were shown when smoking by the spouse was considered as a dichotomous variable. When the data were stratified according to level of passive smoke exposure, a statistically significant trend in the risk estimates was shaam for females (p = 0.05) after adjustment for age and cigarette smoking. However, after ad- justment by logistic regression for age, in- come, occupation, and cigarette smoking, no significant adenocarcinoma risk for pas- sive smoke exposure was found among fe- males. The relatively large proportion of non- smoking female cases 436.5 per cent) observed in this study suggested the impor- tance of other risk factors in adenocarci- noma etiology. A previous study found 19.5 per cent nonsmokers among female ade- nocarcinoma cases (39):.Our study demon- strated a slightly elevated risk among fe- male nonsmokers due to passive smoke exposure, consistent with the findings of Wu et al. (39). Deficiencies in passive smoking data in recent studies include: 1) no commonly established index of aide- stream smoke exposure; 2) a lack of data on other indoor air pollutants such as ra- don; 3) the existence of a probable differ- ential in accuracy of obtaining passive smoke exposure histories between living and deceased subjects; 4) a lack of evidence of changes in the peripheral bronchial epi- thelium of nonsmokers exposed to.side- stream smoke (40); and 5) insufficient numbers of nonsmoking lung cancer cases available for analyses. Despite these limi'- tati'ons, the relation between passive smok- ing and lung cancer deserves further inves- tigation. Although pollutants in the air have long been suspected to contribute to the etiology of lung cancer, epidemiologic evaluation has been hampered ~ by difficulties in defin- ing and measuring air pollution and in eval- uating the effects of confounding variables such as smoking, occupation, and popula- tion mobility (43). A census tract analysis of lung cancer data, total suspended partic- ulate air pollution, and median household income was reported previously for the Denver area (14). Our previous work showed a significant direct relation be- tween male lung cancer rates and total sus- pended particulate air pollution (p < 0.02). However, for both males and females, me- dian household income explained a larger percentage of the variation in lung cancer rates than did particulate air pollution. The data on residence history of cases and controls were analyzed to determine if differences in total suspended particulate air pollution exposure may have accounted for a portion of the adenocarcinoma inci- dence. There were only slight differences between cases and controls in mean or me- dian years of residence in metropolitan Denver. Residence history was defined in terms of exposure-years (years of exposure to high or low total suspended particulates) in order to define an inde: of exposure for each case and control. Although, in Denver, cases commonly experienced more expo- sure-years, no significant differences be- tween cases and controls were detected for males or females. Our data failed to show the presence of a large air pollution effect. Occupational exposures may be impor- tant risk factors for lung cancer (44-51). Prior studies of lung cancer have demon- strated an increased risk for exposure to substances sucb as asbestos, arsenic, nickel, radon daughters, diagnostic radiation, and fossil fuel combustion products (44). Incon- sistent f ndings have been reported regard- ing the importance of occupational factors in adenocarcinoma incidence (9,10). In this study, occupational risks for adenocarci- noma were examined in two ways: 1) an a priori listing of industries and occupations in which workers are at high risk for lung cancer was used to code the work history data from each case or control; and 2) each subject was asked if he or she was ever esposed' to a list of known lung carcinogens

4ir 32 BROWNSON ET AL in the workplace. The exposures (indus, trial, occupational, or pulmonary carcino- gens) were cumulated over the lifetime of the subject, and the analysis was based on a classification of any or no previous ex- posure. Only high-risk occupational history showed a borderline significant risk for ad- enocarcinoma among males after adjust- ment for age and smoking history. The occupational risk was smaller after multiple adjustment for age, income, cigarette smok- ing, and passive smoking. The relations between workplace exposures and adeno- carcinoma risk were unchanged regardless of whether the entire work history of the subject or only the work history 10 or more years prior to diagnosis was used. A difference in risk for lung cancer by social class has been observed whether measured primarily by occupation, income, or education (3). Part of the socioeconomic differential in lung cancer risk is due to smoking habits (52). In this study, educa- tion level and gross income were used as socioeeonomic indicators. Income level showed a stronger association with adeno- carcinoma risk after controlling for age and smoking than did education. Since colon cancer is correlated with socioeconomic status (53), it is possible that the use of colon cancer patients as controls in this study magnified the observed inverse rela- tion between adenocarcinoma and income level. No statistically significant inverse association was noted in adenocarcinoma risk with respect to education level, al- though risk estimates were commonly lower at higher educational levels. The issue of dietary vitamin A and lung cancer risk was not addressed in this study. Evidence is accumulating that a deficiency in dietary vitamin A may result in a higher risk for lung cancer and that a higher intake of vitamin A and its provitamins has an apparent protective effect (28, 54-59). Diet may be less important in our study since recent data have suggested that the inverse relation between vitamin A intake and lung cancer is strong for squamous cell and small cell carcinomas but not for adenocarcinoma (29, 58). This study used a higher proportion of surrogate interviews for cases (68.6 per cent) than of surrogate interviews for con- trols (38.9 per cent). Several investigators have attempted' to characterize the validity of information obtained from surrogate in- terviews (60-62). Pickle et al. (60) found that siblings were best able to describe events that occurred early in life, whereas spouses and offspring best recounted events during adult life. Other studies have found that bias may be introduced because of inaccurate work histories given by next of kin (61) and that spouses may provide ac- curate demographic information and a crude estimate of smoking, but details of employment history and diet may be of lower validity (62). To address this prob- lem, we conducted separate analyses for all respondents and for primary respondents. The results were highly comparable and indicated that some conclusions based on all respondents may have been conservative since adenocarcinoma risk estimates for passive smoking were commonly higher among primary respondents. In light of the changing histopathologic patterns of lung cancer, the findings of this and other recent studies suggest the need to consider the various lung cancer cell types as different diseases. Future research should emphasize accurate histologic typ- ing and the development of cell type- specific etiologic hypotheses. Rereae„crs 1. Mulvihill~JJ. Host factors in human lune turmn: an example of ecopnetics in oncology. JNCI 1976:57:3-7: 2. Sclr.rry OS. Hanun HH. Respiratory tract can- cer. In: Holland JF, Frei E, eds. Cancer medicine. Philadelphia: Leak Fehiger, 1982:1709-44. 3. Fraumeni JF Jr, Blot WJ. Lung and pleura. In: Schottenfeld D. Fraumeni JF Jr. eds. Cancer epi- demiolo=y and prevention. Philadelphia: WB Saunden.,1982:564-82. 4. Krcyberg 1:. Histological ~huu eancer type.: ataor- pholbpcal and biological correlation. Acta Pathol Microbiol Scand Suppl 1962;152:1-92: 5. Vincent RG. Pickren JW, Lane WW, et aL The

RISK FACTORS FOR ADENOCARCINOMA OF THE LUNG 33 changing histopathology of lung cancer. A review of 1682 cases: Cancer 197 7;39:164;-55: 6. Shimizu H. Prestan-Martin S, Cas.grande JT, et al. Epidemiologic eharacteristics of adenoearci- noma of the lung in Los Angeles County. Natl Cancer Inst Monogr 1982:62:161-4. 7: American Cancer Society. Cancer facts and' Sg- ures.1985. New York, 1984. 8. Young Y L Jr, Pollack ES. The incidence of cancer in the United' Statrs. In: Schottanfeld D, Frau- meni JF Jr. eds. Cancer epidemiology.rtd preven- tion. Philadelphia: WB Siunders.1982:138-65. 9. Vincent TN. Satterfield JV. Ackerman LV: Car- cinoma of the lung in women. Cancer 1965;18:559- 70. 10. Stayner LT„ Wegman DH. Smoking, occupation and histopathology of lungtancer: a case-control study with the use of the Third National Cancer Survey. JNCI 1983:70:421-6. 11. Executive Office of the President. Office of Man- agement and Budget. Standard industrial classi- fication manual. Washington DC: US GPO,1972. 12. US Department of Commerce. Office of Federal Statistieal Policy and Ssandards. Standard occu- pational classification manual. Washington DC: US GPO. 1980. 13. Lerchen ML. Personal communication.1984. 14. Brownson RC, Reif JS; Keefe TJ, et aL The relationship between air pollution, income, and hmg cancer in Denver, 1985 (unpublished manu- acriptl: 15. Mantel N. Haensul W. Statistical aspects of the analysis of data from retrospective studies of dis- ease. JNCI 1959:22:719-48. 16. Gan J. The comparison of proportions: a review of signification. Rev Int Stat Inst 1971;39:148-69. 17. Miettinen OS. Estimability and estimation in ase-referent studies. Am J Epidemiol 1976;103:226-35: 1& Miettinen OS. Standardization of risk ratios. Am J Epidemiol 1972:96:363-8. 19. Rothman KJ, BoicrJD Jr. Epidemiologic analysis with a programmable calculator. US GPO, 1979. (US DHEW publication no. (NIH):79-1649):. 20. Mantel N: Chi-square tests with one degree of f»edom: extension of the M.ntel-H.ensul pro- adure. J Am Stat Assoc 1963;59:690-700: 21. Armitage P. Stati.tical methods in medical re- sean:li. Oxford: BlackweU Scientific Pttbliutiona,. 1971. 22. Cos DR. Analysis of binary data. Lomdon: Me- tFiuen. 1970. 23- Engelman L. Stepwi.e logistie regression. In: Dixon WJ, ad BMDP statistical software. Bake- ley, CA: University of California Prar,1981:33a- 44. 24. Schktsalaun JJ. Case-controt studies: design, eonduct, analysis. New York: Oxford University rew mmon 25. Ha~d EC, Seidman H. Smoking and cancer in the United States. Prev Med 1980•9:169-73. 26. Hinds MW, Summermaan GN, Yang HY. at a1. Differences in lung cancer risk from smoking among Japanese. Chinese and Hawaiian women in Hawaii. lnt J' Cancer 1981;27:297-302. 27. Chan WC, Colbourne M'J• Fung SC, et al. Bron- chid cancer in Hong Kong. 19 76-1977. Br J Can- cer 1979;39:182-92: , 28. MacLennarn R, DaCosta J, Day NE, et d. Risk factors for lung cancer in Singapore Chinese. a population with high female incidence rates. Int J Cancer 1977;20:854-60: 29. Byers T, Vena J, Mettlin C, at a1. Dietary vitamin A and lung cancer risk: an analysis by histologic subtypes. Am J' Epidemiol 1984;120:769-76. 30. Auerbach O, Garfinkel L, Parks VR. Histologic type of lung cancer in relation to smoking habits• year of diagnosis and sites of inetastasis: Chest 19 75:67:382- 7. 31. Weiss W. Boucot KR. Seidman H, et aL Risk of lung cancer according to histologic type and cig- arette dosage. JAMA 1972:222:799-801. 32: Davies JM. Lung cancer mortality of workers making chrome pigments. Lancet 1978;1:384.. 33. Hammond EC, Horn D. Smoking and deeth rates-report on forty-four months of follow-up of 187.783 men. 11. Death rates by cause. JAMA 1958:166:1294 -1308: 34. Wynder EL. Goodman MT. Smoking and lung cancer. some unresolved issues. Epidemiol Rev 19815:17 7-207. 35. Hirayama T. Non-smoking wives of heavy smok. ers have a hietier, risk of lung cancer- a study from Japan. Br Med J 1981:282:183-5. 36. Trichopoulos D. Kalandidi A, Sparros L, et al. Lung cancer and passive smoking. lnt J Cancer 1981,2 7:1-40~ 37. Correa P, Pickle LW, Fontham E: et al. Passive smoking and lung cancer. Lancet 1983:2595-7: 38. GarGnkel L, Auerbach 0, Joubert L. Involuntary smoking and lung cancer. a case-control study: JNCI 1985;75:463-9. 39. Wu AH, Henderson BE. Pike MC„at al. Smoking and other risk factors for lung cancer in wromen, JNCI 1985;74:747-51. 40. Garfinkel I. Time trends in lung cancer mortality among nonsmokers and a note on passive smok- ing. JNCI 1981;66:1061-6. 41. Chan W C, Colbourne MJ, Fung SC, et al. Bron- chial cancer in Hong K'ong,1976-77: Br J Cancer 1979;39:182-92. 42. Koo LC, Ho JH-C, Saw D. Active and' pasaive smoking among female lung cancer p.tienu and controls in Hong Kong. J Etp Clin Cancer Res 1993:4:367-75. 43. Naional Research Council, Committee on Medi- cal and Biologic Effects of Environmental Pollu- tants. Airborne particles. Baltimore: University Park Prai,1979.- 44. Fraumeni JF Jr. Rapi>•.tory carcinogenasis: an epidemiologic appraisal. JNCI 1975;55a039-46: 45. Bridbord K. Decou/le P. Fraumeni JF Jr. at al. Estimates of the fraction of cancer in the United States related to occupational factors. Bethe.d., MD: National Cancer Instituu. National Institute of Environmental Health Sciences, and National Institute for Oecupationd Safety and Health, September 15, 1978. 46. Annual report of the chief inspector of factories and workshops for the year, 1932: London: HMSO,

t 34 BROWNSON'ET AL. 1933:103-4. 47. Kuroda S, Kawahata K. Uber die gewerblickie Enutehung des Lungkrebses bei Generatorgasar- beitern. Z Krebsforsch 1936:45:36-9. 48. Hill AB, Faning„ EL. Studies in the incidence of cancer in a factory handling inorganic compounds of arsenic. 1. Mor.ality experience in the factory. Br J Ind Med 1948:5:1-16. 49. Machk W, Gregorius F. Cancer of the respiratory system in the United States chromate-producing industry. Public Health Rep 1948;63:1114-27. 50. Mrrewether ERA. Asbestos and carcinoma of the lung. In: Annual report of the chief inspector of factories for the year 1947: London: HMSO, 1949:79-81. 51. Decoufle P. Occupation. In: Schottenfeld D. Frau- meni JF Jr, eda. Cancer epidemiology and preven- tion. Philadelphia: WB Saunders, 1982:318-35. 52. Wynder EL, Stellman SD. Comparative epide- mioiogy of tobacco-related cancers. Cancer Res 19-o7:37:4608-22. 53. Graham S, Levin M, LiHcnfeld AM. The socioec- onomic distribution of cancer of various sites in Buffslo, Nl': 1948-1952. Cancer 1960:13:18f-91. 54. Bjelke E. Dietary vitamin A and human lung cancer. Int J Cancer 1975:15:561-5. 55. Gregor A. Lee PN, Roe JFC, etal. Comparison of dietary histories in lung cancer cases and controls with special, roference to vitamin A. Nutr Cancer 1980:2:93-7. 56. Shekelle RB, Lepper M. Liu S, et al. Dietary vitamin A and risk of cancer in the Western Electric Study. Lancet 1981:2:1185-9. 57. Mettlin C, Graham S, Swanson M. Vitamin A and lung cancer. JNCI 1979:62:1435-8. 58. Kvale G, Bjelke E, Gart JJ. Dietary habits an& lung cancer ruk. lnt J' Cancer 1983:31:39 "4 -4A5. 59. Hitayama T. Diet and cancer. Nutr Cancer 1979:1:67-81. 60. Pickle LW, Brown LM, Blot WJ. Information available from surrogate respondents in case-con- trol intervie.+ studie.. Am J Epidemiol 1961118:99-108. 61. Greenberg ER, Itoaner B, Hennekens C, et al. An investigation of bias in a study of nuclear shipyard workers. Am J Epidemiol 1985;121:301-8. 62. Lsrchen MI., Samet JM. An assessment of the validity of questionnaire responses provided by a surviving spouse. Am J Epidemiol 1986:123: 481-9.