Document byzwgZXqjO0yo88BZn0543Zk0

Volume 143 Number 6 March 15, 1996 American Journal of EPIDEMIOLOGY Copyright 1996 by The Johns Hopkins University School of Hygiene and Public Health Sponsored by the Society for Epidemiologic Research ORIGINAL CONTRIBUTIONS Family History of Cancer and Risk of Lung Cancer among Lifetime Nonsmoking Women in the United States Downloaded from aje.oxfordjournals.org by guest on October 19, 2010 Anna H. Wu,1 Elizabeth T. H. Fontham,2 Peggy Reynolds,3 Raymond S. Greenberg,4 Patricia Buffler,5 Jonathan Lift,6 Peggy Boyd,7 and Pelayo Correa2 In a multicenter study of lung cancer in lifetime nonsmokers in the United States, 646 female lung cancer patients and 1,252 population controls were interviewed regarding history of cancer in their first-degree relatives. A 30% increased risk (95% confidence interval 0.9-1.8) was found for a history of respiratory tract cancer in parents or siblings after adjustment for exposure to environmental tobacco smoke (ETS) in adult life. Lung cancer, which represented approximately two thirds of the respiratory tract cancers, occurred more frequently in first-degree relatives of lung cancer patients than in comparable relatives of population controls (ETS-adjusted odds ratio = 1.29, 95% confidence interval 0.9-1.9). In particular, a significant threefold increased risk for lung cancer was associated with lung cancer diagnosed in mothers and sisters. The increased risk in relation to family history of lung cancer was observed among parents and siblings who were smokers as well as in those who were nonsmokers. The association with family history of lung cancer was strengthened when the analysis was restricted to adenocarcinoma of the lung (ETS-adjusted odds ratio = 1.50, 95% confidence interval 1.0-2.2). However, there was no association between family history of other cancers and risk of lung cancer in nonsmokers. Am J Epidemiol 1996;143:535-42. family health; lung neoplasms; women Although tobacco smoke is a well-established causal factor for lung cancer, there is accumulating evidence that genetic factors also play a role in lung cancer development. Case-control studies have consis tently found that lung cancer aggregates in family members of lung cancer patients (1-9). A few studies have also observed that all cancers (8, 10), other nonpulmonary smoking-related cancers (2, 9, 10), and reproductive cancers (11) occur significantly more often in first-degree relatives of lung cancer cases than in controls. The extent to which familial susceptibility to the development of lung cancer can be attributed to aggregation of smoking habits in families versus shared genetic susceptibility has been difficult to de termine, because most lung cancer patients in these studies were smokers, and information on smoking by family members was generally absent. In a seminal study conducted by Tokuhata and Lilienfeld (1) in which information on smoking and history of cancer was obtained from both the index subjects and their family members, the familial effect was more marked in nonsmokers than in smokers with lung cancer, and Received for publication March 16, 1995, and in final form De cember 20, 1995. Abbreviations: Cl, confidence interval; ETS, environmental to bacco smoke; OR, odds ratio. 1 School of Medicine, University of Southern California, Los An geles, CA. 2 Department of Pathology, Louisiana State University Medical Center, New Orleans, LA. 3 California Department of Health Services, Emeryville, CA. 4 Medical University of South Carolina, Charleston, SC. 5 School of Public Health, University of California, Berkeley, CA. 6 School of Public Health, Emory University, Atlanta, GA. 7 California Public Health Foundation, Berkeley, CA. Reprint requests to Dr. Anna H. Wu, University of Southern California School of Medicine, Norris Comprehensive Cancer Cen ter, 1441 Eastlake Avenue, MS 44, P.O. Box 33800, Los Angeles, CA 90033-0800. 535 536 Wu et al. Downloaded from aje.oxfordjournals.org by guest on October 19, 2010 the excess of lung cancer was observed in relatives irrespective of their smoking history. However, in two studies with small numbers of lung cancer cases in nonsmokers (less than 35 cases in each study), an increased familial risk was observed only among smokers with lung cancer, not among nonsmokers with lung cancer (8, 9). Using data collected in a recently completed case-control study designed to in vestigate the etiology of lung cancer in lifetime female nonsmokers (12), we compared histories of cancer in first-degree relatives of lung cancer patients and con trols. MATERIALS AND METHODS Study methods have been described in detail else where (12, 13). In brief, this was a population-based case-control study of lung cancer in female lifetime nonsmokers, defined as those who had smoked fewer than 100 cigarettes and had not used any other form of tobacco for more than 6 months. Women aged 20-79 years with a microscopically confirmed diagnosis of primary lung cancer who resided in one of five major metropolitan areas in the United States (Atlanta, Geor gia; Houston, Texas; New Orleans, Louisiana; the San Francisco Bay Area; and Los Angeles County, Cali fornia) comprised the case series. In Atlanta and Hous ton, eligible lung cancer cases were diagnosed from December 1, 1985, through November 30, 1988, whereas in New Orleans and the two California study centers, case accrual continued for 2 additional years. Controls were selected by random digit dialing. The control group was supplemented by random sampling from Health Care Financing Administration files for women aged 65 years or more. Controls were frequen cy-matched to lung cancer cases (2:1 ratio) on age, ethnicity, and study area and were lifetime nonusers of any tobacco products. We completed interviews with 83 percent of eligible lung cancer cases and 70 percent of population controls. During the first 3 years of the study, a second control group was selected from fe male colon cancer patients, mainly to evaluate poten tial recall or response bias associated with being diag nosed with cancer or being ill. In the first report on this study, results on environmental tobacco smoke (ETS) were similar when lung cancer cases were compared with colon cancer or population controls (13). Inter viewing of colon cancer controls was not continued in the last 2 years of the study. Lifetime smoking status was determined in a mul tistep procedure that included reviewing information on tobacco use from hospital charts and contacting the physicians of lung cancer cases and colon cancer con trols. For subjects believed to be nonsmokers or for whom smoking status could not be determined from these sources, we contacted the subjects by telephone and asked specific questions on their lifetime use of tobacco products. The identical telephone screening questions were used for the population control group. At the time of interview, lung cancer cases and con trols were again asked the tobacco use screening ques tions to confirm each subject's reported status of life time non-tobacco use. In addition, a urine sample was collected from all consenting subjects at the end of the interview to determine the level of urinary cotinine and creatinine, as a measure of their recent nonsmok ing status (14). The histologic type of lung cancer was independently confirmed by a pulmonary pathologist (D. S. Greenberg, Department of Pathology, Baylor College of Medicine, Houston, Texas). In the family history section of the interview, all subjects, self- and surrogate respondents, were first asked the number of brothers and sisters (including half-siblings) they had. Subjects were then asked whether there was a history of cancer in the immediate family, which included fathers, mothers, brothers, and sisters (full and half-siblings). We did not distinguish between full and half-siblings in our coding if there was a history of cancer. The specific type of cancer in the affected relative and the relationship of the af fected relative to the index subject was determined. However, the report of cancer in relatives was not independently confirmed, and the relative's age at cancer diagnosis was not obtained. Although we did not specifically ask for the smoking history of affected relatives, in our assessment of lifetime ETS exposure we asked whether the mother, father, or any other household member(s) had smoked in the household of the index subject during her childhood. We classified lung cancer in parents according to whether the mother or father had smoked during the childhood of the index subject. We also classified lung cancer in siblings based on whether there were other smokers in the household (other than parents) during the index sub ject's childhood. We investigated the association between risk of lung cancer and family history of all cancers combined, as well as specifically by cancer site or system, which included the lung, breast, prostate, female genital tract (excluding the cervix), respiratory tract, and digestive system. In addition, family history of all tobaccorelated tumors (buccal cavity, larynx, esophagus, stomach, liver, pancreas, respiratory tract, cervix, bladder, kidney, and renal pelvis) was evaluated. Mul tivariate logistic regression methods were used to compute risk estimates for family history of cancer (15) by comparing lung cancer cases (self- and surro gate respondents combined) with population controls. In tables 1-4, two sets of odds ratios are presented. Am J Epidemiol Vol. 143, No. 6, 1996 Family Cancer History and Lung Cancer Risk in Nonsmokers 537 The odds ratios in the first column (adjusted odds ratios) were adjusted for age (^49, 50-59, 60-69, and >70 years), study area (Los Angeles County, the San Francisco Bay Area, and the southern study ar eas--New Orleans, Atlanta, and Houston), ethnicity (white, black, Asian, or Hispanic), education (<11, 12, and >13 years of schooling), and number of broth ers and sisters. (Details regarding the distribution of cases and controls by these demographic parameters have been previously presented (12).) The second col umn of odds ratios (ETS-adjusted odds ratios) was also adjusted for an index of exposure to ETS (0,1-11, 12-28, 29-47, and >48 smoke-years) during adult life (12). Potential confounders, including employ ment in potentially high-risk occupations (12), dietary intake of antioxidants, and dietary cholesterol, were considered. These factors did not confound the family history-cancer findings and were not included in the final models. RESULTS In table 1, data are presented on reported history of lung cancer among fathers, mothers, brothers, and sisters of lung cancer cases and population controls. The prevalence of any family history of lung cancer was 8.9 percent among lung cancer patients and 6.6 percent among controls (ETS-adjusted odds ratio (OR) = 1.29, 95 percent confidence interval (Cl) 0.88-1.90). Mothers and sisters of lung cancer cases were approximately three times more likely to have a history of lung cancer than comparable relatives of controls; however, there was little difference in such a history among fathers and brothers of lung cancer patients and controls. The increased risk of lung can cer associated with a family history of lung cancer was present after adjustment for ETS exposure; this adjust ment did not uniformly strengthen or weaken the risk estimates for lung cancer in families. The odds ratio associated with lung cancer in mothers was affected very little by adjustment for ETS exposure (adjusted OR = 2.74, ETS-adjusted OR = 2.81), whereas the odds ratio associated with lung cancer in sisters was reduced but remained statistically significant with ad justment for ETS exposure (adjusted OR = 3.58, ETS-adjusted OR = 2.78) (table 1). Few subjects reported a history of lung cancer in more than one family member, and there was no indication of in creasing risk with increasing number of family mem bers with a positive history. Analyses restricted to self-respondents produced risk estimates that were slightly stronger than those calculated for all respon dents combined (data not shown). Downloaded from aje.oxfordjoumals.org by guest on October 19, 2010 TABLE 1. Family history of lung cancer and risk of lung cancer in lifetime nonsmokers, United States, 1985-1990 No. Of population controls No. of lung cancer cases Self- All Adjusted* respondents respondents ORt All lung cancers 95% Clf ETStadjustedf OR 95% Cl Any lung cancer No Yes 1,158 366 570 1.00 1.00 82 39 56 1.33 0.9-1.9 1.29 0.9-1.9 Family member with history of lung cancer Father 30 13 16 1.04 0.6-2.0 0.94 0.5-1.8 Mother 6 6 9 2.74 0.9-8.0 2.81 0.9-8.3 Brother 43 16 21 0.86 0.5-1.5 0.89 0.5-1.6 Sister 8 8 14 3.58 1.5-8.8 2.78 1.0-7.4 No. of family members with lung cancer 1 74 34 49 1.34 0.9-2.0 1.31 0.9-2.0 2 8 5 7 1.29 0.5-3.7 1.14 0.4--3.6 * Adjusted for age (<49, 50-59, 60-69, and >70 years); area (Los Angeles County, San Francisco Bay Area, and southern centers (Houston, Louisiana, and Atlanta combined)); ethnicity (white, black, Asian, or Hispanic); education (11, 12, or >13 years of schooling); and number of brothers and sisters. Thirty-eight subjects were excluded from all analyses because of missing information on education (n = 25) or number of brothers and sisters (n= 13). t OR, odds ratio; Cl, confidence interval; ETS, environmental tobacco smoke. i Adjusted for age, area, ethnicity, education, number of brothers and sisters, and exposure to environmental tobacco smoke in adult life (0, 1-11, 12-28, 29-47, and >48 smoke-years). Am J Epidemiol Vol. 143, No. 6, 1996 538 Wu et al. To determine whether familial risk of lung cancer is associated with the active smoking habits of subjects' relatives, history of lung cancer among the parents and siblings of subjects was stratified by their respective smoking status (table 2). These results are presented for self-respondents only, because of likely incomplete knowledge by proxy respondents of the smoking hab its of parents and siblings of index subjects (12). Risk of lung cancer was elevated among subjects whose parents had a history of lung cancer, irrespective of whether the parents were smokers or nonsmokers. Similarly, history of lung cancer among siblings was associated with an increased risk of lung cancer re gardless of whether siblings were smokers or non smokers. There was a 30 percent (95 percent Cl 1.0-1.8) increased risk in relation to family history of any respiratory cancer (table 3). However, this was ex plained largely by the excess risk associated with lung cancers, which accounted for 68 and 66 percent of the respiratory cancers among lung cancer cases and pop ulation controls, respectively. There was no associa tion between risk of lung cancer and family history of tobacco-related cancers combined (table 3). Respira tory cancers represented 62 percent of the tobaccorelated cancers among lung cancer cases, compared with 47 percent among population controls. However, there was a deficit of digestive tract cancers among first-degree relatives of lung cancer cases in compar ison with population controls. Mothers and sisters of lung cancer cases and population controls showed similar histories of cancer of the breast and of the female genital tract (excluding the cervix). History of prostate cancer did not differ between fathers and brothers of cases in comparison with comparable rel atives of controls (table 3). Lung cancer cases and population controls did not differ with regard to family history of all cancers combined (excluding lung can cer) (adjusted OR = 1.00, 95 percent Cl 0.8-1.2); there was no significantly increased risk in relation to a history of any cancer reported in three or more family members. Results were essentially the same when the analyses were restricted to self-respondents (data not shown) or adjusted for years of ETS expo sure (table 3). Approximately three fourths of the lung cancers in this study population were classified as adenocarci noma (12). Results for any family history of lung cancer were statistically significant (ETS-adjusted OR = 1.50, 95 percent Cl 1.0-2.2) when the analyses were restricted to adenocarcinoma of the lung (table 4). The odds ratios associated with lung cancer in mothers and sisters were further strengthened in anal yses restricted to adenocarcinoma of the lung. Al though the odds ratio for a family history of female genital tract cancers and risk of lung cancer was also greater, this result was not statistically significant. Lung tumors other than adenocarcinomas were few in number; analysis in this subgroup produced unstable risk estimates for family cancer (data not shown). We also examined the effect of family cancer by proband's age at diagnosis of lung cancer (or age at interview for controls). Despite few lung cancers diagnosed among subjects aged 54 years or younger, the risk estimates in Downloaded from aje.oxfordjournals.org by guest on October 19, 2010 TABLE 2. Risk of lung cancer in lifetime nonsmokers, by parents' and siblings' history of lung cancer and smoking habits (self-respondents only), United States, 1985-1990 Lung cancer in relative Parent No No Yes Yes Relative(s) that smoked No Yes No Yes No. of population controls No. of lung cancer cases (self respondents only) Adjusted* ORf 512 175 1.00 687 209 0.86 4 2 1.52 32 15 1.23 95% Clf 0.7-1.1 0.3-8.6 0.6-2.4 ETSfadjusted OR* 1.00 0.82 1.44 1.05 95% Cl 0.6-1.1 0.3-8.3 0.5-2.1 Sibling No No Yes Yes No 863 269 1.00 1.00 Yes 242 88 1.15 0.9-1.5 1.10 0.8-1.5 No 37 17 1.48 0.8-2.7 1.17 0.6-2.3 Yes 13 7 1.39 0.5-3.7 1.49 0.5-4.0 * Adjusted for age (^49, 50-59, 60-69, and >70 years); area (Los Angeles County, San Francisco Bay Area, and southern centers (Houston, Louisiana, and Atlanta combined)); ethnicity (white, black, Asian, or Hispanic); education (sll, 12, or >13 years of schooling); and number of brothers and sisters, t OR, odds ratio; Cl, confidence interval; ETS, environmental tobacco smoke. t Adjusted for age, area, ethnicity, education, number of brothers and sisters, and exposure to environmental tobacco smoke in adult life (0, 1-11, 12-28, 29-47, and >48 smoke-years). Am J Epidemiol Vol. 143, No. 6, 1996 Downloaded from aje.oxfordjournals.org by guest on October 19, 2010 Family Cancer History and Lung Cancer Risk in Nonsmokers 539 TABLE 3. Risk of lung cancer in lifetime nonsmokers, by family history of cancer, United States, 1985-1990 Type of cancer in family No. Of population controls No. of lung cancer cases Self- All respondents respondents All lung cancers Adjusted** ORt * 95% Clf ETStadjustedt OR 95% Cl Respiratory tract No Yes 1,115 349 546 1.00 1.00 125 56 80 1.30 1.0-1.8 1.30 0.9-1.8 Tobacco-related cancers II No 973 314 485 1.00 1.00 Yes 267 91 141 1.06 0.8-1.3 1.10 0.9-1.4 Digestive tractD No Yes 1,055 356 550 1.00 1.00 185 49 76 0.79 0.6-1.1 0.80 0.6-1.1 Breast cancer No Yes 1,133 362 567 1.00 1.00 107 42 58 1.10 0.8-1.6 1.09 0.8-1.6 Female genital tract (excluding cervix) No Yes 1,216 393 608 1.00 1.00 24 11 17 1.33 0.7-2.5 1.42 0.7-2.7 Prostate cancer No Yes 1,180 388 599 1.00 1.00 54 14 23 0.96 0.6-1.6 1.06 0.6-1.8 All cancers combined, excluding lung cancer No 707 231 362 1.00 1.00 Yes 533 174 264 1.00 0.8-1.2 0.99 0.8-1.2 No. of family members with cancer, excluding lung cancer 1 365 119 181 1.01 0.8-1.3 1.00 0.8-1.2 2 122 36 56 0.90 0.6-1.3 0.90 0.6-1.3 3 46 19 27 1.16 0.7-2.0 1.14 0.7-2.0 * Adjusted for age (s49, 50-59, 60-69, and 70 years); area (Los Angeles County, San Francisco Bay Area, and southern centers (Houston, Louisiana, and Atlanta combined)); ethnicity (white, black, Asian, or Hispanic); education (si 1,12, or 213 years of schooling); and number of brothers and sisters. t OR, odds ratio; Cl, confidence interval; ETS, environmental tobacco smoke. i Adjusted for age, area, ethnicity, education, number of brothers and sisters, and exposure to environmental tobacco smoke in adult life (0, 1-11, 12-28, 29-47, and 48 smoke-years). Includes sites listed under International Classification of Diseases for Oncology (29) codes 160-165. II Includes cancers of the buccal cavity and larynx, esophagus, stomach, liver, pancreas, respiratory system, cervix, bladder, kidney, and renal pelvis. H Includes sites listed under International Classification of Diseases for Oncology (29) codes 153, 154, 156, 158, 159. relation to family history of lung cancer and respira tory tract cancer were stronger in the younger age group (s54 years) than in the older age group (^55 years). For example, the ETS-adjusted odds ratio for family history of lung cancer was 1.86 (95 percent Cl 0.51-6.83) in younger subjects (<54 years) and 1.26 (95 percent Cl 0.83-1.90) in older subjects (>55 years). However, a deficit of digestive tract cancers was only apparent in the older age group (ETS-adjusted OR = 0.79, 95 percent Cl 0.57-1.09) and was not observed among younger subjects (ETS-adjusted OR = 1.00, 95 percent Cl 0.36-2.81). There was, however, no statistically significant age interaction with family history of lung cancer, respiratory tract cancer, or digestive tract cancer (data not shown). DISCUSSION This study, to our knowledge, is one of the largest studies to date of family history of cancer in relation to lung cancer among lifetime nonsmokers. The popula tion-based study design reduced the likelihood of se- Am J Epidemiol Vol. 143, No. 6, 1996 540 Wu et al. TABLE 4. Risk of adenocarcinoma of the lung in lifetime nonsmokers, by family history of cancer, United States, 1985-1990 Type of cancer Any lung cancer Any lung cancer in family member Father Mother Brother Sister No. of population controls (n = 1,242) Yes No 82 1,158 No. of lung adenocarcinoma cases (n = 479) Yes No 48 428 Adjusted* ORt ETSf- 95% cit adjustedt OR 95% Cl 1.52 1.0-2.2 1.50 1.0-2.2 30 1,189 6 1,228 43 947 8 974 15 443 1.25 0.7-2.4 1.11 0.6-2.2 8 454 3.20 1.1-9.6 3.24 1.1-9.9 18 378 0.95 0.5-1.7 0.95 0.5-1.7 11 368 4.06 1.6-10.4 3.59 1.3-9.8 No. of family members with lung cancer i 2 74 1,158 8 1,158 41 428 1.50 1.0-2.3 1.49 1.0-2.3 7 428 1.72 0.6-4.9 1.52 0.5-4.8 Respiratory tract 125 1,115 64 412 1.37 1.0-1.9 1.38 1.0-2.0 Tobacco-related cancers II 267 973 110 366 1.08 0.8-1.4 1.15 0.9-1.5 Digestive tractU 185 1,055 58 418 0.79 0.6-1.1 0.80 0.6-1.1 Breast cancer 107 1,133 47 428 1.17 0.8-1.7 1.17 0.8-1.7 Female genital tract (excluding cervix) 24 1,216 14 461 1.50 0.8-3.0 1.59 0.8-3.2 Prostate cancer 54 1,180 19 454 1.01 0.6-1.8 1.13 0.7-2.0 All cancers combined, excluding lung cancer 533 707 207 269 1.04 0.8-1.3 1.02 0.8-1.3 * Adjusted for age (<49, 50-59, 60-69, and >70 years); area (Los Angeles County, San Francisco Bay Area, and southern centers (Houston, Louisiana, and Atlanta combined)); ethnicity (white, black, Asian, or Hispanic); education (si 1, 12, or si3 years of schooling); and number of brothers and sisters. t OR, odds ratio; Cl, confidence interval; ETS, environmental tobacco smoke. i Adjusted for age, area, ethnicity, education, number of brothers and sisters, and exposure to environmental tobacco smoke in adult life (0, 1-11, 12-28, 29-47, and >48 smoke-years). Includes sites listed under International Classification of Diseases for Oncology (29) codes 160-165. II Includes cancers of the buccal cavity and larynx, esophagus, stomach, liver, pancreas, respiratory system, cervix, bladder, kidney, and renal pelvis. 11 Includes sites listed under International Classification of Diseases for Oncology (29) codes 153, 154, 156, 158, 159. Downloaded from aje.oxfordjournals.org by guest on October 19, 2010 lection bias, particularly inasmuch as participation rates were high (12). All women with incident lung cancer in the five geographic areas were eligible for participation, and controls were randomly selected from the same geographic areas as the cases. Our main results show a positive association be tween risk of lung cancer and history of reported lung cancer in first-degree relatives, particularly sisters and mothers. Although the mean number of full and half siblings was slightly greater in cases (2.05 and 2.25 for brothers and sisters, respectively) than in controls (1.98 and 2.15 for brothers and sisters, respectively), all analyses were adjusted for numbers of brothers and sisters. Interpretation of this finding is limited by po tential sources of bias. In particular, recall bias, perti nent to all case-control studies of familial lung cancer based on self-report, must be considered. However, the relative specificity of the association with lung cancer in a family member argues against a general bias of recall of all forms of cancer among lung cancer cases. In addition, first-degree relatives of lung cancer patients showed an excess of respiratory tract cancers and a deficit of digestive tract cancers compared with first-degree relatives of population controls and colon cancer patients (colon cancer patients were inter viewed during the first 3 years of the study and served as a second control group) (data not shown). The prevalence of family history of cancer was slightly greater among self-respondents than among surrogate respondents in this study. The lower esti mate of family history of cancer among surrogate respondents may be due to their unfamiliarity with the family history of probands, although we cannot be certain of this. Because all controls and 61 percent of lung cancer patients were self-respondents, our esti mation of risks using all lung cancer respondents com bined may represent a conservative estimate of the association. In addition, the prevalence of lung cancer Am J Epidemiol Vol. 143, No. 6, 1996 Family Cancer History and Lung Cancer Risk in Nonsmokers 541 Downloaded from aje.oxfordjournals.org by guest on October 19, 2010 in first-degree relatives of probands in this study was somewhat lower than that reported in previous studies. The greater prevalences of lung cancer in first-degree relatives (10-20 percent) reported in previous inter view studies were probably due to the inclusion of smokers (6, 20) and males (20) with lung cancer in those studies. The present study was restricted to fe males who were lifetime nonsmokers. Although a number of previous studies have shown a familial effect of lung cancer in smokers with lung cancer (2, 8-10), the data regarding such an associa tion among nonsmokers with lung cancer are less consistent (1-3, 8-10). This study agrees with studies conducted in Baltimore, Maryland (1), and Connecti cut (10), and suggests that there is a familial effect that is independent of the effect of tobacco smoke. The increased risk of lung cancer in relation to lung cancer in parents and siblings was observed irrespective of the smoking status of the relatives. In a study con ducted in Louisiana with information on the smoking habits of relatives (3), the magnitude of risk associated with lung cancer in female relatives was comparable in case-control comparisons conducted separately for nonsmoker and smoker female relatives. The present findings also suggest that having a female relative (mother or sister) with lung cancer increased the risk of lung cancer more than having a male relative (father or brother) with lung cancer, which is consistent with results from two previous studies (3, 9). Reasons for the higher risks associated with lung cancer in female relatives are not known. It may be that female subjects are more knowledgeable and accurate regarding family histories, particularly those of their mothers and sisters. However, the stud ies conducted by Ooi et al. (3) and Shaw et al. (9) included male and female lung cancer probands, so sex bias in recall is probably not the sole explanation. The greater strength of association among female rel atives might be explained by a greater proportion of lung cancer among males being linked to environmen tal causes (i.e., smoking and occupational exposures) than to genetic predisposition. There are no known biologic reasons for an increased genetic predisposi tion to lung cancer in women as compared with men. Some studies (4, 16) have suggested that female hor mones (i.e., an excess of estrogens) may result in an increased susceptibility to adenocarcinoma of the lung, although other studies (6, 7) do not concur with that suggestion. In this study of lung cancer among female lifetime nonsmokers, only family history of lung cancer was associated with an increased risk, whereas there were no statistically significant associations with family history of all cancers combined or family history of other cancers. The present results are not entirely surprising if the aggregation of relatives with cancer (2, 5, 17-19), particularly smoking-related cancers (5, 9), denotes an aggregation of tobacco use in families or a genotypic susceptibility promoted by tobacco smoke. It is intriguing that there was a deficit of digestive tract cancers among first-degree relatives of lung cancer patients, although this deficit was ob served only among subjects aged 55 years or older. A deficit of lung cancer among family members of he reditary nonpolyposis colorectal cancer patients has been reported in most previous studies (21, 22). To date, studies of lung cancer have disagreed regarding the prevalence of microsatellite instability in lung tu mors (23-25). Because a very high percentage of hereditary nonpolyposis colorectal cancer patients show microsatellite instability (26), it would be of interest in future studies to evaluate whether the pres ence of microsatellite instability in lung tumors differs between lung cancer patients with and without a fam ily history of digestive tract cancers. The findings regarding any family history of lung cancer were strengthened when the analyses were re stricted to adenocarcinoma of the lung. A family his tory of lung cancer was associated with a fourfold increased risk of adenocarcinoma of the lung in a case-control study of this specific cell type in women (6). Among other studies that included all lung cancer cell types, a stronger familial effect for adenocarci noma of the lung in comparison with other cell types was reported in one article (18). However, in another study, a stronger effect was found for small cell car cinoma (20) compared with other cell types, and in a third study, the association was present for all histo logic types of lung cancer (9). Finally, results of this study suggest that a family history of lung cancer tends to be stronger among subjects with early onset of lung cancer. These data are compatible with the segregation analyses conducted by Sellers et al. (27, 28), who found evidence for a major gene influencing earlier age at onset. This analysis represents a first step in demonstrating that lung cancer in nonsmokers tends to run in families more than would be expected by chance. However, at present, we cannot determine what portion of the fa milial clustering is related to common environmental exposures or biologically inherited susceptibility. To further evaluate the contribution of environmental ver sus genetic factors to the familial clustering of lung cancer in nonsmokers, additional information on all family members (i.e., affected and unaffected), includ ing their ages, specific smoking habits, age at lung cancer diagnosis (if relevant), and other lifestyle fac tors (e.g., dietary habits), will be needed. In addition, Am J Epidemiol Vol. 143, No. 6, 1996 Downloaded from aje.oxfordjournals.org by guest on October 19, 2010 542 Wu et al. segregation analysis will be necessary to determine whether the pattern of lung cancer among relatives is compatible with a single major gene, multiple genes, or simply a shared environment. ACKNOWLEDGMENTS This research was supported by grants CA 40095 and CA 57498 from the National Cancer Institute, with additional support from the American Lung Association and the Lou isiana Cancer and Lung Trust Fund Board. REFERENCES 1. Tokuhata GK, Lilienfeld AM. Familial aggregation of lung cancer in humans. J Natl Cancer Inst 1963;30:289-312. 2. Samet JM, Humble CG, Pathak DR. Personal and family history of respiratory disease and lung cancer risk. Am Rev Respir Dis 1986;134:466-70. 3. Ooi WL, Elston RC, Chen VW, et al. Increased familial risk of lung cancer. J Natl Cancer Inst 1986;76:217-22. 4. Gao YT, Blot WJ, Zheng W, et al. Lung cancer among Chinese women. Int J Cancer 1987;40:604-9. 5. Sellers TA, Ooi WL, Elston RC, et al. Increased familial risk for non-lung cancer among relatives of lung cancer patients. Am J Epidemiol 1987;126:237-46. 6. Wu AH, Yu MC, Thomas DC, et al. Personal and family history of lung disease as risk factors for adenocarcinoma of the lung. Cancer Res 1988;48:7279-84. 7. Wu-Williams AH, Dai XD, Blot W, et al. Lung cancer among women in north-east China. Br J Cancer 1990;62:982-7. 8. Osann KE. Lung cancer in women: the importance of smok ing, family history of cancer, and medical history of respira tory disease. Cancer Res 1991;51:4893-7. 9. Shaw GL, Falk RT, Pickle LW, et al. Lung cancer risk associated with cancer in relatives. J Clin Epidemiol 1991 ;44: 429-37. 10. Horwitz R, Smaldone LF, Viscoli CM. An ecogenetic hypoth esis for lung cancer in women. Arch Intern Med 1988; 148: 2609-12. 11. Sellers TA, Potter JD, Folsom AR. Association of incident lung cancer with family history of female reproductive cancers: The Iowa Women's Health Study. Genet Epidemiol 1991;8:199-208. 12. Fontham ETH, Correa P, Reynolds P, et al. Environmental tobacco smoke and lung cancer in nonsmoking women: a multicenter study. JAMA 1994;271:1752-9. 13. Fontham ETH, Correa P, Wu-Williams A, et al. Lung cancer in nonsmoking women: a multicenter case-control study. Can cer Epidemiol Biomark Prev 1991;1:35-43. 14. Haley NJ, Axelrad CM, Tilton KA. Validation of self-reported smoking behavior: biochemical analysis of cotinine and thio cyanate. Am J Public Health 1983;73:1204-7. 15. Breslow NE, Day NE. Statistical methods in cancer research. Vol 1. The analysis of case-control studies. Lyon, France: International Agency for Research on Cancer, 1980. (IARC scientific publication no. 32). 16. Taioli E, Wynder EL. Endocrine factors and adenocarcinoma of the lung in women. J Natl Cancer Inst 1994;86:869-70. 17. McDuffie HH. Clustering of cancer in families of patients with primary lung cancer. J Clin Epidemiol 1991;44:69-76. 18. Lynch HT, Kimberling WJ, Markvicka SE, et al. Genetics and smoking-associated cancers: a study of 485 families. Cancer 1986;57:1640-6. 19. Ambrosone CB, Rao U, Michalek AM, et al. Lung cancer histologic types and family history of cancer. Cancer 1993; 72:1192-8. 20. Sellers TA, Elston RC, Atwood LD, et al. Lung cancer histo logic type and family history of cancer. Cancer 1992;69: 86-91. 21. Benatti P, Sassatelli R, Roncucci L, et al. Tumor spectrum in hereditary non-polyposis colorectal cancer (HNPCC) and in families with "suspected HNPCC": a population-based study in Northern Italy. Int J Cancer 1993;54:371-7. 22. Watson P, Lynch HT. The tumor spectrum in HNPCC. Anti cancer Res 1994;14:1635-40. 23. Peltomaki P, Aaltonen LA, Sistonene P, et al. Genetic map ping of a locus predisposing to human colorectal cancer. Science 1993;260:810-12. 24. Shridhar V, Siegfried J, Hunt J, et al. Genetic instability of microsatellite sequences in many non-small cell lung carcino mas. Cancer Res 1994;54:2084-7. 25. Merlo A, Mabry M, Gabrielson E, et al. Frequent microsatel lite instability in primary small cell lung cancer. Cancer Res 1994;54:2098-101. 26. Aaltonen LA, Peltomaki P, Mecklin JP, et al. Replication errors in benign and malignant tumors from hereditary non polyposis colorectal cancer patients. Cancer Res 1994;54: 1645- 8. 27. Sellers TA, Bailey-Wilson JE, Elston RC, et al. Evidence for mendelian inheritance in the pathogenesis of lung cancer. J Natl Cancer Inst 1990;82:1272-9. 28. Sellers TA, Chen PL, Potter JD, et al. Segregation analysis of smoking-associated malignancies: evidence for mendelian in heritance. Am J Med Genet 1994;52:308-14. 29. World Health Organization. International classification of dis eases, ninth revision, clinical modification. 4th ed. Vol 1 and 2. Salt Lake City, UT: Med-Index Publications, 1993. Am J Epidemiol Vol. 143, No. 6, 1996