|
|
||
 |
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 13 (May 1), 2007: pp. 1705-1712 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2006.08.1455 Sex Differences in Lung Cancer Survival: Do Tumors Behave Differently in Elderly Women?
From the Divisions of General Internal Medicine and Pulmonary, Critical Care, and Sleep Medicine, Department of Medicine; and the Department of Health Policy, Mount Sinai School of Medicine, New York, NY Address reprint requests to Juan P. Wisnivesky, MD, MPH, Department of Medicine, Mount Sinai School of Medicine, 1 Gustave L. Levy Pl, Box 1087, New York, NY 10029; e-mail: juan.wisnivesky{at}mssm.edu
Purpose: Women with lung cancer appear to have better survival. Whether this results from better response to treatment, different tumor biology, or a longer life expectancy is not well understood. This study sought to assess sex differences in the natural history of lung cancer after controlling for unrelated causes of death and type of treatment. Methods: This study included 18,967 elderly patients with stage I and II non–small-cell lung cancer diagnosed between 1991 and 1999 from the Surveillance, Epidemiology, and End Results registry linked to Medicare records. Patients were grouped into three categories according to the treatment received: surgery, radiation or chemotherapy but no surgery, and untreated cases. We used stratified and multivariate analyses to evaluate sex differences in survival using three methods to control for competing risks: lung cancer–specific survival, overall survival adjusting for comorbidities, and relative survival. Sensitivity analysis was used to test whether potential differences in smoking could account for the observed association of sex with survival. Results: Women in all treatment groups had better lung cancer–specific, overall, and relative survival than did men (P < .0001). Stratified and multivariate analyses showed that women had better survival than did men after controlling for confounders. Sensitivity analyses showed that potential sex differences in smoking did not explain our findings. Conclusion: In this national, population-based sample, elderly women with early lung cancer had better risk-adjusted survival regardless of the type of treatment. That sex differences were observed among untreated patients suggests that lung cancer in women may have a different natural history.
Lung cancer appears to behave differently in women. Several studies have reported sex differences in the clinical presentation, histology, and outcomes of lung cancer. Women are more likely to present with lung adenocarcinoma, tend to receive diagnosis at an earlier age, and are more likely to be diagnosed with localized disease.1-6 Women may also be more predisposed to molecular aberrations resulting from the carcinogenic effects of tobacco,7-9 but do not appear to be more susceptible to develop lung cancer.10,11 A number of studies have also reported that women with lung cancer have better survival.1-4,6,12-21 It is unclear, however, whether lung cancer in women has a different natural history than in men. Most prior studies were limited to patients who received treatment for their cancer.1,6,13,15-17 However, evaluating sex differences in tumor behavior requires studying the natural history of the disease among untreated cases. Sex differences in survival among treated patients may be related to better responses to treatment among women and not caused by differences in tumor biology. Additionally, lack of adjustment for sex-related differences in deaths resulting from unrelated causes may have confounded the results in some of these studies. This study sought to evaluate whether elderly women with lung cancer have longer survival after accounting for sex differences in life expectancy and other factors known to influence mortality rates, and assess whether the survival advantage is present among treated as well as untreated patients.
Data Sources Patients were identified from the Surveillance, Epidemiology, and End Results (SEER) registry linked to Medicare records. The SEER program collects information on all new cases of cancer from 14 population-based registries covering approximately 26% of the US population. The Medicare database has information on all Medicare claims and encompasses 97% of individuals age 65 years or older. For patients age 65 years or older included in the SEER registries, 94% have been linked to Medicare.22 We linked the SEER-Medicare registry with county-level data from the Area Resource File to obtain information on physician availability in the patient's area of residence.23
Study Cohort
Sociodemographic Variables, Comorbid Disease, and Determination of Survival To assess local access to care, we constructed two strata using data on physician availability: one containing the patients who resided in areas with the lowest quartile of physicians per 1,000 population, and the other consisting of the remaining patients. All patients in the study were insured by Medicare. Patients were classified as members of a health maintenance organization (HMO) if Medicare records indicated that the patient had HMO coverage more than 75% of the follow-up time. To evaluate the burden of comorbid disease in our cohort, we used the Deyo adaptation of the Charlson comorbidity index.26 We calculated the patient's comorbidity index for each year of follow-up using data from Medicare inpatient, outpatient, and physician records.27 Given that Medicare does not collect complete data for persons in HMOs (1,043 patients, 5% of the cohort), we could not calculate this index for these individuals. Patients were classified as having undergone resection if SEER data indicated that a surgical procedure (segmentectomy, wedge resection, lobectomy, or pneumonectomy) had been performed.28 Patients were classified as having received radiotherapy if the SEER radiation code indicated that the patient underwent beam radiation. We considered patients as treated with adjuvant chemotherapy if Medicare claims indicated that the individual received a platinum-based regimen within 3 months of diagnosis.29 Patients were grouped into three categories based on the type of treatment received: surgery (including patients who received postoperative radiotherapy or chemotherapy), radiation or chemotherapy but no surgery, and untreated cases. Survival was determined as the interval from the date of cancer diagnosis to the SEER date of death. Those surviving past December 31, 1999, were classified as censored. The cause of death was coded according to information provided in the SEER registry, which uses state death certificates as a primary source.
Statistical Analysis We assessed sex differences in prognosis by comparing unadjusted survival rates of men and women for the entire cohort, conducting stratified analyses within relevant subgroups, and determining the effect of sex on survival while controlling for important confounders using multiple regression analyses. These analyses were performed using the three methods described in the preceding explanation and stratified by the type of treatment received (surgery, chemotherapy or radiation, and untreated cases). Lung cancer–specific survival curves were constructed using the Kaplan-Meier method.31 We used Cox regression to evaluate for sex differences in overall survival while adjusting for comorbidities.32 These models were fitted, including the patient's comorbidity index for each year of follow-up as a time-dependent covariate.33 Similarly, we used Cox regression for lung cancer–specific and overall survival analyses involving adjustments for potential confounders. A multiple regression analysis of relative survival was performed using the Hakulinen method,34 conceived as an additive relative survival model and considering that mortality resulting from lung cancer is added to the mortality resulting from other causes. The SEER-Medicare database does not include smoking information. Smoking habits differ among men and women, and smoking status has been linked in some35-37 but not all38,39 studies to overall survival of lung cancer patients. Therefore, the potential association of sex with prognosis may be confounded by smoking status. To evaluate this possibility, we performed stratified analyses among patients with and without chronic obstructive pulmonary disease (COPD). Although COPD is not a sensitive surrogate for a positive smoking history, if the effect of sex on survival was completely confounded by smoking status, the association between sex and lung cancer prognosis should not persist among patients with COPD, most of whom can be assumed to be current or past smokers. Sensitivity analysis can also be utilized to adjust for unmeasured covariates. Thus, we used this approach to assess whether sex differences in smoking could explain the observed association of sex with survival.40 In our analysis, we used published data on smoking patterns of men and women with lung cancer and the relative hazard of death associated with smoking, to evaluate the robustness of our findings across different scenarios. Analyses were performed with SAS 9.0 (SAS Institute, Cary, NC) and SURV3 software (Finnish Cancer Registry, Helsinki, Finland).
Patient Characteristics There were 18,967 patients in this study; 8,231 (43%) were female and 10,736 (57%) were male. The clinicopathologic characteristics of these patients are reported in Table 1. Women were more frequently diagnosed at an older age and were more likely to be white and to reside in areas with higher median income. Adenocarcinomas were more common among women. Female patients were more likely to present with stage IA disease compared with male patients. Among resected cases, women were less likely to receive postoperative radiotherapy (11% v 13%; P = .005) but equally likely to undergo adjuvant chemotherapy (< 2% for both groups; P = .34).
Association of Sex and Survival: Unadjusted and Stratified Analysis Lung cancer–specific and overall survival (adjusted for comorbidities) were significantly better for women (P < .0001). Median cancer-specific survival for women was 51 months (95% CI, 48 to 54 months) compared with 37 months (95% CI, 35 to 39 months) for men. The 5-year relative survival rates were significantly higher in women than in men (46% v 38%, respectively; P < .0001). Stratified analysis showed that women had increased lung cancer-specific, overall, and relative survival across several subgroups (Table 2). The survival advantage for women was observed among patients who underwent surgery, those treated with only radiation or chemotherapy, and untreated patients (Fig 1A-C). Among patients with COPD, most of whom have a positive smoking history, women with lung cancer also had better prognosis. Sex differences were present in patients with all histologic types except for squamous cell carcinoma tumors; among these cases, no significant sex differences in survival were observed.
Multivariate and Sensitivity Analyses We performed multiple regression analyses, stratifying by type of treatment, to evaluate whether sex-related survival differences persisted after adjusting for age, race, marital status, income, geographic area, physician availability, COPD status, histology, and stage. As seen in Table 3, women had a lower risk of lung cancer mortality regardless of the type of treatment received (hazard ratio, 0.74 to 0.78). Similarly, female sex was associated with decreased overall mortality in a Cox model that included a time-varying comorbidity effect. According to these models, women had a 22% to 27% decreased risk of death. Multiple regression analysis of the relative risk of death also showed a survival advantage for women across all treatment groups.
For patients treated with surgery and radiation or chemotherapy, we found a statistically significant interaction between sex and tumor histology (P  .02) indicating that the survival advantage for women varied across different histologic types. On the basis of these models, women with squamous cell carcinoma had similar survival to men. No significant interaction effect was found between sex and age group, or COPD status (P > .9 for both comparisons). In sensitivity analyses (Table 4), we found that potential sex differences in smoking did not seem to explain the observed association between sex and survival. For example, adjusted sex differences in survival persisted even if women were four times more likely than men to be never-smokers, and smoking history had a hazard ratio of 1.5, or never-smokers were three times more common among women and smoking history had a hazard ratio of 2.
In this national, population-based study of 18,967 elderly patients with stage I and II non–small-cell lung cancer, women had better survival than men regardless of the type of treatment received. Because these conclusions are critically dependent on adequately controlling for sex differences in comorbidities and overall life expectancy, we used three different analytic techniques. All three approaches provided the same result; sex differences in survival were consistently present even after controlling for other prognostic factors. The survival advantage of women was specific to adenocarcinoma and large-cell carcinoma cases, as we found no sex differences in survival among patients with squamous cell carcinomas in both the stratified and multivariate analyses. These findings suggest that lung cancer in elderly women may have a different natural history and tumor biology than in men. Further evidence from studies exploring mechanisms mediating these differences in prognosis should be sought to validate our findings based on epidemiologic data. A major potential confounder of the association between sex and survival is the effect of deaths due to unrelated conditions, which are greater in men. In this study, we showed a survival advantage for women with lung cancer using three different methods to adjust for this bias. The analysis focusing on lung cancer–specific mortality controls for competing risks by classifying as censored observations individuals dying as a result of unrelated causes. To be accurate, this method requires reliable data on the cause of death. This information in SEER is abstracted from death certificates, and inaccuracies in the cause of death in this document have been reported.41,42 However, for lung cancer the underlying cause of death has been found to be more than 90% accurate in a large registry.43 Another assumption of a lung cancer–specific survival approach is that cancer mortality is independent of competing-risk mortality. Albeit, this assumption may be approximately true, and adjustment for age usually prevents serious problems. When comparing the overall mortality of males and females, we controlled for comorbid conditions using Cox regression. We not only adjusted for comorbidities at presentation, but also used information on the burden of comorbid illnesses for each year of follow-up, because these may change over time. Although this approach should more thoroughly control for competing risks, it has not been validated. The third method, examining relative survival, estimates the ratio of the observed to the expected survival of a comparable group from the general population. We used US population life-tables stratified by age, sex, and calendar period to estimate expected survival. These tables include the effect of deaths caused by lung cancer. Nonetheless, deaths resulting from a specific disease generally constitutes only a small fraction of the total population mortality and have a negligible effect on estimates of expected survival, even for common cancers.30,44 Prior studies evaluating the effect of sex on lung cancer prognosis suggest that women have a survival advantage.1-4,6,12-19 The majority of patients in these studies, however, underwent surgery or chemotherapy. The present study extends these results to patients with untreated early-stage lung cancer, suggesting that lung cancer has a different natural history among women. The reasons for this survival advantage have not been identified. Hormonal influences may play a role in lung cancer progression given that studies have confirmed the presence of estrogen receptors in malignant lung tissue45,46 and hormones have also been linked to the pathogenesis of lung cancer.47,48 In addition, genetic and metabolic factors have been proposed as a potential explanation for the survival benefit experienced by women.49-52 We were not able to directly control for the effect of potential sex differences in smoking habits because data on tobacco exposure are not recorded in the SEER-Medicare registry. Sex differences in smoking patterns have been described among lung cancer patients, and some studies have reported that the patient's smoking status may be associated with worse overall survival35-37,53,54 but not to lung cancer–specific survival.36,54,55 Conversely, a large systematic review of studies evaluating predictors of survival in lung cancer patients reported that most studies found that smoking was not an independent prognostic variable.39 Because smoking is associated with many factors that may contribute to poorer survival, establishing a relationship between smoking and lung cancer prognosis requires careful adjustment for competing risks. In our study, we used two different approaches to control for the potential effect of sex differences in smoking habits in our results. We showed that sex differences in survival were present in analyses restricted to patients with COPD56 (most of whom we justifiably assumed to be current or past smokers), suggesting that our results do not result from sex differences in smoking status. Sensitivity analyses also suggested that the observed association between female sex and improved survival does not appear to be explained by sex differences in smoking status. Several strengths and limitations should be noted. The generalizability of our findings should be strong. The SEER registry contains population-based data; therefore, it is less affected by referral patterns that might be associated with hospital-based series. We focused on Medicare beneficiaries who were 65 years of age and older; thus, we could not explore whether there are similar sex differences in survival for younger patients. However, prior studies have shown sex differences in lung cancer survival among all age groups.4,19 We used an aggregate measure of income as a surrogate for socioeconomic status of each patient; however, we realize that this measure may be a suboptimal marker of the patient's socioeconomic status. Additionally, we did not address survival differences in patients with advanced lung cancer. We focused on stage I and II cases because treatment of these patients is well standardized. Additionally, the natural history of disease is best studied by following the progression of a cohort of patients diagnosed at an early and uniform point in the course of the disease.57 In summary, this study shows that elderly women with lung cancer have improved survival independent of sex differences in life expectancy resulting from unrelated causes of death. The observed survival advantage among untreated patients suggests that lung cancer in women has a different natural history and, potentially, a different tumor biology. This information has implications for the design and analysis of randomized trials, which may benefit from stratification by sex, and may be incorporated in prognostic classifications. SEER-Medicare data shows that women with stage I and II non–small-cell lung cancer have 5-year lung cancer–specific survival rates that are 5% to 10% higher than those in men. More accurate staging is important in communicating prognostic information to lung cancer patients. Sex differences in survival, together with data on comorbidities and patient preferences should also be incorporated into the decision-making process for treatments such as the use of adjuvant chemotherapy for early-stage non–small-cell lung cancer. These findings deserve further analysis to evaluate the biologic basis of sex-specific differences because an improved understanding of the biology of lung cancer may reveal new targets for treating this devastating disease.
The authors indicated no potential conflicts of interest.
Conception and design: Juan P. Wisnivesky Financial support: Juan P. Wisnivesky Collection and assembly of data: Juan P. Wisnivesky Data analysis and interpretation: Juan P. Wisnivesky, Ethan A. Halm Manuscript writing: Juan P. Wisnivesky, Ethan A. Halm Final approval of manuscript: Juan P. Wisnivesky, Ethan A. Halm
Detailed Results of the Statistical Analysis
We thank National Cancer Institute, the Health Care Finance Administration, Information Management Services (IMS) Inc, and SEER in the creation of the SEER-Medicare Database; and Angela Fahey of IMS and Paul Dickman, PhD, for helpful discussions.
Supported by Grant No. K08 HS013312 from the Agency for Healthcare Research and Quality (J.P.W.). Presented in abstract form at the 71st Annual Meeting of the American College of Chest Physicians, October 29–November 3, 2005, Montreal, Quebec, Canada. Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2007 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
2 test. We used three methods to adjust for sex-related differences in life expectancy. First, we computed lung cancer–specific survival for men and women, because it allowed us to control for unrelated causes of death. To estimate cancer-specific survival, deaths attributed to causes other than lung cancer were censored at the date of death. Second, we computed overall survival (all-cause mortality) while adjusting for comorbidities. The underlying hypothesis of these analyses is that the effect of competing risks should be similar regardless of patient's sex after accounting for differences in the burden of comorbidities. Finally, we estimated the relative survival of men and women using the life-table method.
