- Email: [email protected]
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women
Lung Cancer (2005) 47, 165—172
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women J.O. Ebberta, B.A. Williamsb, Z. Sunc, M.C. Aubryd, J.A. Wampflerb, Y.I. Garcese, R.L. Meyerc, P. Yangc,∗ a
Division of Primary Care Internal Medicine, Department of Internal Medicine, Nicotine Research Program, Mayo Clinic College of Medicine, Rochester, MN 55905, USA b Division of Biostatistics, Mayo Clinic College of Medicine, Rochester, MN 55905, USA c Division of Epidemiology, Mayo Clinic College of Medicine, Rochester, MN 55905, USA d Division of Anatomic Pathology, Mayo Clinic College of Medicine, Rochester, MN 55905, USA e Division of Radiation Oncology, Mayo Clinic College of Medicine, Rochester, MN 55905, USA Received 6 April 2004 ; received in revised form 28 June 2004; accepted 2 July 2004 KEYWORDS
Summary
Lung neoplasms; Prognosis; Smoking cessation; Carcinoma; Small-cell; Carcinoma; Non-small-cell lung; Sex
Background: Previous studies have attempted to investigate the impact of smoking cessation on lung cancer survival but have been limited by small numbers of former smokers and incomplete data. Methods: Over a six-year period, 5229 patients with non-small-cell lung cancer (NSCLC) and small-cell lung cancer (SCLC) were enrolled in a prospective cohort of whom 2052 were former smokers. Patient’s characteristics were obtained from medical records and a baseline interview. Vital status was determined through multiple sources. Cox proportional hazards models were used to estimate the effect of smoking abstinence on post-diagnosis mortality. Results: For all patients with NSCLC, the median survival among never, former, and current smokers was 1.4 years, 1.3 years, and 1.1 years, respectively (P < 0.01). Female NSCLC patients had a significantly lower risk of mortality with a longer duration of smoking abstinence (RR per 10 years of smoking abstinence = 0.85; 95% CI: 0.75, 0.97). No effect of smoking abstinence on mortality was observed for women with SCLC or for men with either histologic group. Conclusions: The identification of smoking history as a prognostic factor in lung cancer survival supports previous research suggesting a direct biologic effect of smoking on survival. However, this effect may vary by sex and type of lung cancer. © 2004 Elsevier Ireland Ltd. All rights reserved.
∗ Corresponding author. Present address: Department of Health Sciences and Cancer Center at Mayo Clinic College of Medicine, 200 First Street SW, Rochester, MN 55905, USA. Tel.: +1 507 266 5369; fax: +1 507 266 2478. E-mail addresses: [email protected] (J.O. Ebbert), [email protected] (B.A. Williams), [email protected] (Z. Sun), [email protected] (M.C. Aubry), wampfl[email protected] (J.A. Wampfler), [email protected] (Y.I. Garces), [email protected] (R.L. Meyer), [email protected] (P. Yang).
0169-5002/$ — see front matter © 2004 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.lungcan.2004.07.045
166
1. Introduction In 2004, lung cancer is expected to account for an estimated 13% of cancer diagnoses and 28% of all cancer deaths in the United States [1]. The fiveyear survival rate for all stages of lung cancer is 15% [1], which has not changed significantly over the last several decades [2]. A personal history of cigarette smoking has been associated with decreased overall survival among patients receiving treatment for non-smallcell lung cancer (NSCLC) [3—8], small-cell lung cancer (SCLC) [9], and all cell types combined [10,11]. Several studies have attempted to assess the impact of smoking cessation on overall survival for NSCLC [7] and SCLC [12,13], as well as on disease-free survival after SCLC treatment [14—16]. However, the small numbers of former smokers in these studies have made it difficult to assess the impact of the duration of smoking abstinence prior to lung cancer diagnosis on survival after lung cancer diagnosis. A recent systematic review of the literature has suggested that smoking history is a promising host-related factor with a predictive value that has yet to be fully defined [17]. In order to better clarify the relationship between the duration of smoking abstinence and survival in patients with lung cancer, we analyzed a prospective cohort of 5229 patients with newly diagnosed primary NSCLC and SCLC with a focus on 2052 former smokers.
2. Methods 2.1. Study cohort Only patients providing informed consent as approved by the Mayo Clinic Institutional Review Board were included in this study. Between, 1 January 1997 and 31 December 2002, 5229 patients with confirmed NSCLC and SCLC at Mayo Clinic, Rochester, Minnesota, were prospectively enrolled and actively followed. New patients were ascertained daily from a computerized pathology reporting system that identifies all lung cancer cases that received care in our institution. All pathologic diagnoses were made or confirmed histologically without knowing the patients’ smoking history and sex. A pulmonary pathologist (MCA) performed an independent review to adjudicate inconsistencies and discrepancies [18]. Information abstracted from medical records for each patient during a baseline visit at the
J.O. Ebbert et al. time of diagnosis included demographics (age, sex, and race), history of tobacco exposure, comorbidities, lung cancer pathology, anatomical site, and treatment. Clinical stage was assigned by results from available chest radiography, computerized tomography, bone scans, PET scans, and magnetic resonance imaging. When the patient received any type of diagnostic procedure and therapy elsewhere, authorization for the release of medical information was requested and copies of the relevant medical records were abstracted. Once the records were obtained, a Mayo clinician assigned the disease stage, and the information was gathered according to the criteria for histology, staging [19,20], and treatment. Tobacco use history included age of regular smoking initiation, average cigarettes smoked per day, total years smoked, and other types of tobacco products used. Never smokers were defined by selfreport as having smoked fewer than 100 cigarettes per lifetime. Former smokers were defined as reporting at least six months of smoking abstinence at the time of diagnosis. Current smokers were daily cigarette smokers or those with less than six months of smoking abstinence at the time of diagnosis.
2.2. Cohort follow-up Vital status of each patient was verified through the Mayo Clinic registration database, next-ofkin reports, death certificates and obituary documents, the Mayo Clinic Tumor Registry, and the Social Security Death Index website. For all patients, clinical status information was obtained from the most recent clinical note or the last mailed study questionnaire. The mailed study questionnaire obtained information on new diseases and treatment occurring after the initial diagnosis, an updated family history of lung or other cancers, and changes in smoking status, body weight, appetite, and dietary supplements. We also obtained detailed tobacco use history for patients with incomplete information in the medical records [21,22]. For deceased patients, the follow-up packet was sent to the next-of-kin to obtain proxy information. In January of each year between 1998 and 2003, vital status of every patient coded as alive during the previous years in our database was confirmed. Among former smoking NSCLC patients, 57.3% died, with survivors followed for a mean of 1.9 years. Among former smoking SCLC patients, 70.6% died, with survivors followed for a mean of 1.5 years. Living patients were censored at the date of last follow-up.
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women
2.3. Statistical analysis Survival following lung cancer diagnosis was analyzed. Survival was defined as the time in days from lung cancer diagnosis to death or the last known date the patient was reported to be alive. The primary independent variable was number of years having quit smoking prior to lung cancer diagnosis (as a continuous variable). Smoothing splines from a Cox proportional hazards model were fit for the entire cohort and separately by sex to determine the proper functional form of years since quitting on post-diagnosis mortality. All splines displayed relatively linear, monotonic relationships between years since quitting and mortality, hence years since quitting was analyzed as a continuous variable in subsequent analyses. A Cox proportional hazards model was developed which included main effects for years since quitting and sex plus an interaction term between years since quitting and sex. This model facilitated estimating relative risks for years since quitting by sex and determining whether the effect of years since quitting on mortality differed by sex. Cox models were also developed for NSCLC and SCLC adjusting for known predictors of survival. For NSCLC patients, these variables were age at diagnosis, packs of cigarettes per day smoked before quitting, years smoked, histologic cell type, tumor grade, disease stage, and treatment. The model for SCLC patients retained the same variables except tumor histology and grade. Hazard ratios or relative risks (RR) by sex were estimated from the adjusted model, and the interaction term between sex and years since quitting was tested for statistical significance. The analysis described was preceded by an analysis comparing survival across never, former, and current smokers among NSCLC and SCLC patients, respectively. Median survival was estimated within each smoking group as the time point in years where the Kaplan—Meier survival function equaled 50%. Differences in survival across groups were tested using log-rank tests. To further illustrate the effect of years since quitting on post-diagnosis mortality by sex, estimated survival from the adjusted Cox proportional hazards model was plotted over follow-up time across four different intervals of years since quitting for NSCLC (≤10, 11—20, 21—30, and >30 years) and for SCLC (<5, 5—10, 11—19, and ≥20 years). In estimating these survival rates by duration of smoking abstinence, all adjuster variables in the Cox model were fixed at their average value.
167
All baseline characteristics were reported across the aforementioned duration of abstinence groups for NSCLC and SCLC. To evaluate potential bias due to missing data, we compared survival, disease stage, and demographic characteristics between those with and without a smoking quit date. Testing for significant differences was completed using Chi-square tests for categorical variables and Wilcoxon rank sum tests for continuous variables. Categorical variables are reported as counts and percentages and continuous variables as medians and inter-quartile ranges. All analyses were completed using SAS v 6.12. Two-sided P-values of <0.05 were considered statistically significant.
3. Results The inception cohort in this study was comprised of 5229 primary lung cancer patients with 4618 NSCLC [23] and 611 SCLC cases. The overall median survival was 1.1 years, and the median survival was 1.4 years for NSCLC and 0.9 years for SCLC. Among the 4577 patients with known smoking status, 513 (11%) were never smokers, 2637 (58%) were former smokers, and 1427 (31%) were current smokers. The median survival for NSCLC never, former, and current smokers was 1.4 years, 1.3 years, and 1.1 years, respectively. Differences in survival across the three groups were statistically significant (P < 0.01). Using an adjusted Cox proportional hazards model, the relative risk (RR) of post-diagnosis mortality among current compared to never smokers was 1.15 (95% CI: 1.00, 1.32), and for former compared to never smokers was 0.99 (95% CI: 0.86, 1.13). For SCLC, there were no significant differences in survival among never, former, and current smokers [current versus never, adjusted RR 1.08 (95% CI: 0.58, 1.99); former versus never, adjusted RR 1.05 (95% CI: 0.57, 1.94)]. Among the former smokers, 2052 cases (1851 NSCLC; 201 SCLC) reported year smoking abstinence began so duration of smoking abstinence prior to diagnosis could be derived. Among the 2052 former smokers, 1342 (65%) were men and 710 (35%) were women. For both NSCLC and SCLC patients, increasing duration of smoking abstinence was associated with an older age of lung cancer diagnosis and fewer years smoked (Table 1). For NSCLC, a larger proportion of men and fewer packs per day (PPD) [≤1 PPD] were associated with an increasing duration of smoking abstinence. Further, an increasing duration of smoking abstinence was associated with a larger proportion of adenocarcinomas (P < 0.01) and well-
168
J.O. Ebbert et al.
Table 1
Patient demographics in a lung cancer inception cohort with 2052 former smokers
NSCLC (N = 1851)
Duration of smoking abstinence ≤10 (n = 765) 11—20 (n = 526)
Age at diagnosis (IQRa ) 68 (61, 74) Male 455, 59.5% Packs per day (PPD) PPD ≤ 1 334, 48.0% 1 < PPD ≤ 2 303, 43.5% PPD > 2 59, 8.5% Unknown 69 No. of years smoked (IQR) 43 (35, 50) Histology Adenocarcinoma 355, 46.5% Squamous 253, 33.1% Other NSCLC 156, 20.4% Grade Well-differentiated 79, 10.6% Moderately differentiated 328, 43.9% Poorly/undifferentiated 247, 33.0% Non-gradable 94, 12.6% Unknown 17 Stage I 221, 29.3% II 75, 10.0% III 218, 29.0% IV 239, 31.7% Unknown 12 Treatment within six months of diagnosis Chemotherapy 189, 30.9% Radiation 185, 30.3% Surgery 333, 54.5%
21—30 (n = 336)
>30 (n = 224)
P-value
69 (63, 75) 357, 67.9%
73 (67, 78) 239, 71.1%
75 (69, 79) 173, 77.2%
<0.01 <0.01
228, 48.0% 193, 40.6% 54, 11.4% 51 35 (30, 42)
166, 55.0% 112, 37.1% 24, 7.9% 34 30 (21, 35)
127, 65.8% 57, 29.1% 12, 6.1% 28 20 (14, 25)
<0.01
302, 57.4% 152, 28.9% 72, 13.7%
199, 59.2% 83, 24.7% 54, 16.1%
136, 60.7% 46, 20.5% 42, 18.8%
<0.01
78, 15.2% 233, 45.4% 142, 27.7% 60, 11.7% 13
50, 15.4% 143, 44.1% 88, 27.2% 43, 13.3% 12
51, 88, 55, 25, 5
23.3% 40.2% 25.1% 11.4%
<0.01
152, 29.3% 47, 9.1% 139, 26.8 180, 34.7% 8
102, 30.5% 26, 7.8% 96, 28.7% 110, 32.9% 2
78, 21, 53, 70, 2
35.1% 9.5% 23.9% 31.5%
0.53
134, 31.7% 128, 30.3% 232, 54.8%
73, 27.1% 74, 27.5% 156, 58.0%
43, 24.7% 40, 23.0% 107, 61.5%
<0.01
0.24 0.24 0.34
SCLC (N = 201)
Age at diagnosis (IQR) Male Packs per day (PPD) PPD ≤ 1 1 < PPD ≤ 2 PPD > 2 Unknown No. of years smoked (IQR) Stage Extensive Limited a
Duration of smoking abstinence <5 (n = 68) 5—10 (n = 41)
11—19 (n = 53)
20+ (n = 39)
P-value
65 (59, 71) 33, 48.5%
71 (63, 75) 24, 58.5%
72 (67, 77) 33, 62.3%
78 (70, 82) 28, 71.8%
<0.01 0.12
34, 51.5% 26, 39.4% 6, 9.1% 2 45 (40, 50)
13, 34.2% 20, 52.6% 5, 13.2% 3 48 (37, 50)
21, 43.8% 17, 35.4% 10, 20.8% 5 39 (32, 44)
20, 60.6% 10, 30.3% 3, 9.1% 6 31 (25, 40)
0.13
32, 48.5% 34, 51.5%
21, 51.2% 20, 48.8%
31, 58.5% 22, 41.5%
21, 55.3% 17, 44.7%
<0.01 0.73
IQR = Interquartile range.
differentiated tumor types (P < 0.01). Among never smokers who developed NSCLC, 74.7% had adenocarcinoma and 23.7% had well-differentiated tumors. Among patients with SCLC, there were 15 never smokers, all of whom reported environmental tobacco smoke exposure [22]. No significant differences were observed between the numbers of
patients receiving chemotherapy (P = 0.07) or radiation therapy (P = 0.17) within six months of diagnosis across duration of abstinence. Among the NSCLC patients, the adjusted relative risk (RR) of post-diagnosis mortality by years of smoking abstinence was significant for women, but not for men (Table 2). A significant interaction
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women
169
Table 2 Relative risks of post-diagnosis mortality by years of cigarette smoking abstinence in a lung cancer inception cohort with 2052 former smokers NSCLC Relative risk 95% CI P-value
a
Adjusted relative riska,b 95% CI P-value
Females
Males
All patients
0.94 (0.85, 1.04) 0.23
1.05 (0.99, 1.11) 0.12
1.03 (0.97, 1.08) 0.33
0.85 (0.75, 0.97) 0.01
1.00 (0.91, 1.10) 0.98
0.96 (0.88, 1.04) 0.30
SCLC Relative risk 95% CI P-value
c
Adjusted relative riskc,d 95% CI P-value a b c d
1.15 (0.99, 1.34) 0.07
1.14 (1.04, 1.26) <0.01
1.16 (1.07, 1.26) <0.01
1.04 (0.84, 1.28) 0.80
1.02 (0.89, 1.16) 0.74
1.04 (0.84, 1.28) 0.74
Per 10 years quitting smoking. Adjusted for age at diagnosis, packs per day smoked, years smoked, histology, tumor grade, stage, and treatment. Per five years quitting smoking. Adjusted for age at diagnosis, packs per day smoked, years smoked, stage, and treatment.
between sex and years of smoking abstinence on post-diagnosis mortality was observed in the adjusted model (P = 0.01). Women had significantly lower risk of post-diagnosis mortality with a longer duration of smoking abstinence (adjusted RR per 10 years of smoking abstinence = 0.85; 95% CI: 0.75, 0.97). Although a 10-year interval was chosen to illustrate the effect on survival, our results suggest the benefit consistently increases with the number of years since smoking abstinence. There was not a point of the continuous duration variable at which the risk drops dramatically. The magnitude of this effect in women was seen in all cell types except squamous cell carcinoma
(data not shown). Survival over five years of followup for women by duration of abstinence group is shown in Fig. 1. The sex duration of abstinence relationship was not significantly different across disease stages; however, a notably stronger effect was seen in women with stage IV lung cancer [RR per 10 years of abstinence = 0.76 (95% CI: 0.64—0.91)] than in other stages. For men, the RR per 10 years of abstinence was not significantly different from one, with no differences in risk across histologic subtypes. No effect of smoking abstinence on the risk of post-diagnosis mortality was observed in either men or women with SCLC, after adjusting for other known predictors of survival.
Fig. 1. Survival by years since abstinence smoking for women (adjusted for age, packs per day, years smoked, histology, grade, stage, treatment).
170 Non-small-cell lung cancer (NSCLC) former smokers with missing data on duration of smoking abstinence (n = 195) had worse survival following lung cancer diagnosis than patients with complete data on duration of abstinence (P < 0.01; estimated oneyear survival, 35% versus 59%). This survival disadvantage could be partly attributed to a significantly higher proportion of late stage cancer (P < 0.01) and poorly differentiated among patients with missing data (P < 0.01). In addition, these patients smoked fewer years and were younger at diagnosis (P’s < 0.05). There were no differences in patients with complete or missing data with respect to sex, packs of cigarettes per day, and histologic cell type.
4. Discussion In this study, we observed that for NSCLC patients, never smokers survived longer than current smokers after lung cancer diagnosis. This relationship was not observed among the SCLC patients. After adjusting for other known predictors of post-diagnosis mortality, female former smokers with NSCLC experienced a 15% risk reduction for mortality per 10 years of smoking abstinence. This relationship was consistent across stages, but strongest in women with stage IV NSCLC. For every 10 years of abstinence, female former smokers with stage IV NSCLC experienced a 24% reduction in post-diagnosis mortality. Duration of smoking abstinence was not a predictor of mortality for men with lung cancer or women with SCLC in our study. Our observation of an effect of smoking status on survival for NSCLC is consistent with previous reports in the literature. In a study of 587 patients undergoing surgical resection for adenocarcinoma, current smokers (smoking within one year of admission) were 2.38 times more likely to die following surgical resection compared to nonsmokers [7]. In a study of 30 male and female patients who underwent curative resection for stage I adenocarcinoma, smokers had a relative risk of 5.87 for mortality following surgical resection compared to non-smokers [24]. In a retrospective study of 1635 patients with NSCLC treated in Brazil, smoking was associated with worse survival at one and two years after treatment [5]. In a study of 223 women with primary lung cancer, smokers with adenocarcinoma had a RR of 2.1 (P = 0.057) for dying within 12 months of diagnosis compared to never smokers with adenocarcinoma adjusting for age, tumor stage, and treatment [10]. Our observation was that the RR of post-diagnosis death was higher among current smokers with NSCLC compared to
J.O. Ebbert et al. never smokers after multivariable adjustment (RR 1.15; 95% CI: 1.00, 1.32) supports the hypothesis that smoking status prior to lung cancer diagnosis may be an important prognostic variable. Our observation of a differential effect of smoking history on lung cancer mortality by sex is also consistent with previously published reports in the literature. In a case-control study of 463 men and 212 women with histologically confirmed lung cancer (NSCLC and SCLC), the risk of death for ever smokers was significantly higher than for never smokers for the entire sample of cases (RR 1.4; 95% CI: 1.1, 1.8), but a differential effect of survival by smoking history was seen between men and women [25]. The risk of death was higher for female current smokers (RR 1.9; 95% CI: 1.2—3.1) and past smokers (RR 1.4; 95% CI: 0.8, 2.3) compared to never smokers, but no effect of smoking on survival was observed in men. We observed that women had a significantly decreased risk of mortality for each 10 years of smoking abstinence (RR 0.85; 95% CI: 0.75, 0.97), and this relationship was not observed in men. This finding supports the hypothesis of previous investigators that the effect of cigarette smoking on survival is heterogeneous in nature [25]. We did not observe a significant difference in overall survival by smoking history in patients with SCLC, which is inconsistent with several studies in the medical literature. In a multicenter study of 766 SCLC patients participating in three German multicenter trials, non-smokers had a significantly longer median survival (13.6 months) compared to smokers (9.9 months) [9]. However, this study included patients enrolled in chemotherapy—radiotherapy clinical trials, and the patients may have had a worse prognosis than the patients in our cohort. Evidence for this speculation is supported by a longer median survival in our cohort (12.0 months) compared to the German clinical trial cohort (10.6 months) and a lower proportion of extensive stage patients [105/201 (52%) versus 463/757 (61%), respectively]. One possible explanation is that smoking status may be an important prognostic variable for patients with more advanced disease than among all SCLC patients. A few prior studies have attempted to assess the impact of smoking cessation on overall survival with SCLC [13]. In a study of 112 SCLC patients receiving chemotherapy, a significant trend was observed suggesting that patients who quit smoking prior (NSPrior) to diagnosis had the best survival, patients who quit at the time of diagnosis (NS-Dx) had intermediate survival, and patients who continued to smoke (S) had the worst survival [13]. Survival differences remained after adjusting for disease
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women extent, performance status, and type of protocol treatment. However, this effect was primarily attributable to the NS-Prior patients, since no statistically significant difference was observed between the NS-Dx and S patients. We did not observe this relationship in our 201 SCLC patients stratifying by sex and adjusting for age at diagnosis, packs per day, years smoked, disease stage, and treatment. One possible explanation is that Johnston-Early’s study did not adjust for other significant confounding factors, as we did in our study. Another possible explanation is that SCLC patients generally progress and die quickly after diagnosis and the small impact of duration of smoking abstinence on survival is not detectable. There are several limitations to this study. First of all, we assessed all-cause mortality rather than cause-specific mortality due to limitations in obtaining precise cause of death information for some patients. The patients are comprised of those in the area surrounding the Mayo Clinic (Olmsted County) and those from outside that area. For 98% of 138 patients residing in Olmsted County at the time of lung cancer diagnosis who died, we were able to identify causes of death. Death from lung cancer occurred in 100% of never smokers (n = 8), 94% of former smokers (n = 70), and 94% of current smokers (n = 60). For patients outside of Olmsted County, we did not have complete data for cause-specific mortality. Because survival after lung cancer diagnosis is often short, assuming most deaths were due to lung cancer may be reasonable. Furthermore, previous investigators have proposed that all-cause mortality may be a more appropriate analysis for assessing the impact of smoking than cancerspecific or competing causes of death alone [8]. A second limitation is potential selection bias, either due to referral practice or self-selected participation in our study. We attempted to assess selection bias by comparing former smokers with complete smoking information with those who had missing data (9.5%). We observed a shorter duration of smoking, a higher proportion of advanced disease stage at diagnosis, and shorter survival in patients with missing data compared to patients with complete data. Furthermore, the baseline characteristics of patients with missing data were compared to the characteristics of patients with complete data across duration of abstinence groups. Patients with missing data were characterized by a median age of 69 years; 61.5% were male; 51.8% smoking one pack per day or less; smoked a median of 30 years; 46.9% had adenocarcinoma; 29.9% had squamous cell carcinoma; and 7.7% had well-differentiated tumors. These features are more consistent with the characteristics of patients having a shorter duration of
171
smoking abstinence than those with longer durations. In light of the increased mortality found in patients with missing data, one could surmise that the indirect relationship between duration of smoking abstinence and mortality could have been accentuated had data been available on all patients. In addition we do not have data on performance status for all patients, particularly the early-stage patients at the time of diagnosis. If smoking status was correlated with performance status, then performance status may explain the observed differences in survival. However, in a previous study of SCLC, no association between performance status and smoking status was found [13]. The third limitation is the lack of diversity in our patient population with 96% being Caucasians of a European origin. We re-analyzed the data using only Caucasian patients and found almost identical results on the association between NSCLC survival in women and duration of smoking cessation compared to all patients combined. Another limitation of this study is that we could not control for comorbid illness. However, in a cohort study of 1155 patients with lung cancer, current smoking at diagnosis was found to be an important predictor of shortened lung cancer survival that was independent of comorbidity [8]. In conclusion, we observed a difference in survival by smoking status for NSCLC. We also observed a difference in survival between women and men in relation to duration of smoking abstinence prior to lung cancer diagnosis, which may reflect the complex interplay between biological and behavioral differences [26]. Our finding that duration of smoking abstinence predicts survival in women with NSCLC supports previous research suggesting that there is a direct biological effect of smoking on lung cancer survival [8], and our findings suggest that this effect may differ by sex.
Acknowledgements We would like to thank Drs. Randolph S. Marks, Scott H. Okuno, and Claude Deschamps for their contributions to this work. We would like to thank Susan Ernst for her technical assistance with the manuscript.
References [1] American Cancer Society, Cancer Facts and Figures, 2004. Available at http://www.cancer.org/docroot/STT/ stt O.asp. Accessed August 16, 2004.
172 [2] Ginsberg RJ, Vokes EE, Raben A. Cancer of the lung. In: DeVita VT, Hellmans S, Rosenberg SA, editors. Principles and Practice of Oncology. Philadelphia: Raven; 1999. p. 849—950. [3] Anonymous. Factors influencing long term survival after resection of bronchogenic carcinoma; an analysis of 1,118 cases. Zhonghua, Yi Xue Za Zhi, Chi Med J 1974;191(11):682—686. [4] Fujisawa T, Iizasa T, Saitoh Y, Sekine Y, et al. Smoking before surgery predicts poor long-term survival in patients with stage I non-small-cell lung carcinomas. J Clin Oncol 1999;17(7):2086—91. [5] Martins SJ, Pereira JR. Clinical factors and progrnosis in non-small-cell lung cancer. Am J Clin Oncol 1999;22:453—7. [6] Sioris T, Husgafvel-Pursianen K, Karjalainen A, Anttila S, et al. Survival in operable non-small-cell lung cancer: role of p53 mutations, tobacco smoking and asbestos exposure. Int J Cancer 2000;86:590—4. [7] Sobue T, Suzuki T, Fujimoto I, Doi O, et al. Prognostic factors for surgically treated lung adenocarcinoma patients, with special reference to smoking habit. Jpn J Cancer Res 1991;82:33—9. [8] Tammemagi DVM, McLaughlin JR, Brendan J, Mullen M, et al. A study of smoking, p53 tumor suppressor gene alterations and non-small-cell lung cancer. Ann Epidemiol 2000;10(3):176—85. [9] Wolf M, Holle R, Hans KPD, et al. Analysis of prognostic factors in 766 patients with small-cell lung cancer (SCLC): the role of sex as a predictor for survival. Br J Cancer 1991;63:986—92. [10] Hinds MW, Yang HY, Stemmermann G, Lee JJ, et al. Smoking history and lung cancer survival in women. J Natl Cancer Inst 1982;68(3):395—9. [11] Xavier F, Henn LdA, Oliveira M, Orlandine L. Smoking and its relation to the histological type, survival, and prognosis among patients with primary lung cancer. Revista Paulista De Medicina 1996;114:1298—302. [12] Bergman B, Sorenson S. Smoking and effect of chemotherapy in small-cell lung cancer. Eur Respir J 1988;1:932—7. [13] Johnston-Early A, Cohen MH, Minna JD, Paxton LM, et al. Smoking abstinence and small-cell lung cancer survival. an association. JAMA 1980;244:2175—9. [14] Kawahara M, Ushijima S, Kamimori T, Kodama N, et al. Second primary tumours in more than two-year disease-free
J.O. Ebbert et al.
[15]
[16]
[17] [18]
[19] [20] [21]
[22] [23]
[24]
[25] [26]
survivors of small-cell lung cancer in Japan: the role of smoking cessation. Br J Cancer 1998;78:409—12. Richardson GE, Tucker MA, Venzon DJ, Linnoila RI, et al. Smoking cessation after successful treatment of smallcell lung cancer is associated with fewer smoking-related second primary cancers. Ann Intern Med 1993;119:383— 90. Tucker MA, Murray N, Shaw EG, Ettinger DS, et al. Second primary cancers related to smoking and treatment of small-cell lung cancer. J Natl Cancer Inst 1997;89(23): 1782—8. Brundage MD, Davies D, Mackillop WJ. Prognostic factors in non-small-cell lung cancer. a decade of progress. Chest 2002;122(3):1037—57. World Health Organization. Histological typing of lung and pleural tumors. World Health Organization’s International Histological Classification of Tumors. 3rd ed. Berlin, Heidelberg: Springer Verlag; 1999. Greene FL. American cancer society and American Joint Committee on cancer. AJCC Cancer Staging Manual. 6th ed. New York: Springer-Verlag; 2002. Mountain CF. Revisions in the international system for staging lung cancer. Chest 1997;111:1710—7. Yang P, Wentzlaff KA, Katzmann JA, Allen MS, et al. Alpha1-antitrypsin deficiency allele carriers in lung cancer patients. Cancer Epidemiol Biomarkers Prev 1999;8: 461—5. de Andrade M, Ebbert JO, Wampfler JA, Miller DL, et al. Environmental tobacco smoke exposure in women with lung cancer. Lung Cancer 2004;43:127—34. Visbal AL, Williams BA, Nichols FC, Marks RS, et al. Gender differences in non-small-cell lung cancer survival: an analysis of 4618 patients diagnosed between 1997—2002. Ann Thorac Surg 2004;78(1):209—15. Isobe T, Hiyama K, Yoshid Y, Fujiwara Y, et al. Prognostic significance of p53 and ras gene abnormalities in lung adenocarcinoma patients with stage I disease after curative resection. Jpn J Cancer Res 1994;85:1240—6. Goodman MT, Kolonel LN, Wilkens LR, Yoshizawa CN, et al. Smoking history and survival among lung cancer patients. Cancer Causes Control 1990;1:155—63. Payne S. Smoke like a man, die like a man?: a review of the relationship between gender, sex and lung cancer. Soc Sci Med 2001;53:1067—80.
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women J.O. Ebberta, B.A. Williamsb, Z. Sunc, M.C. Aubryd, J.A. Wampflerb, Y.I. Garcese, R.L. Meyerc, P. Yangc,∗ a
Division of Primary Care Internal Medicine, Department of Internal Medicine, Nicotine Research Program, Mayo Clinic College of Medicine, Rochester, MN 55905, USA b Division of Biostatistics, Mayo Clinic College of Medicine, Rochester, MN 55905, USA c Division of Epidemiology, Mayo Clinic College of Medicine, Rochester, MN 55905, USA d Division of Anatomic Pathology, Mayo Clinic College of Medicine, Rochester, MN 55905, USA e Division of Radiation Oncology, Mayo Clinic College of Medicine, Rochester, MN 55905, USA Received 6 April 2004 ; received in revised form 28 June 2004; accepted 2 July 2004 KEYWORDS
Summary
Lung neoplasms; Prognosis; Smoking cessation; Carcinoma; Small-cell; Carcinoma; Non-small-cell lung; Sex
Background: Previous studies have attempted to investigate the impact of smoking cessation on lung cancer survival but have been limited by small numbers of former smokers and incomplete data. Methods: Over a six-year period, 5229 patients with non-small-cell lung cancer (NSCLC) and small-cell lung cancer (SCLC) were enrolled in a prospective cohort of whom 2052 were former smokers. Patient’s characteristics were obtained from medical records and a baseline interview. Vital status was determined through multiple sources. Cox proportional hazards models were used to estimate the effect of smoking abstinence on post-diagnosis mortality. Results: For all patients with NSCLC, the median survival among never, former, and current smokers was 1.4 years, 1.3 years, and 1.1 years, respectively (P < 0.01). Female NSCLC patients had a significantly lower risk of mortality with a longer duration of smoking abstinence (RR per 10 years of smoking abstinence = 0.85; 95% CI: 0.75, 0.97). No effect of smoking abstinence on mortality was observed for women with SCLC or for men with either histologic group. Conclusions: The identification of smoking history as a prognostic factor in lung cancer survival supports previous research suggesting a direct biologic effect of smoking on survival. However, this effect may vary by sex and type of lung cancer. © 2004 Elsevier Ireland Ltd. All rights reserved.
∗ Corresponding author. Present address: Department of Health Sciences and Cancer Center at Mayo Clinic College of Medicine, 200 First Street SW, Rochester, MN 55905, USA. Tel.: +1 507 266 5369; fax: +1 507 266 2478. E-mail addresses: [email protected] (J.O. Ebbert), [email protected] (B.A. Williams), [email protected] (Z. Sun), [email protected] (M.C. Aubry), wampfl[email protected] (J.A. Wampfler), [email protected] (Y.I. Garces), [email protected] (R.L. Meyer), [email protected] (P. Yang).
0169-5002/$ — see front matter © 2004 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.lungcan.2004.07.045
166
1. Introduction In 2004, lung cancer is expected to account for an estimated 13% of cancer diagnoses and 28% of all cancer deaths in the United States [1]. The fiveyear survival rate for all stages of lung cancer is 15% [1], which has not changed significantly over the last several decades [2]. A personal history of cigarette smoking has been associated with decreased overall survival among patients receiving treatment for non-smallcell lung cancer (NSCLC) [3—8], small-cell lung cancer (SCLC) [9], and all cell types combined [10,11]. Several studies have attempted to assess the impact of smoking cessation on overall survival for NSCLC [7] and SCLC [12,13], as well as on disease-free survival after SCLC treatment [14—16]. However, the small numbers of former smokers in these studies have made it difficult to assess the impact of the duration of smoking abstinence prior to lung cancer diagnosis on survival after lung cancer diagnosis. A recent systematic review of the literature has suggested that smoking history is a promising host-related factor with a predictive value that has yet to be fully defined [17]. In order to better clarify the relationship between the duration of smoking abstinence and survival in patients with lung cancer, we analyzed a prospective cohort of 5229 patients with newly diagnosed primary NSCLC and SCLC with a focus on 2052 former smokers.
2. Methods 2.1. Study cohort Only patients providing informed consent as approved by the Mayo Clinic Institutional Review Board were included in this study. Between, 1 January 1997 and 31 December 2002, 5229 patients with confirmed NSCLC and SCLC at Mayo Clinic, Rochester, Minnesota, were prospectively enrolled and actively followed. New patients were ascertained daily from a computerized pathology reporting system that identifies all lung cancer cases that received care in our institution. All pathologic diagnoses were made or confirmed histologically without knowing the patients’ smoking history and sex. A pulmonary pathologist (MCA) performed an independent review to adjudicate inconsistencies and discrepancies [18]. Information abstracted from medical records for each patient during a baseline visit at the
J.O. Ebbert et al. time of diagnosis included demographics (age, sex, and race), history of tobacco exposure, comorbidities, lung cancer pathology, anatomical site, and treatment. Clinical stage was assigned by results from available chest radiography, computerized tomography, bone scans, PET scans, and magnetic resonance imaging. When the patient received any type of diagnostic procedure and therapy elsewhere, authorization for the release of medical information was requested and copies of the relevant medical records were abstracted. Once the records were obtained, a Mayo clinician assigned the disease stage, and the information was gathered according to the criteria for histology, staging [19,20], and treatment. Tobacco use history included age of regular smoking initiation, average cigarettes smoked per day, total years smoked, and other types of tobacco products used. Never smokers were defined by selfreport as having smoked fewer than 100 cigarettes per lifetime. Former smokers were defined as reporting at least six months of smoking abstinence at the time of diagnosis. Current smokers were daily cigarette smokers or those with less than six months of smoking abstinence at the time of diagnosis.
2.2. Cohort follow-up Vital status of each patient was verified through the Mayo Clinic registration database, next-ofkin reports, death certificates and obituary documents, the Mayo Clinic Tumor Registry, and the Social Security Death Index website. For all patients, clinical status information was obtained from the most recent clinical note or the last mailed study questionnaire. The mailed study questionnaire obtained information on new diseases and treatment occurring after the initial diagnosis, an updated family history of lung or other cancers, and changes in smoking status, body weight, appetite, and dietary supplements. We also obtained detailed tobacco use history for patients with incomplete information in the medical records [21,22]. For deceased patients, the follow-up packet was sent to the next-of-kin to obtain proxy information. In January of each year between 1998 and 2003, vital status of every patient coded as alive during the previous years in our database was confirmed. Among former smoking NSCLC patients, 57.3% died, with survivors followed for a mean of 1.9 years. Among former smoking SCLC patients, 70.6% died, with survivors followed for a mean of 1.5 years. Living patients were censored at the date of last follow-up.
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women
2.3. Statistical analysis Survival following lung cancer diagnosis was analyzed. Survival was defined as the time in days from lung cancer diagnosis to death or the last known date the patient was reported to be alive. The primary independent variable was number of years having quit smoking prior to lung cancer diagnosis (as a continuous variable). Smoothing splines from a Cox proportional hazards model were fit for the entire cohort and separately by sex to determine the proper functional form of years since quitting on post-diagnosis mortality. All splines displayed relatively linear, monotonic relationships between years since quitting and mortality, hence years since quitting was analyzed as a continuous variable in subsequent analyses. A Cox proportional hazards model was developed which included main effects for years since quitting and sex plus an interaction term between years since quitting and sex. This model facilitated estimating relative risks for years since quitting by sex and determining whether the effect of years since quitting on mortality differed by sex. Cox models were also developed for NSCLC and SCLC adjusting for known predictors of survival. For NSCLC patients, these variables were age at diagnosis, packs of cigarettes per day smoked before quitting, years smoked, histologic cell type, tumor grade, disease stage, and treatment. The model for SCLC patients retained the same variables except tumor histology and grade. Hazard ratios or relative risks (RR) by sex were estimated from the adjusted model, and the interaction term between sex and years since quitting was tested for statistical significance. The analysis described was preceded by an analysis comparing survival across never, former, and current smokers among NSCLC and SCLC patients, respectively. Median survival was estimated within each smoking group as the time point in years where the Kaplan—Meier survival function equaled 50%. Differences in survival across groups were tested using log-rank tests. To further illustrate the effect of years since quitting on post-diagnosis mortality by sex, estimated survival from the adjusted Cox proportional hazards model was plotted over follow-up time across four different intervals of years since quitting for NSCLC (≤10, 11—20, 21—30, and >30 years) and for SCLC (<5, 5—10, 11—19, and ≥20 years). In estimating these survival rates by duration of smoking abstinence, all adjuster variables in the Cox model were fixed at their average value.
167
All baseline characteristics were reported across the aforementioned duration of abstinence groups for NSCLC and SCLC. To evaluate potential bias due to missing data, we compared survival, disease stage, and demographic characteristics between those with and without a smoking quit date. Testing for significant differences was completed using Chi-square tests for categorical variables and Wilcoxon rank sum tests for continuous variables. Categorical variables are reported as counts and percentages and continuous variables as medians and inter-quartile ranges. All analyses were completed using SAS v 6.12. Two-sided P-values of <0.05 were considered statistically significant.
3. Results The inception cohort in this study was comprised of 5229 primary lung cancer patients with 4618 NSCLC [23] and 611 SCLC cases. The overall median survival was 1.1 years, and the median survival was 1.4 years for NSCLC and 0.9 years for SCLC. Among the 4577 patients with known smoking status, 513 (11%) were never smokers, 2637 (58%) were former smokers, and 1427 (31%) were current smokers. The median survival for NSCLC never, former, and current smokers was 1.4 years, 1.3 years, and 1.1 years, respectively. Differences in survival across the three groups were statistically significant (P < 0.01). Using an adjusted Cox proportional hazards model, the relative risk (RR) of post-diagnosis mortality among current compared to never smokers was 1.15 (95% CI: 1.00, 1.32), and for former compared to never smokers was 0.99 (95% CI: 0.86, 1.13). For SCLC, there were no significant differences in survival among never, former, and current smokers [current versus never, adjusted RR 1.08 (95% CI: 0.58, 1.99); former versus never, adjusted RR 1.05 (95% CI: 0.57, 1.94)]. Among the former smokers, 2052 cases (1851 NSCLC; 201 SCLC) reported year smoking abstinence began so duration of smoking abstinence prior to diagnosis could be derived. Among the 2052 former smokers, 1342 (65%) were men and 710 (35%) were women. For both NSCLC and SCLC patients, increasing duration of smoking abstinence was associated with an older age of lung cancer diagnosis and fewer years smoked (Table 1). For NSCLC, a larger proportion of men and fewer packs per day (PPD) [≤1 PPD] were associated with an increasing duration of smoking abstinence. Further, an increasing duration of smoking abstinence was associated with a larger proportion of adenocarcinomas (P < 0.01) and well-
168
J.O. Ebbert et al.
Table 1
Patient demographics in a lung cancer inception cohort with 2052 former smokers
NSCLC (N = 1851)
Duration of smoking abstinence ≤10 (n = 765) 11—20 (n = 526)
Age at diagnosis (IQRa ) 68 (61, 74) Male 455, 59.5% Packs per day (PPD) PPD ≤ 1 334, 48.0% 1 < PPD ≤ 2 303, 43.5% PPD > 2 59, 8.5% Unknown 69 No. of years smoked (IQR) 43 (35, 50) Histology Adenocarcinoma 355, 46.5% Squamous 253, 33.1% Other NSCLC 156, 20.4% Grade Well-differentiated 79, 10.6% Moderately differentiated 328, 43.9% Poorly/undifferentiated 247, 33.0% Non-gradable 94, 12.6% Unknown 17 Stage I 221, 29.3% II 75, 10.0% III 218, 29.0% IV 239, 31.7% Unknown 12 Treatment within six months of diagnosis Chemotherapy 189, 30.9% Radiation 185, 30.3% Surgery 333, 54.5%
21—30 (n = 336)
>30 (n = 224)
P-value
69 (63, 75) 357, 67.9%
73 (67, 78) 239, 71.1%
75 (69, 79) 173, 77.2%
<0.01 <0.01
228, 48.0% 193, 40.6% 54, 11.4% 51 35 (30, 42)
166, 55.0% 112, 37.1% 24, 7.9% 34 30 (21, 35)
127, 65.8% 57, 29.1% 12, 6.1% 28 20 (14, 25)
<0.01
302, 57.4% 152, 28.9% 72, 13.7%
199, 59.2% 83, 24.7% 54, 16.1%
136, 60.7% 46, 20.5% 42, 18.8%
<0.01
78, 15.2% 233, 45.4% 142, 27.7% 60, 11.7% 13
50, 15.4% 143, 44.1% 88, 27.2% 43, 13.3% 12
51, 88, 55, 25, 5
23.3% 40.2% 25.1% 11.4%
<0.01
152, 29.3% 47, 9.1% 139, 26.8 180, 34.7% 8
102, 30.5% 26, 7.8% 96, 28.7% 110, 32.9% 2
78, 21, 53, 70, 2
35.1% 9.5% 23.9% 31.5%
0.53
134, 31.7% 128, 30.3% 232, 54.8%
73, 27.1% 74, 27.5% 156, 58.0%
43, 24.7% 40, 23.0% 107, 61.5%
<0.01
0.24 0.24 0.34
SCLC (N = 201)
Age at diagnosis (IQR) Male Packs per day (PPD) PPD ≤ 1 1 < PPD ≤ 2 PPD > 2 Unknown No. of years smoked (IQR) Stage Extensive Limited a
Duration of smoking abstinence <5 (n = 68) 5—10 (n = 41)
11—19 (n = 53)
20+ (n = 39)
P-value
65 (59, 71) 33, 48.5%
71 (63, 75) 24, 58.5%
72 (67, 77) 33, 62.3%
78 (70, 82) 28, 71.8%
<0.01 0.12
34, 51.5% 26, 39.4% 6, 9.1% 2 45 (40, 50)
13, 34.2% 20, 52.6% 5, 13.2% 3 48 (37, 50)
21, 43.8% 17, 35.4% 10, 20.8% 5 39 (32, 44)
20, 60.6% 10, 30.3% 3, 9.1% 6 31 (25, 40)
0.13
32, 48.5% 34, 51.5%
21, 51.2% 20, 48.8%
31, 58.5% 22, 41.5%
21, 55.3% 17, 44.7%
<0.01 0.73
IQR = Interquartile range.
differentiated tumor types (P < 0.01). Among never smokers who developed NSCLC, 74.7% had adenocarcinoma and 23.7% had well-differentiated tumors. Among patients with SCLC, there were 15 never smokers, all of whom reported environmental tobacco smoke exposure [22]. No significant differences were observed between the numbers of
patients receiving chemotherapy (P = 0.07) or radiation therapy (P = 0.17) within six months of diagnosis across duration of abstinence. Among the NSCLC patients, the adjusted relative risk (RR) of post-diagnosis mortality by years of smoking abstinence was significant for women, but not for men (Table 2). A significant interaction
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women
169
Table 2 Relative risks of post-diagnosis mortality by years of cigarette smoking abstinence in a lung cancer inception cohort with 2052 former smokers NSCLC Relative risk 95% CI P-value
a
Adjusted relative riska,b 95% CI P-value
Females
Males
All patients
0.94 (0.85, 1.04) 0.23
1.05 (0.99, 1.11) 0.12
1.03 (0.97, 1.08) 0.33
0.85 (0.75, 0.97) 0.01
1.00 (0.91, 1.10) 0.98
0.96 (0.88, 1.04) 0.30
SCLC Relative risk 95% CI P-value
c
Adjusted relative riskc,d 95% CI P-value a b c d
1.15 (0.99, 1.34) 0.07
1.14 (1.04, 1.26) <0.01
1.16 (1.07, 1.26) <0.01
1.04 (0.84, 1.28) 0.80
1.02 (0.89, 1.16) 0.74
1.04 (0.84, 1.28) 0.74
Per 10 years quitting smoking. Adjusted for age at diagnosis, packs per day smoked, years smoked, histology, tumor grade, stage, and treatment. Per five years quitting smoking. Adjusted for age at diagnosis, packs per day smoked, years smoked, stage, and treatment.
between sex and years of smoking abstinence on post-diagnosis mortality was observed in the adjusted model (P = 0.01). Women had significantly lower risk of post-diagnosis mortality with a longer duration of smoking abstinence (adjusted RR per 10 years of smoking abstinence = 0.85; 95% CI: 0.75, 0.97). Although a 10-year interval was chosen to illustrate the effect on survival, our results suggest the benefit consistently increases with the number of years since smoking abstinence. There was not a point of the continuous duration variable at which the risk drops dramatically. The magnitude of this effect in women was seen in all cell types except squamous cell carcinoma
(data not shown). Survival over five years of followup for women by duration of abstinence group is shown in Fig. 1. The sex duration of abstinence relationship was not significantly different across disease stages; however, a notably stronger effect was seen in women with stage IV lung cancer [RR per 10 years of abstinence = 0.76 (95% CI: 0.64—0.91)] than in other stages. For men, the RR per 10 years of abstinence was not significantly different from one, with no differences in risk across histologic subtypes. No effect of smoking abstinence on the risk of post-diagnosis mortality was observed in either men or women with SCLC, after adjusting for other known predictors of survival.
Fig. 1. Survival by years since abstinence smoking for women (adjusted for age, packs per day, years smoked, histology, grade, stage, treatment).
170 Non-small-cell lung cancer (NSCLC) former smokers with missing data on duration of smoking abstinence (n = 195) had worse survival following lung cancer diagnosis than patients with complete data on duration of abstinence (P < 0.01; estimated oneyear survival, 35% versus 59%). This survival disadvantage could be partly attributed to a significantly higher proportion of late stage cancer (P < 0.01) and poorly differentiated among patients with missing data (P < 0.01). In addition, these patients smoked fewer years and were younger at diagnosis (P’s < 0.05). There were no differences in patients with complete or missing data with respect to sex, packs of cigarettes per day, and histologic cell type.
4. Discussion In this study, we observed that for NSCLC patients, never smokers survived longer than current smokers after lung cancer diagnosis. This relationship was not observed among the SCLC patients. After adjusting for other known predictors of post-diagnosis mortality, female former smokers with NSCLC experienced a 15% risk reduction for mortality per 10 years of smoking abstinence. This relationship was consistent across stages, but strongest in women with stage IV NSCLC. For every 10 years of abstinence, female former smokers with stage IV NSCLC experienced a 24% reduction in post-diagnosis mortality. Duration of smoking abstinence was not a predictor of mortality for men with lung cancer or women with SCLC in our study. Our observation of an effect of smoking status on survival for NSCLC is consistent with previous reports in the literature. In a study of 587 patients undergoing surgical resection for adenocarcinoma, current smokers (smoking within one year of admission) were 2.38 times more likely to die following surgical resection compared to nonsmokers [7]. In a study of 30 male and female patients who underwent curative resection for stage I adenocarcinoma, smokers had a relative risk of 5.87 for mortality following surgical resection compared to non-smokers [24]. In a retrospective study of 1635 patients with NSCLC treated in Brazil, smoking was associated with worse survival at one and two years after treatment [5]. In a study of 223 women with primary lung cancer, smokers with adenocarcinoma had a RR of 2.1 (P = 0.057) for dying within 12 months of diagnosis compared to never smokers with adenocarcinoma adjusting for age, tumor stage, and treatment [10]. Our observation was that the RR of post-diagnosis death was higher among current smokers with NSCLC compared to
J.O. Ebbert et al. never smokers after multivariable adjustment (RR 1.15; 95% CI: 1.00, 1.32) supports the hypothesis that smoking status prior to lung cancer diagnosis may be an important prognostic variable. Our observation of a differential effect of smoking history on lung cancer mortality by sex is also consistent with previously published reports in the literature. In a case-control study of 463 men and 212 women with histologically confirmed lung cancer (NSCLC and SCLC), the risk of death for ever smokers was significantly higher than for never smokers for the entire sample of cases (RR 1.4; 95% CI: 1.1, 1.8), but a differential effect of survival by smoking history was seen between men and women [25]. The risk of death was higher for female current smokers (RR 1.9; 95% CI: 1.2—3.1) and past smokers (RR 1.4; 95% CI: 0.8, 2.3) compared to never smokers, but no effect of smoking on survival was observed in men. We observed that women had a significantly decreased risk of mortality for each 10 years of smoking abstinence (RR 0.85; 95% CI: 0.75, 0.97), and this relationship was not observed in men. This finding supports the hypothesis of previous investigators that the effect of cigarette smoking on survival is heterogeneous in nature [25]. We did not observe a significant difference in overall survival by smoking history in patients with SCLC, which is inconsistent with several studies in the medical literature. In a multicenter study of 766 SCLC patients participating in three German multicenter trials, non-smokers had a significantly longer median survival (13.6 months) compared to smokers (9.9 months) [9]. However, this study included patients enrolled in chemotherapy—radiotherapy clinical trials, and the patients may have had a worse prognosis than the patients in our cohort. Evidence for this speculation is supported by a longer median survival in our cohort (12.0 months) compared to the German clinical trial cohort (10.6 months) and a lower proportion of extensive stage patients [105/201 (52%) versus 463/757 (61%), respectively]. One possible explanation is that smoking status may be an important prognostic variable for patients with more advanced disease than among all SCLC patients. A few prior studies have attempted to assess the impact of smoking cessation on overall survival with SCLC [13]. In a study of 112 SCLC patients receiving chemotherapy, a significant trend was observed suggesting that patients who quit smoking prior (NSPrior) to diagnosis had the best survival, patients who quit at the time of diagnosis (NS-Dx) had intermediate survival, and patients who continued to smoke (S) had the worst survival [13]. Survival differences remained after adjusting for disease
Duration of smoking abstinence as a predictor for non-small-cell lung cancer survival in women extent, performance status, and type of protocol treatment. However, this effect was primarily attributable to the NS-Prior patients, since no statistically significant difference was observed between the NS-Dx and S patients. We did not observe this relationship in our 201 SCLC patients stratifying by sex and adjusting for age at diagnosis, packs per day, years smoked, disease stage, and treatment. One possible explanation is that Johnston-Early’s study did not adjust for other significant confounding factors, as we did in our study. Another possible explanation is that SCLC patients generally progress and die quickly after diagnosis and the small impact of duration of smoking abstinence on survival is not detectable. There are several limitations to this study. First of all, we assessed all-cause mortality rather than cause-specific mortality due to limitations in obtaining precise cause of death information for some patients. The patients are comprised of those in the area surrounding the Mayo Clinic (Olmsted County) and those from outside that area. For 98% of 138 patients residing in Olmsted County at the time of lung cancer diagnosis who died, we were able to identify causes of death. Death from lung cancer occurred in 100% of never smokers (n = 8), 94% of former smokers (n = 70), and 94% of current smokers (n = 60). For patients outside of Olmsted County, we did not have complete data for cause-specific mortality. Because survival after lung cancer diagnosis is often short, assuming most deaths were due to lung cancer may be reasonable. Furthermore, previous investigators have proposed that all-cause mortality may be a more appropriate analysis for assessing the impact of smoking than cancerspecific or competing causes of death alone [8]. A second limitation is potential selection bias, either due to referral practice or self-selected participation in our study. We attempted to assess selection bias by comparing former smokers with complete smoking information with those who had missing data (9.5%). We observed a shorter duration of smoking, a higher proportion of advanced disease stage at diagnosis, and shorter survival in patients with missing data compared to patients with complete data. Furthermore, the baseline characteristics of patients with missing data were compared to the characteristics of patients with complete data across duration of abstinence groups. Patients with missing data were characterized by a median age of 69 years; 61.5% were male; 51.8% smoking one pack per day or less; smoked a median of 30 years; 46.9% had adenocarcinoma; 29.9% had squamous cell carcinoma; and 7.7% had well-differentiated tumors. These features are more consistent with the characteristics of patients having a shorter duration of
171
smoking abstinence than those with longer durations. In light of the increased mortality found in patients with missing data, one could surmise that the indirect relationship between duration of smoking abstinence and mortality could have been accentuated had data been available on all patients. In addition we do not have data on performance status for all patients, particularly the early-stage patients at the time of diagnosis. If smoking status was correlated with performance status, then performance status may explain the observed differences in survival. However, in a previous study of SCLC, no association between performance status and smoking status was found [13]. The third limitation is the lack of diversity in our patient population with 96% being Caucasians of a European origin. We re-analyzed the data using only Caucasian patients and found almost identical results on the association between NSCLC survival in women and duration of smoking cessation compared to all patients combined. Another limitation of this study is that we could not control for comorbid illness. However, in a cohort study of 1155 patients with lung cancer, current smoking at diagnosis was found to be an important predictor of shortened lung cancer survival that was independent of comorbidity [8]. In conclusion, we observed a difference in survival by smoking status for NSCLC. We also observed a difference in survival between women and men in relation to duration of smoking abstinence prior to lung cancer diagnosis, which may reflect the complex interplay between biological and behavioral differences [26]. Our finding that duration of smoking abstinence predicts survival in women with NSCLC supports previous research suggesting that there is a direct biological effect of smoking on lung cancer survival [8], and our findings suggest that this effect may differ by sex.
Acknowledgements We would like to thank Drs. Randolph S. Marks, Scott H. Okuno, and Claude Deschamps for their contributions to this work. We would like to thank Susan Ernst for her technical assistance with the manuscript.
References [1] American Cancer Society, Cancer Facts and Figures, 2004. Available at http://www.cancer.org/docroot/STT/ stt O.asp. Accessed August 16, 2004.
172 [2] Ginsberg RJ, Vokes EE, Raben A. Cancer of the lung. In: DeVita VT, Hellmans S, Rosenberg SA, editors. Principles and Practice of Oncology. Philadelphia: Raven; 1999. p. 849—950. [3] Anonymous. Factors influencing long term survival after resection of bronchogenic carcinoma; an analysis of 1,118 cases. Zhonghua, Yi Xue Za Zhi, Chi Med J 1974;191(11):682—686. [4] Fujisawa T, Iizasa T, Saitoh Y, Sekine Y, et al. Smoking before surgery predicts poor long-term survival in patients with stage I non-small-cell lung carcinomas. J Clin Oncol 1999;17(7):2086—91. [5] Martins SJ, Pereira JR. Clinical factors and progrnosis in non-small-cell lung cancer. Am J Clin Oncol 1999;22:453—7. [6] Sioris T, Husgafvel-Pursianen K, Karjalainen A, Anttila S, et al. Survival in operable non-small-cell lung cancer: role of p53 mutations, tobacco smoking and asbestos exposure. Int J Cancer 2000;86:590—4. [7] Sobue T, Suzuki T, Fujimoto I, Doi O, et al. Prognostic factors for surgically treated lung adenocarcinoma patients, with special reference to smoking habit. Jpn J Cancer Res 1991;82:33—9. [8] Tammemagi DVM, McLaughlin JR, Brendan J, Mullen M, et al. A study of smoking, p53 tumor suppressor gene alterations and non-small-cell lung cancer. Ann Epidemiol 2000;10(3):176—85. [9] Wolf M, Holle R, Hans KPD, et al. Analysis of prognostic factors in 766 patients with small-cell lung cancer (SCLC): the role of sex as a predictor for survival. Br J Cancer 1991;63:986—92. [10] Hinds MW, Yang HY, Stemmermann G, Lee JJ, et al. Smoking history and lung cancer survival in women. J Natl Cancer Inst 1982;68(3):395—9. [11] Xavier F, Henn LdA, Oliveira M, Orlandine L. Smoking and its relation to the histological type, survival, and prognosis among patients with primary lung cancer. Revista Paulista De Medicina 1996;114:1298—302. [12] Bergman B, Sorenson S. Smoking and effect of chemotherapy in small-cell lung cancer. Eur Respir J 1988;1:932—7. [13] Johnston-Early A, Cohen MH, Minna JD, Paxton LM, et al. Smoking abstinence and small-cell lung cancer survival. an association. JAMA 1980;244:2175—9. [14] Kawahara M, Ushijima S, Kamimori T, Kodama N, et al. Second primary tumours in more than two-year disease-free
J.O. Ebbert et al.
[15]
[16]
[17] [18]
[19] [20] [21]
[22] [23]
[24]
[25] [26]
survivors of small-cell lung cancer in Japan: the role of smoking cessation. Br J Cancer 1998;78:409—12. Richardson GE, Tucker MA, Venzon DJ, Linnoila RI, et al. Smoking cessation after successful treatment of smallcell lung cancer is associated with fewer smoking-related second primary cancers. Ann Intern Med 1993;119:383— 90. Tucker MA, Murray N, Shaw EG, Ettinger DS, et al. Second primary cancers related to smoking and treatment of small-cell lung cancer. J Natl Cancer Inst 1997;89(23): 1782—8. Brundage MD, Davies D, Mackillop WJ. Prognostic factors in non-small-cell lung cancer. a decade of progress. Chest 2002;122(3):1037—57. World Health Organization. Histological typing of lung and pleural tumors. World Health Organization’s International Histological Classification of Tumors. 3rd ed. Berlin, Heidelberg: Springer Verlag; 1999. Greene FL. American cancer society and American Joint Committee on cancer. AJCC Cancer Staging Manual. 6th ed. New York: Springer-Verlag; 2002. Mountain CF. Revisions in the international system for staging lung cancer. Chest 1997;111:1710—7. Yang P, Wentzlaff KA, Katzmann JA, Allen MS, et al. Alpha1-antitrypsin deficiency allele carriers in lung cancer patients. Cancer Epidemiol Biomarkers Prev 1999;8: 461—5. de Andrade M, Ebbert JO, Wampfler JA, Miller DL, et al. Environmental tobacco smoke exposure in women with lung cancer. Lung Cancer 2004;43:127—34. Visbal AL, Williams BA, Nichols FC, Marks RS, et al. Gender differences in non-small-cell lung cancer survival: an analysis of 4618 patients diagnosed between 1997—2002. Ann Thorac Surg 2004;78(1):209—15. Isobe T, Hiyama K, Yoshid Y, Fujiwara Y, et al. Prognostic significance of p53 and ras gene abnormalities in lung adenocarcinoma patients with stage I disease after curative resection. Jpn J Cancer Res 1994;85:1240—6. Goodman MT, Kolonel LN, Wilkens LR, Yoshizawa CN, et al. Smoking history and survival among lung cancer patients. Cancer Causes Control 1990;1:155—63. Payne S. Smoke like a man, die like a man?: a review of the relationship between gender, sex and lung cancer. Soc Sci Med 2001;53:1067—80.