American Journal of Volume 136 Number 6 September 15,1992
EPIDEMIOLOGY Copyright © 1992 by The Johns Hopkins University School of Hygiene and Public Health Sponsored by the Society for Epidemiologic Research
ORIGINAL CONTRIBUTIONS Preexisting Lung Disease and Lung Cancer among Nonsmoking Women
Michael C. R. Alavanja,1 Ross C. Brownson,2 John D. Boice, Jr.,1 and Ed Hock3
Preexisting lung disease was examined as a risk factor for lung cancer in a populationbased, case-control study of nonsmoking women in Missouri conducted between June 1, 1986, and April 1, 1991. A history of lung disease was reported by approximately 4 1 % of 618 cases and 35% of 1,402 controls (odds ratio (OR) = 1.2; 95% confidence interval (Cl) 1.0-1.5. The risk was more pronounced when next-of-kin interviews were excluded (OR = 1.5). Previous lung disease was significantly related both to adenocarcinoma (OR = 1.4), which accounted for 62% of the cancers, and to all other cell types of lung cancer combined (OR = 1.8). Despite having discontinued smoking for more than 15 years, long-term ex-smokers were at a 2.2-fold risk of lung cancer compared with lifetime nonsmokers. Among lifetime nonsmokers, significant risks were noted for asthma (OR = 2.7) and pneumonia (OR = 1.5). Emphysema (OR = 2.6) and tuberculosis (OR = 2.0) were also significantly related to lung cancer, but only among former smokers. Chronic bronchitis was linked to elevated risks of nonadenocarcinomas only (OR = 2.3). Pleurisy was not reported more frequently by cases than by controls. Approximately 16% of all lung cancers among nonsmoking women could be attributed to previous lung diseases, most notably asthma, pneumonia, emphysema, and tuberculosis. Am J Epidemiol 1992;136:623-32. adenocarcinoma, asthma, emphysema, lung diseases, lung neoplasms, pneumonia, tuberculosis
Between 9 and 22 percent of the 55,000 annual lung cancers in women and 1 and 13
percent of the 102,000 annual lung cancers in men are estimated to occur in non-
Received for publication June 13, 1991, and in final form March 19, 1992 Abbreviations: Cl, confidence interval; OR, odds ratio. 1 Epidemiology and Biostatistics Program, National Cancer Institute, Executive Plaza North, Rockville, MD. 2 Division of Chronic Disease Prevention and Health Promotion, Missouri Department of Health, Columbia, MO. 3 Information Management Services (IMS), Rockville, MD. Reprint requests to Dr. Michael C. R. Alavanja, Epidemiology and Biostatistics Program, National Cancer Institute, 6130 Executive Blvd., Executive Plaza North, Room
543, Rockville, MD 20852 The authors thank the Health Care Financing Administration for their help in selecting population-based controls for their study; Ms Kelly LaVern for her expert assistance in typing the manuscript; the staff of Survey Research Associates, Inc., for their assistance, in particular Sandi Ezrine, Patsy Henderson, and Joan Huber; and our panel of pathologists for their expert help, especially Drs Tim Loy and Ellis Ingram of the University of Missouri School of Medicine, Columbia, MO, and Dr. Jeffery Meyer of the Mayo Clinic, Rochester, MN.
623
624
Alavanja et al.
smokers (1-3). The reasons nonsmokers develop lung cancer have not been well defined, although the possible relation to previous nonmalignant lung disease has attracted attention for several decades (413) and continues to be the focus of recent studies (14-22). To investigate further the possible role that tuberculosis, chronic bronchitis, emphysema, asthma, and pneumonia might play in the development of lung cancer, we conducted a population-based, casecontrol study of lung cancer among nonsmoking women in the state of Missouri. MATERIALS AND METHODS Cases
Approximately 3,475 women with primary lung cancer were reported to the Missouri Cancer Registry between June 1, 1986, and April 1, 1991. Smoking status and race are routinely reported to the registry, which facilitated the selection of white, nonsmoking women for the study. Selection was limited to whites because of the small numbers of members of other racial/ethnic groups. Overall, 650 (19 percent) of all lung cancers arose in nonsmokers between the ages of 30 and 84 years. Physicians denied permission to interview 24 of these cases (4 percent), and an additional 8 (1 percent) refused to be interviewed, which left 618 cases (95 percent). This group included 432 lifetime nonsmokers (70 percent) and 186 former smokers (30 percent). Two hundred sixteen cases were personally interviewed by telephone; because of death or serious illness, next-ofkin interviews were obtained for the other 402 cases. Among the 131 cases under 65 years of age, 118 (90.1 percent) had a valid Missouri driver's license at the time of diagnosis, and all 487 cases 65 years or older were registered with the Health Care Financing Administration. In addition to the diagnosis of lung cancer reported by the registry, tissue slides from 468 cases (76 percent) were reviewed for histologic verification. Slides for these cases were examined simultaneously by three pathologists (T.L., J.M., and E.I.) using a multiheaded microscope; they did not know the
referring pathologists' diagnoses. For surgical specimens, consensus diagnoses were based on the classification criteria outlined by the World Health Organization (23). When only cytologic material was available, consensus diagnoses were based on standard cytologic criteria (24). Controls
A population-based sample of white, nonsmoking female controls was selected from two sources. For women aged 30-64 years, names and addresses were randomly generated from driver's license files provided by the Missouri Department of Revenue. For women aged 65-84 years, names and addresses were randomly generated from lists provided by the Health Care Financing Administration that include an estimated 95 percent of the women of this age group (25). The number and age distribution of the controls selected for interview were frequency matched in 5-year age strata at a ratio of approximately 2.2 to 1, based on the number of cases reported to the cancer registry in previous years. Most telephone numbers for cases and controls were obtained from a company providing computer matching of names and addresses with telephone numbers. Telephone numbers not identified in this manner were sought using directory assistance and local directories, and by contacting relatives. If a telephone number was still not found, a letter was mailed to the study subject requesting a telephone number for telephone interview. Of the 1,862 potential nonsmoking controls, 1,527 (82 percent) agreed to respond to the initial screening interview, of whom 1,402 (75 percent) agreed to complete the full telephone questionnaire. No controls completing the full telephone questionnaire had lung cancer. Two controls and three cases could not remember their history of nonmalignant respiratory disease and were eliminated from the analyses. Questionnaire administration and design
At the time of cancer diagnosis, the mean age of the cases was 71.5 years; the mean
Preexisting Lung Disease and Lung Cancer
age of the controls, based on the date of cancer diagnosis in the corresponding case, was 69.9 years. Interviews of cases who developed lung cancer after the study began and of their corresponding controls were conducted within 8 months of the cancer diagnosis. The interviews of 80 percent of these women were completed within 4 months of diagnosis. Interviews of cases reported to the registry before the beginning of the study and of their corresponding controls were completed within 2.5 years of the cancer diagnosis. All interviews obtained information only on the period of life preceding the date of cancer diagnosis. A telephone-administered screening questionnaire was used to determine and/or verify the eligibility of cases and controls on age, gender, race, and smoking status. Both lifetime nonsmokers and former smokers were included in the study. A lifetime nonsmoker was defined as a woman who had never smoked more than 100 cigarettes or used any other tobacco products for more than 6 months during her lifetime. To minimize the effect of previous smoking, a former smoker was defined as a woman who had ceased using all tobacco products 15 years or more before the interview (median period of smoking cessation = 24 years). Information on residential history, passive smoking exposure, family health history, reproductive history, occupational history, diet, and previous lung disease was obtained from a structured questionnaire administered by trained interviewers. Subjects were asked whether they had ever had a diagnosis by a physician of any form of nonmalignant lung disease, including asthma, abscesses of the lung, chronic bronchitis, chronic obstructive pulmonary disease, pneumonia, pleurisy, tuberculosis, and other lung conditions. For each positive response, additional questions were asked regarding the age and date of first diagnosis. For cases who could not be interviewed because of death or ill health, a spouse, adult child, or sibling familiar with the health history of the case was sought for a next-of-kin interview. Nextof-kin interviews were conducted for 65 percent of the cases and none of the controls.
625
As a quality control procedure, a 5 percent sample of study subjects was selected for a second interview at the end of the field interview period. The second interview focused on active and passive smoking and previous lung disease using the same questions and structured interview format as the original interview. Analysis
Multivariate logistic regression methods for frequency-matched case-control studies were used to compute estimated odds ratios for each variable (26). Confidence intervals were determined by Gait's method (27). Passive smoking and previous smoking history were treated as dichotomous variables in the adjustment process in all adjusted analyses presented. More complex methods of adjustment (for example, by duration or intensity of exposure) did not produce estimates of risk that differed from those presented. Age was treated as a continuous variable in the adjustment process. The percentage of attributable risk for the population (28) was computed as follows: Percentage of attributable risk = P(OR - 1)/[1 + (P)(OR - 1)] x 100, where P is the prevalence of exposure in the control population and OR is the odds ratio for exposure to the independent variable under study. RESULTS
Most women in our series developed lung cancer after the age of 70 years, were married, and had completed high school (table 1). There were few differences between the 618 cases and 1,402 controls in any of the demographic characteristics evaluated. However, the proportion of former smokers (women who had quit smoking more than 15 years previously; median period of cessation = 24 years), was about twice as high among lung cancer cases (30 percent) as among controls (17 percent). Pathologic material from 468 cases was available for review. Adenocarcinoma was the most frequent lung cancer cell type (62
626
Alavanja et al.
TABLE 1. Sociodemographic characteristics of nonsmoking women with lung cancer and controls at the time of cancer diagnosis: Missouri, 1986-1991 Controls (n = 1,402)
Cases (n = 618) Characteristic No.
%
Age at interview (years) 74
46 85 193 294
7 14 32 47
103 233 457 609
7 17 32 43
Education (years) 12 Unknown
240 228 121 29
38 37 20 5
537 478 354 33
38 34 25 2
Marital status Married Widowed Separated Divorced Never married
292 269 3 28 26
48 43
EPIDEMIOLOGY Copyright © 1992 by The Johns Hopkins University School of Hygiene and Public Health Sponsored by the Society for Epidemiologic Research
ORIGINAL CONTRIBUTIONS Preexisting Lung Disease and Lung Cancer among Nonsmoking Women
Michael C. R. Alavanja,1 Ross C. Brownson,2 John D. Boice, Jr.,1 and Ed Hock3
Preexisting lung disease was examined as a risk factor for lung cancer in a populationbased, case-control study of nonsmoking women in Missouri conducted between June 1, 1986, and April 1, 1991. A history of lung disease was reported by approximately 4 1 % of 618 cases and 35% of 1,402 controls (odds ratio (OR) = 1.2; 95% confidence interval (Cl) 1.0-1.5. The risk was more pronounced when next-of-kin interviews were excluded (OR = 1.5). Previous lung disease was significantly related both to adenocarcinoma (OR = 1.4), which accounted for 62% of the cancers, and to all other cell types of lung cancer combined (OR = 1.8). Despite having discontinued smoking for more than 15 years, long-term ex-smokers were at a 2.2-fold risk of lung cancer compared with lifetime nonsmokers. Among lifetime nonsmokers, significant risks were noted for asthma (OR = 2.7) and pneumonia (OR = 1.5). Emphysema (OR = 2.6) and tuberculosis (OR = 2.0) were also significantly related to lung cancer, but only among former smokers. Chronic bronchitis was linked to elevated risks of nonadenocarcinomas only (OR = 2.3). Pleurisy was not reported more frequently by cases than by controls. Approximately 16% of all lung cancers among nonsmoking women could be attributed to previous lung diseases, most notably asthma, pneumonia, emphysema, and tuberculosis. Am J Epidemiol 1992;136:623-32. adenocarcinoma, asthma, emphysema, lung diseases, lung neoplasms, pneumonia, tuberculosis
Between 9 and 22 percent of the 55,000 annual lung cancers in women and 1 and 13
percent of the 102,000 annual lung cancers in men are estimated to occur in non-
Received for publication June 13, 1991, and in final form March 19, 1992 Abbreviations: Cl, confidence interval; OR, odds ratio. 1 Epidemiology and Biostatistics Program, National Cancer Institute, Executive Plaza North, Rockville, MD. 2 Division of Chronic Disease Prevention and Health Promotion, Missouri Department of Health, Columbia, MO. 3 Information Management Services (IMS), Rockville, MD. Reprint requests to Dr. Michael C. R. Alavanja, Epidemiology and Biostatistics Program, National Cancer Institute, 6130 Executive Blvd., Executive Plaza North, Room
543, Rockville, MD 20852 The authors thank the Health Care Financing Administration for their help in selecting population-based controls for their study; Ms Kelly LaVern for her expert assistance in typing the manuscript; the staff of Survey Research Associates, Inc., for their assistance, in particular Sandi Ezrine, Patsy Henderson, and Joan Huber; and our panel of pathologists for their expert help, especially Drs Tim Loy and Ellis Ingram of the University of Missouri School of Medicine, Columbia, MO, and Dr. Jeffery Meyer of the Mayo Clinic, Rochester, MN.
623
624
Alavanja et al.
smokers (1-3). The reasons nonsmokers develop lung cancer have not been well defined, although the possible relation to previous nonmalignant lung disease has attracted attention for several decades (413) and continues to be the focus of recent studies (14-22). To investigate further the possible role that tuberculosis, chronic bronchitis, emphysema, asthma, and pneumonia might play in the development of lung cancer, we conducted a population-based, casecontrol study of lung cancer among nonsmoking women in the state of Missouri. MATERIALS AND METHODS Cases
Approximately 3,475 women with primary lung cancer were reported to the Missouri Cancer Registry between June 1, 1986, and April 1, 1991. Smoking status and race are routinely reported to the registry, which facilitated the selection of white, nonsmoking women for the study. Selection was limited to whites because of the small numbers of members of other racial/ethnic groups. Overall, 650 (19 percent) of all lung cancers arose in nonsmokers between the ages of 30 and 84 years. Physicians denied permission to interview 24 of these cases (4 percent), and an additional 8 (1 percent) refused to be interviewed, which left 618 cases (95 percent). This group included 432 lifetime nonsmokers (70 percent) and 186 former smokers (30 percent). Two hundred sixteen cases were personally interviewed by telephone; because of death or serious illness, next-ofkin interviews were obtained for the other 402 cases. Among the 131 cases under 65 years of age, 118 (90.1 percent) had a valid Missouri driver's license at the time of diagnosis, and all 487 cases 65 years or older were registered with the Health Care Financing Administration. In addition to the diagnosis of lung cancer reported by the registry, tissue slides from 468 cases (76 percent) were reviewed for histologic verification. Slides for these cases were examined simultaneously by three pathologists (T.L., J.M., and E.I.) using a multiheaded microscope; they did not know the
referring pathologists' diagnoses. For surgical specimens, consensus diagnoses were based on the classification criteria outlined by the World Health Organization (23). When only cytologic material was available, consensus diagnoses were based on standard cytologic criteria (24). Controls
A population-based sample of white, nonsmoking female controls was selected from two sources. For women aged 30-64 years, names and addresses were randomly generated from driver's license files provided by the Missouri Department of Revenue. For women aged 65-84 years, names and addresses were randomly generated from lists provided by the Health Care Financing Administration that include an estimated 95 percent of the women of this age group (25). The number and age distribution of the controls selected for interview were frequency matched in 5-year age strata at a ratio of approximately 2.2 to 1, based on the number of cases reported to the cancer registry in previous years. Most telephone numbers for cases and controls were obtained from a company providing computer matching of names and addresses with telephone numbers. Telephone numbers not identified in this manner were sought using directory assistance and local directories, and by contacting relatives. If a telephone number was still not found, a letter was mailed to the study subject requesting a telephone number for telephone interview. Of the 1,862 potential nonsmoking controls, 1,527 (82 percent) agreed to respond to the initial screening interview, of whom 1,402 (75 percent) agreed to complete the full telephone questionnaire. No controls completing the full telephone questionnaire had lung cancer. Two controls and three cases could not remember their history of nonmalignant respiratory disease and were eliminated from the analyses. Questionnaire administration and design
At the time of cancer diagnosis, the mean age of the cases was 71.5 years; the mean
Preexisting Lung Disease and Lung Cancer
age of the controls, based on the date of cancer diagnosis in the corresponding case, was 69.9 years. Interviews of cases who developed lung cancer after the study began and of their corresponding controls were conducted within 8 months of the cancer diagnosis. The interviews of 80 percent of these women were completed within 4 months of diagnosis. Interviews of cases reported to the registry before the beginning of the study and of their corresponding controls were completed within 2.5 years of the cancer diagnosis. All interviews obtained information only on the period of life preceding the date of cancer diagnosis. A telephone-administered screening questionnaire was used to determine and/or verify the eligibility of cases and controls on age, gender, race, and smoking status. Both lifetime nonsmokers and former smokers were included in the study. A lifetime nonsmoker was defined as a woman who had never smoked more than 100 cigarettes or used any other tobacco products for more than 6 months during her lifetime. To minimize the effect of previous smoking, a former smoker was defined as a woman who had ceased using all tobacco products 15 years or more before the interview (median period of smoking cessation = 24 years). Information on residential history, passive smoking exposure, family health history, reproductive history, occupational history, diet, and previous lung disease was obtained from a structured questionnaire administered by trained interviewers. Subjects were asked whether they had ever had a diagnosis by a physician of any form of nonmalignant lung disease, including asthma, abscesses of the lung, chronic bronchitis, chronic obstructive pulmonary disease, pneumonia, pleurisy, tuberculosis, and other lung conditions. For each positive response, additional questions were asked regarding the age and date of first diagnosis. For cases who could not be interviewed because of death or ill health, a spouse, adult child, or sibling familiar with the health history of the case was sought for a next-of-kin interview. Nextof-kin interviews were conducted for 65 percent of the cases and none of the controls.
625
As a quality control procedure, a 5 percent sample of study subjects was selected for a second interview at the end of the field interview period. The second interview focused on active and passive smoking and previous lung disease using the same questions and structured interview format as the original interview. Analysis
Multivariate logistic regression methods for frequency-matched case-control studies were used to compute estimated odds ratios for each variable (26). Confidence intervals were determined by Gait's method (27). Passive smoking and previous smoking history were treated as dichotomous variables in the adjustment process in all adjusted analyses presented. More complex methods of adjustment (for example, by duration or intensity of exposure) did not produce estimates of risk that differed from those presented. Age was treated as a continuous variable in the adjustment process. The percentage of attributable risk for the population (28) was computed as follows: Percentage of attributable risk = P(OR - 1)/[1 + (P)(OR - 1)] x 100, where P is the prevalence of exposure in the control population and OR is the odds ratio for exposure to the independent variable under study. RESULTS
Most women in our series developed lung cancer after the age of 70 years, were married, and had completed high school (table 1). There were few differences between the 618 cases and 1,402 controls in any of the demographic characteristics evaluated. However, the proportion of former smokers (women who had quit smoking more than 15 years previously; median period of cessation = 24 years), was about twice as high among lung cancer cases (30 percent) as among controls (17 percent). Pathologic material from 468 cases was available for review. Adenocarcinoma was the most frequent lung cancer cell type (62
626
Alavanja et al.
TABLE 1. Sociodemographic characteristics of nonsmoking women with lung cancer and controls at the time of cancer diagnosis: Missouri, 1986-1991 Controls (n = 1,402)
Cases (n = 618) Characteristic No.
%
Age at interview (years) 74
46 85 193 294
7 14 32 47
103 233 457 609
7 17 32 43
Education (years) 12 Unknown
240 228 121 29
38 37 20 5
537 478 354 33
38 34 25 2
Marital status Married Widowed Separated Divorced Never married
292 269 3 28 26
48 43
