1 Division of Biostatistics, School of Public Health, University of Minnesota, Minneapolis, MN, USA.
2 Division of Preventive and Behavioral Medicine, University of Massachusetts Medical School, Worcester, MA, USA.
3 Department of Epidemiology, School of Public Health and Tropical Medicine, Tulane University, New Orleans, LA, USA.
4 Department of Biostatistics and Epidemiology, School of Medicine, University of Pennsylvania, Philadelphia, PA, USA (current address).
5 Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, PA, USA.
Correspondence: LE Eberly, 420 Delaware St. SE, MMC 303, Minneapolis, MN, 55455, USA. E-mail: lynn{at}biostat.umn.edu
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
Methods In all, 6613 MRFIT baseline smokers alive at trial end in 1982 had acceptable FEV1 measures and complete smoking history; men were classified as during-trial long-term quitters (N = 1292), intermittent quitters (1961), and never quitters (3360). Proportional hazards models for LCM were fit with quintiles of average FEV1, adjusted for age, height, race, smoking history, and other risk factors.
Results For long-term, intermittent, and never quitters respectively, mean baseline cigarettes/ day was 28, 32, and 35; trial-averaged FEV1 was 3201, 3146, and 3082 ml; and average decline in FEV1 was 46.0, 54.6, and 62.5 ml/year. With median post-trial mortality follow-up of 18 years, there were 363 lung cancer deaths. Age-adjusted LCM rates varied across FEV1 quintiles from 50 (lowest quintile) to 11 (highest quintile), 58 to 11, and 76 to 20, per 10 000 person-years, for long-term quitters, intermittent quitters, and never quitters, respectively. Multivariate adjusted hazard ratios for 100 ml higher FEV1 were 0.92 [P = 0.004], 0.95 [P = 0.003], and 0.95 [P < 0.0001] respectively.
Conclusions These results demonstrate the strong predictive value of FEV1 for lung cancer among cigarette smokers independent of smoking history; results did not differ by during-trial quit status. FEV1 may be a biological marker for smoking dose or it may be that genetic susceptibilities to both decreased FEV1 and lung cancer are associated.
Accepted 14 March 2003
Poor clinical outcomes after a diagnosis of lung cancer have generated considerable interest in both early detection of lung cancer (and hopefully improved subsequent survival) and chemoprevention to reduce the risk of incident lung cancer among current and former cigarette smokers. The utilization of new screening techniques, and probably of chemopreventive agents, could be greatly enhanced by improved methods to identify higher risk groups of smokers and ex-smokers for developing lung cancer. The identification of high-risk individuals could reduce the number of smokers and ex-smokers to be screened for lung cancer, and could enhance the potential benefits versus likely adverse side-effects of chemopreventive agents.
Screening methods to detect lung cancer, recently reviewed,1 using chest X-ray and sputum cytology have demonstrated little benefit in reducing lung cancer mortality (LCM) in clinical trials.24 Newer methods using computerized axial tomography (CT) of the lung are being evaluated and preliminary data suggest that earlier detection of lung cancer may be feasible.59 Other screening methods to identify those cigarette smokers at higher risk for lung cancer include the evaluation of genetic markers of the metabolism of cigarette carcinogens, and/or host susceptibility, and clinical measures such as symptoms of chronic bronchitis or decreased pulmonary function.1015 Lange and colleagues10 showed that those with percentage-predicted forced expiratory volume in 1 second (FEV1) <40 had a risk of lung cancer death 3.9 times higher than those with 80+; that study included smokers and non-smokers.
A previous investigation of lung cancer and pulmonary function (as measured by FEV1) in the Multiple Risk Factor Intervention Trial (MRFIT) cohort16 examined the association of FEV1 with LCM among smokers at 7.5 years of post-trial mortality follow-up. A single measure of FEV1 was a significant predictor of LCM with a multivariate adjusted hazard ratio (HR) of 0.49 per 1 litre higher FEV1 (P = 0.001); the adjustment included multiple measures of smoking dose. In other work, Townsend and colleagues17 examined the associations between FEV1 and during-trial smoking cessation, concluding that quitters, compared with continuing smokers, had smaller decreases in FEV1 across the MRFIT trial years. The LCM in the MRFIT cohort of smokers examined relative to smoking dose18 demonstrated an increased risk over 10 years of 30% for every 10 cigarette/day increase (HR = 1.30, P < 0.001), and relative to the randomized intervention groups19 demonstrated equivalent risk over 15.8 years (HR = 1.17, P > 0.05). With almost five times as many lung cancer deaths now than in the original MRFIT exploration of FEV1,16 we can re-examine with greater power both the overall association among smokers between FEV1 and LCM and whether or not that association differs according to during-trial smoking cessation.
![]() |
Methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
Participants for the trial were screened over three visits prior to randomization. At Screen 1, 8194 of the randomized men (63.7%) reported smoking cigarettes (average 34 cigarettes/day). At Screen 2, serum thiocyanate (a chemical measure of smoking dose)28 was collected, a physical examination was performed, and pulmonary function (FEV1) was determined. At Screen 3, a detailed smoking history was collected. Of the 12 866 men who were eligible and enrolled in the trial from December 1973 through February 1976, 6428 were randomized to special intervention (SI) and 6438 were randomized to usual care (UC). SI men were given dietary advice to lower blood cholesterol, smoking cessation counselling, and hypertension medication using a stepped care approach; UC men were not offered interventions at the clinical centres and were referred back to their usual source of medical care.21 All participants were asked to return to their MRFIT clinical centre (22 centres in 18 US cities) once a year for 6 years for a comprehensive evaluation, including FEV1 measurement, assessment of risk factors, medical history update, and morbidity status. Results regarding the randomized intervention have been reported elsewhere.20,22 Follow-up rates in MRFIT were excellent with fewer than 10% of annual visits missed in each year.22
Pulmonary function
Pulmonary function (FEV1) was measured at Screen 2 and annually thereafter; techniques and quality control procedures have been described elsewhere.29 FEV1 was measured using a 10-l Stead-Wells water-filled spirometer (Stead-Wells, Braintree, MA, USA) and adjusted for ambient room temperature. Standardization of pulmonary function measures and calibration checks were not given high priority at the start of the trial; a rigorous standardization programme was not introduced until several years into the trial. The standardization29 met standards subsequently developed by the American Thoracic Society30 and the Epidemiology and Standardization Project.31 By Annual Visit 3, approximately 85% of the tracings met the standards.16,32
The baseline FEV1 measure used in this paper is based on Annual Visit 3 data because of the smaller proportion of acceptable tracings at true baseline (Screen 2) and at Annual Visits 1 and 2. Trial-averaged FEV1 is the average of FEV1 determinations at Annual Visits 3, 4, 5, and 6. Average annual change in FEV1 is the slope in FEV1 across Annual Visits 3, 4, 5, and 6.
Smoking status
Smoking status is defined in this study based on whether an individual smoked at baseline and at Annual Visits. Never smokers are defined as individuals who reported never smoking at Screen 3. Ex-smokers are defined as those who reported at Screen 3 that they had been a smoker in the past and were not current smokers. Long-term quitters are defined as those participants who reported smoking at Screen 3 and reported not smoking at each of Annual Visits 2, 3, 4, 5, and 6 (regardless of smoking status reported at Annual Visit 1). Never quitters are defined as those participants who reported smoking at Screen 3 and at each of Annual Visits 1 through 6. The intermediate group, intermittent quitters, are defined as those participants who reported smoking at Screen 3 and are neither long-term quitters nor never quitters (i.e. they reported not smoking at one or more of Annual Visits 1 through 6, but not at all of 2 through 6). If a participant missed one or more of these visits then only the visits he attended were used in determining smoking status.
Mortality ascertainment
During the trial and at termination of active intervention on 28 February 1982, vital status was ascertained by clinic staff through contact with the participant, or with family or friends, and through searches of publicly accessible files if the participant was thought to be deceased. Causes of death were verified by clinical staff and coded using International Classification of Diseases, Ninth Revision (ICD-9).3335 Post-trial mortality through December 1990 was determined by matching identifying information, provided by each participant during screening, with US National Death Index (NDI) records.3638 Death certificates were obtained to ascertain underlying cause of death, coded independently by two nosologists; a third nosologist adjudicated any disagreements. Death dates and corresponding ICD-9 or ICD-1033,39 causes from January 1991 through December 1999 were obtained using the NDI-Plus service. Mortality ascertainment is estimated to be approximately 100% complete using these data sources.38 ICD-9 code 162 and ICD-10 codes C33C34 were used to identify lung cancer deaths.
Statistical methods
Among the smokers, five groups of participants were created according to quintiles of trial-averaged FEV1. Participant characteristics were summarized within each FEV1 quintile and tested for group differences with ANOVA or 2 analysis, as appropriate; sub-group differences among the smoking groups were similarly explored. Death rates per 10 000 person-years, age-adjusted by the direct method to the full cohort of MRFIT screenees (N = 361 662), and Kaplan-Meier curves for time from sixth anniversary of randomization to lung cancer death were computed for each quintile. Univariate and multivariate proportional hazards models40 were carried out on time to lung cancer death to test quintile differences; adjusting variables were for known lung cancer risk factors (age at randomization, cigarettes/day, thiocyanate, age at which smoking began, use of filter cigarettes, cigarette tar and nicotine content), MRFIT design variables (DBP, fasting cholesterol, randomized intervention group [SI or UC]), and potential moderators of the FEV1 effect (race, height, body mass index [BMI]); analyses were stratified by clinical centre. These models were repeated using a single FEV1 from Annual Visit 3 and then using quintiles of the average annual change (slope) in FEV1, additionally adjusting for Annual Visit 3 FEV1. Models using trial-averaged FEV1 were run separately within smoking sub-groups (never quitters, intermittent quitters, long-term quitters). The associations of FEV1 with mortality were compared for deaths within 10 years to deaths after 10 years, and separately for the SI compared with the UC men. P-values given are two-tailed; no adjustments for multiple comparisons were made.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
Comparing across smoking sub-groups, long-term and intermittent quitters smoked significantly fewer cigarettes/day at baseline (2 d.f., F-test P < 0.0001), showed lower thiocyanate values (P < 0.0001), and consumed fewer alcoholic drinks/week (P < 0.0001) than never quitters. Height and cholesterol were not significantly different at level 0.05. Over 6 trial years, average cigarettes/day smoked were 30.5 for never quitters, 15.0 for intermittent quitters, and 0.9 for long-term quitters (P < 0.0001). FEV1 was highest for long-term quitters and lowest for never quitters (P < 0.0001), while FEV1 declined on average 46.0, 54.6, and 62.5 ml per year for the three groups respectively (P < 0.0001).
Over a median 18 years post-trial follow-up, there were 2547 deaths, including 363 from lung cancer. There were 127, 87, 62, 52, and 35 lung cancer deaths in the lowest to highest FEV1 quintiles, corresponding to decreasing age-adjusted death rates of 64.1, 41.4, 30.6, 25.6, and 16.5 deaths per 10 000 person-years respectively (Table 2). Kaplan-Meier curves of survival by FEV1 quintiles showed significant differences (Figure 1
; log rank test
2 = 112.4, d.f. = 4, P < 0.0001). Survival probability began dropping immediately for the lowest quintile, after about 2 years for the second quintile, and not until about 7 years for the third through fifth quintiles. There were 238, 82, and 43 lung cancer deaths among the never, intermittent, and long-term quitters, respectively, with decreasing age-adjusted death rates of 52.6, 27.8, and 21.0 deaths per 10 000 person-years. In contrast, lung cancer death rates among MRFIT self-identified never smokers (N = 1792) and ex-smokers (N = 3205) were 3.6 and 11.5 deaths per 10 000 person-years (11 and 62 deaths), respectively.
|
|
Proportional hazards regression results across smoking sub-groups showed quintile-specific HR that were very similar to those in the smokers overall (Table 3). There was a slightly weaker linear association of FEV1 with mortality among never quitters (HR per 100 ml higher 0.95, 95% CI: 0.93, 0.97, P < 0.0001) than among long-term quitters (HR = 0.92, 95% CI: 0.88, 0.97, P = 0.03). There was no significant interaction between FEV1 quintile and smoking sub-groups (P = 0.80).
|
Results among all smokers using a single measure of FEV1 showed similar results (Table 4) as those using an averaged FEV1. The multivariate-adjusted HR per 100 ml higher FEV1 was 0.95 (95% CI: 0.93, 0.97, P < 0.0001). Results among all smokers using the average annual change (slope) in FEV1 were not as strong, with no significant results for comparisons across quintiles of average annual change (results not shown), but a marginally significant linear trend (HR = 0.89, 95% CI: 0.80, 0.99, P = 0.04). Here the HR represents an 11% decrease in risk associated with a 100-ml higher average annual change (e.g. a participant whose FEV1 is unchanged in a year compared with a participant whose FEV1 declines by 100 ml in a year).
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
We used proportional hazards multiple regression with FEV1 (controlling for age, height, and race) rather than with percentage predicted FEV1 (based on normative equations dependent on age, height, and race). Our approach does not require an internal or external comparison group from which the normative equations are derived, and it can reduce the variance of estimated effects because no additional variability is introduced with the use of a normative equation. Our approach allows independent estimation of the effects of and also direct adjustment for age, height, and race. For MRFIT baseline never smokers and ex-smokers, smoking status reported at subsequent Annual Visits was not used; re-defining those never and ex-smokers who smoked during the trial did not affect our results.
The more recent application of CT scan to the early detection of lung cancer has once again raised hopes of diagnosis at a stage of development that permits survival to be favourably affected. Given the relatively low incidence (even in heavy smokers) and the long latency of lung cancer, a definitive randomized trial of CT scan screening would need to be both large and long-term, and thus expensive; the US National Cancer Institute is currently funding CT scan studies (http://researchportfolio.cancer.gov). Additional predictors of lung cancer such as pulmonary function, over and above the rate and duration of cigarette smoking, could be valuable in identifying those smokers most likely to develop lung cancer to enable a smaller, more economical randomized screening trial.
A basic problem with CT of the lungs (and other potential early detection methods) is the relatively low positive predictive value of a CT-identified lesion. Such a lesion usually leads to further diagnostic testing which is both expensive and of potential risk for the patient, i.e. thorachotomy and biopsy. Incorporating additional screening via FEV1 could improve the positive predictive value of the CT screening. In this study, the lung cancer death rate among smokers in the lowest quintile of FEV1 was 64.1 per 10 000 person-years and 16.5 per 10 000 person-years in the highest FEV1 quintile. Assuming a 10-year follow-up, sensitivity of 80%, and specificity of 95%, then the positive predictive value of a CT-identified lesion would be about 52% for those with low FEV1 and only 21% for those with the highest FEV1. In other words, about one in every two lesions identified in those with low FEV1 would be a cancer as opposed to only one in five for those with the highest FEV1. It is important to note that the FEV1 values in this study were not low enough to cause significant disability; participants with very low FEV1 or a history of chronic pulmonary disease were excluded from the original MRFIT trial.
In the absence of a preventive or therapeutic intervention of proven efficacy for lung cancer (other than smoking cessation) the implications for public health practice are thus limited to using the FEV1 as an additional indicator of risk, perhaps useful for exclusion criteria or risk stratification in clinical trials of lung cancer, and, potentially, as a further incentive to quit smoking.
In this paper, we presented evidence that a well-established and inexpensive measurement, the FEV1, is a strong predictor of lung cancer mortality (and of all-cause mortality) independent of smoking dose. This strong association held true among various sub-groups of smokers defined according to their 6-year smoking cessation pattern. There are two primary hypotheses regarding the mechanism by which FEV1 might affect lung cancer occurrence. The first is that the increased risk of lung cancer seen here among smokers with decreased pulmonary function could be attributed to a genetic susceptibility to both decreased lung function and lung cancer (i.e. host susceptibility among smokers).44 The second is that lung function may be an additional marker of the biological (carcinogenic) effects of smoking exposure beyond measures of duration, dose, and characteristics of cigarette smoking.34 Using the combination of FEV1 and smoking status may identify smokers at unusually high risk of lung cancer.
KEY MESSAGES
|
![]() |
Acknowledgments |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 Kubik AK, Parkin DM, Zatloukal P. Czech study on lung cancer screening: Post-trial follow-up of lung cancer deaths up to year 15 since enrolment. Cancer 2000;89(11Suppl.):236368.[CrossRef][ISI][Medline]
3 Melamed MR. Lung cancer screening results in the National Cancer Institute New York study. Cancer 2000;89(11Suppl.):235662.[CrossRef][ISI][Medline]
4 Kubik A, Haerting J. Survival and mortality in a randomized study of lung cancer detection. Neoplasma 1990;37:46775.[ISI][Medline]
5 Hasegawa M, Sone S, Takashima S et al. Growth rate of small lung cancers detected on mass CT screening. Br J Radiol 2000;73:125259.
6 Sone S, Li F, Yang ZG et al. Results of three-year mass screening programme for lung cancer using mobile low-dose spiral computed tomography scanner. Br J Cancer 2001;84:2532.[CrossRef][ISI][Medline]
7 Mori K, Tominaga K, Hirose T, Sasagawa M, Yokoyama K, Moriyama N. Utility of low-dose helical CT as a second step after plain chest radiography for mass screening for lung cancer. J Thoracic Imaging 1997;12:17380.[ISI][Medline]
8 Wagner H, Ruckdeschel JC. Lung cancer. In: Reintgen D, Clark R (eds). Cancer Screening. St Louis, MO: Mosbyp, 1996, pp. 11849.
9 Marcus PM, Bergstralh EJ, Fagerstrom RM et al. Lung cancer mortality in the Mayo Lung Project: impact of extended follow-up. J Natl Cancer Inst 2000;92:130816.
10 Lange P, Nyboe J, Appleyard M, Jensen G, Schnohr P. Ventilatory function and chronic mucus hypersecretion as predictors of death from lung cancer. Am Rev Respir Dis 1990;141:61317.[ISI][Medline]
11 Amos CI, Caporaso NE, Weston A. Host factors in lung cancer risk: a review of interdisciplinary studies. Cancer Epidemiol Biomarkers Prev 1992;1:50513.[Abstract]
12 Benowitz NL, Perez-Stable EJ, Herrera B, Jacob P III. Slower metabolism and reduced intake of nicotine from cigarette smoking in Chinese-Americans. J Natl Cancer Inst 2002;94:10815.
13 Garte S, Gaspari L, Alexandrie AK et al. Metabolic gene polymorphism frequencies in control populations. Cancer Epidemiol Biomarkers Prev 2001;10:123948.
14 Stucker I, Boffetta P, Antilla S et al. Lack of interaction between asbestos exposure and glutathione S-transferase M1 and T1 genotypes in lung carcinogenesis. Cancer Epidemiol Biomarkers Prev 2001;10:125358.
15 Park JY, Lee SY, Jeon HS et al. Polymorphism of the DNA repair gene XRCC1 and risk of primary lung cancer. Cancer Epidemiol Biomarkers Prev 2002;11:2328.
16 Kuller L, Ockene J, Meilahn E, Svendsen K. Relation of forced expiratory volume in one second (FEV1) to lung cancer mortality in the Multiple Risk Factor Intervention Trial (MRFIT). Am J Epidemiol 1990;132:26574.[Abstract]
17 Townsend MC, DuChene AG, Morgan J, Browner WS. Pulmonary function in relation to cigarette smoking and smoking cessation. Prev Med 1991;20:62137.[ISI][Medline]
18 Kuller LH, Ockene JK, Meilahn E, Wentworth DN, Svendsen KH, Neaton JD. Cigarette smoking and mortality. Prev Med 1991;20:63854.[ISI][Medline]
19 Shaten BJ, Kuller LH, Kjelsberg MO et al. Lung cancer mortality after 16 years in MRFIT participants in intervention and usual care groups. Ann Epidemiol 1997;7:12536.[CrossRef][ISI][Medline]
20 Benfari R. Multiple Risk Factor Intervention Trial (MRFIT): III. The model for intervention. Prev Med 1981;10:42642.[ISI][Medline]
21 Multiple Risk Factor Intervention Trial Research Group. Statistical design considerations in the NHLI Multiple Risk Factor Intervention Trial. J Chronic Dis 1997;30:26175.[CrossRef]
22 Multiple Risk Factor Intervention Trial Research Group. Multiple Risk Factor Intervention Trial: Risk factor changes and mortality results. JAMA 1982;248:146577.[Abstract]
23 Cutler J, Neaton J, Hulley S, Kuller L, Paul O, Stamler J. Coronary heart disease and all-causes mortality in the Multiple Risk Factor Intervention Trial, subgroup findings and comparisons with other trials. Prev Med 1985;14:293311.[CrossRef][ISI][Medline]
24 Stamler J. The Multiple Risk Factor Intervention Trial (MRFIT). In: Hofmann H (ed.). Primary and Secondary Prevention of Coronary Heart Disease Results of New Trials. New York/Berlin: Springer-Verlag, 1985, pp. 833.
25 Multiple Risk Factor Intervention Trial Research Group. Relationship between baseline risk factors and coronary heart disease and total mortality in the Multiple Risk Factor Intervention Trial. Prev Med 1986;15:25473.[ISI][Medline]
26 Multiple Risk Factor Intervention Trial Research Group. Coronary heart disease death, non-fatal acute myocardial infarction and other clinical outcomes in the MRFIT. Am J Cardiol 1986;58:113.[ISI][Medline]
27 Multiple Risk Factor Intervention Trial Research Group. The Multiple Risk Factor Intervention Trial: Quality control of technical procedures and data acquisition. Control Clin Trials 1986;7(Suppl.):1S202S.
28 Hughes G, Hymowitz N, Ockene J, Simmon N, Vogt TM. The Multiple Risk Factor Intervention Trial (MRFIT). Forum: The multiple risk factor intervention trial (MRFIT). The methods and impact of intervention over four years. V. Intervention on smoking. Prev Med 1981;10:476500.[ISI][Medline]
29 Townsend M, Morgan J, Durkin D, DuChene AG, Lamb S. Quality control aspects of pulmonary function testing in the Multiple Risk Factor Intervention Trial. Control Clin Trials 1986;7:179S92S.[CrossRef][Medline]
30 American Thoracic Society. Snowbird workshop on standardization of spirometry. Am Rev Respir Dis 1979;119:83138.[ISI][Medline]
31 Ferris BJ. Epidemiology standardization project. Am Rev Respir Dis 1978;118(6 Part 2):1120.[ISI][Medline]
32 Kuller L, Ockene J, Townsend M, Browner W, Meilahn E, Wentworth D. The epidemiology of pulmonary function and COPD mortality in the Multiple Risk Factor Intervention Trial. Am Rev Respir Dis 1989; 140:S76S81.[ISI][Medline]
33 International Classification of Diseases, Ninth Revision Clinical Modification, Vol. 1. Ann Arbor, MI: Edwards Brothers, 1981.
34 Multiple Risk Factor Intervention Trial Research Group. Multiple Risk Factor Intervention Trial: risk factor changes and mortality results. JAMA 1982;248:146577.[Abstract]
35 International Classification of Diseases, Ninth Revision, North American Clinical Modification. Vol. 1. Ann Arbor, MI: Edwards Brothers, Inc., 1981.
36 Multiple Risk Factor Intervention Trial Research Group. Mortality after 16 years for participants randomized to the Multiple Risk Factor Intervention Trial. Circulation 1996;94:94651.
37 Horm J. Assignment of Probabilistic Scores to National Death Index Record Matches. Hyattsville, MD: National Center for Health Statistics, 1996.
38 Wentworth D, Neaton JD, Rasmussen W. An evaluation of the Social Security Administration Master Beneficiary Record Index File and the National Death Index in the ascertainment of vital status. Am J Public Health 1982;73:127074.[ISI]
39 World Health Organization. International Classification of Diseases, Tenth Revision. Geneva: World Health Organization, 1992.
40 Cox DR. Regression models and life tables with discussion. J Royal Stat SocB. 1972;34:187220.
41 Nomura A, Stemmermann GN, Chyou PH, Marcus EB, Buist AS. Prospective study of pulmonary function and lung cancer. Am Rev Respir Dis 1991;144:30711.[ISI][Medline]
42 Van den Eeden SK, Friedman GD. Forced expiratory volume (1 second) and lung cancer incidence and mortality. Epidemiology 1992;3:25357.[ISI][Medline]
43 Islam SS, Schottenfeld D. Declining FEV1 and chronic productive cough in cigarette smokers: a 25-year prospective study of lung cancer incidence in Tecumseh, Michigan. Cancer Epidemiol Biomarkers Prev 1994;3:28998.[Abstract]
44 Cohen BH. Hypothesis: Is pulmonary dysfunction the common denominator for the multiple effects of cigarette smoking? Lancet 1978;ii:102427.