Postoperative radiotherapy for elderly patients with stage III lung cancer†‡§
We acknowledge the efforts of the Applied Research Branch, Division of Cancer Prevention and Population Science, National Cancer Institute; the Office of Information Services and the Office of Strategic Planning, Health Care Finance Administration; Information Management Services, Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare Database. The interpretation and reporting of these data are the sole responsibilities of the authors.
The sponsors of the study had no role in the study design, data collection, analysis, and interpretation, or writing of the report.
Dr. Wisnivesky had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
The potential role of postoperative radiation therapy (PORT) for patients who have completely resected, stage III nonsmall cell lung cancer (NSCLC) with N2 disease remains controversial. By using population-based data, the authors of this report compared the survival of a concurrent cohort of elderly patients who had N2 disease treated with and without PORT.
By using the Surveillance, Epidemiology, and End Results (SEER) registry linked to Medicare records, 1307 patients were identified who had stage III NSCLC with N2 lymph node involvement diagnosed between 1992 and 2005. Propensity scoring methods and instrumental variable analysis were used to compare the survival of patients who did and did not receive PORT after controlling for selection bias.
Overall, 710 patients (54%) received PORT. Propensity score analysis indicated that PORT was not associated with improved survival in patients with N2 disease (hazard ratio [HR], 1.11; 95% confidence interval [CI], 0.97-1.27). Analyses that were limited to patients who did or did not receive chemotherapy, who received intermediate-complexity or high-complexity radiotherapy planning, or adjusted for time trends produced similar results. The instrumental variable estimator for the absolute improvement in 1-year and 3-year survival with PORT was −0.04 (95% CI, −0.15 to 0.08) and −0.08 (95% CI, −0.24 to 0.15), respectively.
The current data suggested that PORT is not associated with improved survival for elderly patients with N2 disease. These findings have important clinical implications, because SEER data indicate that a large percentage of elderly patients currently receive PORT despite the lack of definitive evidence about its effectiveness. The potential effectiveness of PORT should be evaluated further in randomized controlled trials. Cancer 2012. © 2012 American Cancer Society.
Surgical resection is considered the standard of care for patients who are diagnosed with clinically early stage nonsmall cell lung cancer (NSCLC).1 However, a considerable proportion of patients who undergo surgery for localized NSCLC have involvement of N2 lymph nodes identified on pathologic staging.2-4 These patients have higher rates of local recurrence and worse survival compared with patients who have pathologic N0 or N1 disease.5, 6 Because of this increased risk of local recurrence, postoperative radiation therapy (PORT) has been proposed as a means to improve the outcomes of patients with N2 lymph node involvement who undergo resection.7
There is considerable debate regarding the potential benefit and harms of PORT for patients with NSCLC. Although some studies have demonstrated improvements in survival with PORT, those findings have not been confirmed in other trials.8-12 Current recommendations against the use of PORT are based on the results from a meta-analysis published in 1998 and that subsequently was updated to include more recent studies.13, 14 That meta-analysis analyzed data from 2343 patients with stage I to III NSCLC from 11 prospective trials (some initiated as early as 1965) and demonstrated a significant adverse effect of PORT on survival. In subgroup analysis, the detrimental survival effect seemed to be limited to those with stage I or II disease; the results for N2-positive patients favored PORT, although the difference was not statistically significant. However, these results are subject to several important limitations.15, 16 The radiation therapy (RT) techniques used in many of the trials were not consistent with current standards (eg, cobalt machines, single RT field), and some studies used larger than conventional daily fractions, which may have resulted in suboptimal effectiveness while increasing toxicity. In addition, most of these randomized controlled trials (RCTs) were limited to highly selected, younger patients who fulfilled the strict inclusion and exclusion criteria. Thus, there are limited data regarding the effectiveness of PORT for elderly patients with N2 disease who are treated in the community. More recently, several large RCTs have established adjuvant chemotherapy as the standard of care for stage IIIA NSCLC17, 18; thus, it is also important to re-evaluate the role of PORT among patients who receive chemotherapy. In this study, we used nationally representative data from the Surveillance, Epidemiology, and End Results (SEER)-Medicare registry to evaluate whether PORT was associated with improved survival among elderly patients with resected NSCLC and N2 lymph node involvement.
MATERIALS AND METHODS
We used the National Cancer Institute's SEER database, which integrates data from 17 regional cancer registries.19 SEER has been linked to Medicare enrollment and claims data using unique patient identifiers.20 Among individuals aged ≥65 years in SEER, approximately 93% have been identified in the Medicare Enrollment File and included in the SEER-Medicare registry.
We used SEER-Medicare data to identify all patients with primary, completely resected NSCLC who had N2 lymph node involvement diagnosed between 1992 and 2005. We excluded patients who were diagnosed at autopsy or using death certificate data as well as patients in health maintenance organizations or who lacked Part B Medicare coverage (coverage for outpatient care) at the time of diagnosis, because we were not able to ascertain comorbidities or chemotherapy receipt for such patients. We also excluded patients who underwent limited resection (wedge resection or segmentectomy), received preoperative chemotherapy or RT, or died during the perioperative period (within 30 days of surgery). The final cohort consisted of 1307 patients with resected, N2 NSCLC.
Sociodemographic information was obtained from SEER and Medicare databases. To evaluate the burden of comorbidities, we used the Deyo adaptation of the Charlson Comorbidity Index, applying lung cancer-specific condition weights.21, 22 In terms of histology, cases were classified as adenocarcinoma, bronchioalveolar cell carcinoma, squamous cell carcinoma, large cell carcinoma, or other histologic type. Stage was classified according to the most recent American Joint Committee on Cancer criteria.6, 23
Patients who underwent resection were categorized accordingly if SEER data or Medicare physician claims indicated that they underwent lobectomy or pneumonectomy (surgical codes 30-70). PORT use was ascertained from SEER and Medicare claims.24 Patients who received PORT were categorized accordingly if SEER codes indicated that they received postoperative external-beam RT or if Medicare inpatient, outpatient, or physician claims contained any code indicating PORT use within 4 months of surgery.25 RT complexity was determined using planning and simulation codes from Medicare physician claims.26, 27 Postoperative receipt of chemotherapy (platinum-based or other regimens) was determined from Medicare claims using published algorithms.28
Postoperative events can influence decisions regarding PORT use. Thus, the presence or absence of common surgical complications was identified using Medicare claims.29-31 We also used Medicare data to ascertain use of home health services. To be eligible for Medicare home services, beneficiaries must be homebound; thus, we used this information as a proxy for poor performance status.
Differences in distribution of baseline characteristics between patients who received or did not receive PORT were evaluated using the chi-square test. The Kaplan-Meier method was used to estimate unadjusted survival rates among patients in the 2 treatment groups. Survival was calculated from the date of resection to the Medicare date of death. Those who survived past December 31, 2007 (the date of the last follow-up) were classified as censored observations.
We used propensity score analyses to control for differences in the baseline characteristics of patients who did and did not receive PORT. Propensity scores can be thought of as a measure of the likelihood that a patient will be assigned to a certain treatment (ie, PORT vs no PORT) on the basis of his or her pretreatment characteristics. To perform the propensity score analyses, we estimated the probability that each patient would receive PORT using logistic regression.32 The model included variables for the patients' sociodemographic characteristics, comorbidities, cancer-related factors (histology, grade, tumor size, T classification, location, histology, number of lymph nodes evaluated, and number of positive lymph nodes), type of resection, postoperative complications, and use of home health services during the postoperative period. Once the model was fitted, we used regression analyses to evaluate whether the baseline covariates were balanced across study groups after adjusting for propensity scores.
Cox regression analysis was used to compare the survival of who did and did not receive PORT, adjusting for propensity scores in 3 ways. First, we included the propensity score as a continuous covariate in a Cox model comparing the survival of patients who did and did not receive PORT. In a second approach, we classified patients into quintiles based on their propensity for PORT and then fitted a stratified Cox model. Finally, we matched patients who did and did not receive PORT by their propensity scores and compared survival among study groups using a marginal Cox model with a robust sandwich variance estimator.33
To assess the potential effectiveness of PORT with and without chemotherapy, we conducted secondary analyses adjusting as well as stratifying for chemotherapy receipt. In addition, we conducted stratified analyses to assess the effectiveness of intermediate-complexity and high-complexity RT planning. Finally, we performed propensity score analyses adjusting for year of diagnosis to control for possible time trends in the use of other lung cancer treatments.
We performed an instrumental variable (IV) analysis to control for unmeasured confounders. IV analysis is a technique that attempts to simulate a randomized controlled trial using observational data. We used geographic variability in the use of PORT for the IV analysis.34-36 Different geographic regions (Health Care Service Areas [HCSAs]) in SEER were classified as high-use or low-use areas based on the proportion of patients in the HCSAs that received PORT (excluding areas with ≤5 patients). Areas in which the proportion of patients that received PORT was above the median were classified as high-utilization areas. The IV was calculated as the difference between the adjusted 1-year and 3-year survival in the high-use and low-use areas, divided by the probability of receiving PORT in those regions.34 Thus, the IV estimate represents the absolute difference in 1-year and 3-year survival among patients who did and did not receive PORT. Adjusted survival was estimated using a Cox model controlling for age, sex, race, and use of chemotherapy. The confidence interval (CI) of the IV estimate was obtained using bootstrap.34
On the basis of the number of deaths observed among patients in the cohort, we estimated that the study had >80% power to detect a 15% to 20% decreased hazard of death with PORT at the P = .05 significance level. Analyses were performed using the SAS software package (SAS Institute Inc., Cary, NC). The study was reviewed by the institutional review board and was classified as exempt.
Overall, 710 patients (54%; 95% CI, 51%-57%) with N2 disease received PORT. The baseline characteristics of the study cohort are listed in Table 1. Patients who received PORT were more likely to be younger (P < .0001) and to have higher incomes (P = .03), but there were no significant differences among groups in the distribution of sex, race/ethnicity, income, or comorbidities (P > .05 for all comparisons). Similarly, the tumor status and histology of cancers treated with and without PORT were similar. However, the number of positive N2 lymph nodes varied among patients who received and did not receive PORT (P = .05). Rates of lobectomy and pneumonectomy were similar in the 2 groups (P = .11); patients who received RT were more likely to receive adjuvant chemotherapy (P < .0001). Except for postoperative use of chemotherapy (not included in the model), all covariates were well balanced among the groups after adjusting for propensity scores (Table 1).
Table 1. Baseline Characteristics of Patients With N2 Nonsmall Cell Lung Cancer According to the Receipt of Postoperative Radiation Therapy Treatment
|Age, y|| || ||<.0001||.96|
| ≤70||297 (42)||177 (30)|| || |
| 71-75||241 (34)||196 (33)|| || |
| >75||172 (24)||224 (38)|| || |
|Sex|| || || || |
| Women||371 (52)||279 (47)||.04||.99|
|Race|| || || || |
| White||612 (85)||518 (87)||.77||.89|
| African American||46 (7)||37 (6)|| || |
| Hispanic||18 (3)||15 (3)|| || |
| Other||34 (5)||27 (4)|| || |
|Median income in zip code of residence|| || || || |
| Lowest quartile||134 (19)||145 (24)||.03||.94|
| Second quartile||182 (26)||155 (26)|| || |
| Third quartile||187 (26)||124 (21)|| || |
| Highest quartile||204 (29)||172 (29)|| || |
|Charlson Comorbidity Score|| || || || |
| ≤1||324 (46)||245 (41)||.14||.98|
| 1-2||223 (31)||190 (31)|| || |
| >2||163 (23)||162 (27)|| || |
|Tumor classification|| || || || |
| T1||172 (24)||148 (25)||.98||.99|
| T2||419 (59)||348 (58)|| || |
| T3||119 (17)||101 (17)|| || |
|No. of lymph nodes evaluated|| || || || |
| ≤5||170 (24)||126 (21)||.19||.70|
| 6-9||163 (23)||124 (21)|| || |
| ≥10||377 (53)||347 (58)|| || |
|No. of positive lymph nodes|| || || || |
| ≤2||311 (44)||300 (50)||.05||.89|
| 3-4||156 (22)||108 (18)|| || |
| ≥5||243 (34)||189 (32)|| || |
|Histology|| || || || |
| Adenocarcinoma||455 (64)||379 (63)||.71||.98|
| Squamous cell carcinoma||199 (28)||167 (28)|| || |
| Large cell carcinoma||38 (5)||31 (5)|| || |
| Other||18 (3)||20 (3)|| || |
|Type of surgery|| || ||.11||.98|
| Lobectomy||623 (88)||506 (85)|| || |
| Pneumonectomy||87 (12)||91 (15)|| || |
|Adjuvant chemotherapy|| || || || |
| None||455 (64)||461 (77)||<.0001||.002b|
| Platinum-based||223 (31)||118 (20)|| || |
| Other chemotherapy||32 (5)||18 (3)|| || |
On unadjusted analysis, the survival of patients who did and did not receive PORT did not differ significantly (P = .30). Similarly, propensity score analyses did not reveal an improved survival with the use of PORT (Table 2). A Cox model adjusting for propensity scores indicated that postoperative survival was not improved with PORT (hazard ratio, 1.11; 95% CI, 0.97-1.27). Analyses that were stratified (hazard ratio, 1.12; 95% CI, 0.98-1.28) or matched (HR, 1.10; 95% confidence interval, 0.95-1.27) by propensity score had similar findings. Secondary analyses adjusting for the receipt of adjuvant chemotherapy or limiting the sample to patients who did or did not receive postoperative chemotherapy, or with intermediate-complexity and high-complexity RT planning also indicated that PORT was not associated with improved survival (HR, 1.06-1.25). Finally, the association between PORT use and survival remained unchanged when analyses were repeated adjusting for potential time trends in the use of other lung cancer treatments.
Table 2. Propensity Score Analysis: Comparison of Survival of Patients With Resected N2 Disease who Did and Did Not Receive Postoperative Radiation Therapy
|Primary analysis: Entire cohort|| || |
| Adjusting for propensity scores||1.11 (0.97-1.27)||1.13 (0.99-1.30)|
| Stratified by propensity score quintiles||1.12 (0.98-1.28)||1.14 (1.00-1.30)|
| Matched analysis||1.10 (0.95-1.27)||1.12 (0.97-1.29)|
|Stratified analyses|| || |
| Limited to patients who received chemotherapyb|| || |
| Adjusting for propensity scores||1.17 (0.88-1.56)||—|
| Stratified by propensity score quintiles||1.18 (0.89-1.58)||—|
| Matched analysis||1.22 (0.88-1.67)||—|
| Limited to patients who did not receive chemotherapy|| || |
| Adjusting for propensity scores||1.13 (0.97-1.32)||—|
| Stratified by propensity score quintiles||1.14 (0.98-1.33)||—|
| Matched analysis||1.25 (0.96-1.32)||—|
| Limited to patients who received intermediate-complexity RT planning|| || |
| Adjusting for propensity scores||1.09 (0.94-1.26)||1.10 (0.94-1.27)|
| Stratified by propensity score quintiles||1.09 (0.94-1.27)||1.10 (0.95-1.28)|
| Matched analysis||1.05 (0.91-1.21)||1.06 (0.91-1.22)|
| Limited to patients who received high-complexity RT planning|| || |
| Adjusting for propensity scores||1.09 (0.87-1.38)||1.17 (0.92-1.48)|
| Stratified by propensity score quintiles||1.13 (0.89-1.43)||1.20 (0.95-1.54)|
| Matched analysis||1.09 (0.95-1.27)||1.12 (0.97-1.29)|
| Adjusting for time trends|| || |
| Adjusting for propensity scores||1.06 (0.92-1.22)||1.07 (0.93-1.24)|
| Stratified by propensity score quintiles||1.07 (0.93-1.22)||1.10 (0.95-1.25)|
| Matched analysis||1.05 (0.91-1.21)||1.07 (0.93-1.24)|
Assessment of PORT use in the HCSAs in SEER revealed that approximately 34.1% of patients in the low-use areas received PORT compared with 66.7% in the high-use areas. The IV estimate indicated that PORT was associated with a 0.04 decrease in 1-year cumulative survival (95% CI, −0.15 to 0.08). Similarly, 3-year survival among patients who did and did not receive PORT did not differ significantly (absolute difference in cumulative survival, −0.08; 95% CI, −0.24 to 0.15).
The role of PORT in patients with completely resected stage III NSCLC who have involvement of N2 lymph nodes remains uncertain. However, despite the lack of conclusive evidence from randomized trials, PORT is used frequently to treat elderly patients with N2 disease.37 By using a population-based registry, we demonstrated that PORT was not associated with improved survival in elderly patients with resected NSCLC who had N2 disease identified at pathologic staging. These data suggest that PORT should not be used routinely to treat these patients outside research trials.
The prognosis of patients who undergo lung cancer resection depends greatly on the extent of lymph node involvement. Although long-term survival may be achieved by approximately 70% of patients without lymph node metastasis, it decreases to between 20% and 35% for those with microscopic N2 disease.5, 6 Because 20% to 40% of patients with N2 disease experience treatment failure caused by locoregional relapse, PORT frequently is used to reduce the risk of local recurrence. Several small, randomized controlled trials conducted from 1965 to the 1990s assessing the role of PORT after NSCLC resection produced discordant results.8-12 Enthusiasm for PORT diminished after the 1998 PORT Meta-Analysis Trialists Group reported a 7% absolute greater mortality associated with PORT, particularly for patients with N0 or N1 disease.13, 14, 38 The results from that meta-analyses also indicated a nonsignificant survival benefit for patients with N2 lymph node involvement, suggesting that PORT may be beneficial for patients with more advanced disease.
More recent data have renewed the interest in PORT. Several recent studies have suggested that PORT may be of benefit for patients with completely resected N2 NSCLC.37, 39-41 A subgroup analysis of the Adjuvant Navelbine International Trialist Association trial demonstrated that PORT led to longer overall survival in patients with resected N2 NSCLC.41 An analysis of the SEER database (not linked to Medicare) in patients with resected NSCLC between 1998 and 2002 indicated that PORT was associated with longer survival for patients with N2 disease.37 However, SEER data do not contain information about comorbidities, complications of surgery, or use of adjuvant chemotherapy.28 Because a lower number of comorbidities and chemotherapy use are associated with PORT treatment, lack of adjustment for these covariates may explain the discordant results among studies. In addition, previous studies used standard regression analysis to compare the outcomes of patients who did and did not receive PORT and did not perform more advanced statistical techniques to control for selection bias. Our findings are consistent with the conclusions of the PORT Meta-Analysis Trialists Group suggesting that, until definitive data are available, use of PORT should we limited to patients enrolled in clinical trials.
The Lung Adjuvant Radiotherapy Trial or LungArt is an ongoing, multicenter, phase 3, randomized controlled trial evaluating the effectiveness of PORT in patients with resected NSCLC who have N2 lymph node involvement. The study, sponsored by a French cooperative group, will recruit a total of 700 patients randomized to PORT versus a control arm. Our results highlighting the importance of enrolling patients into this study should provide useful data to assess assumptions regarding the potential effect of PORT on survival and to determine the appropriate duration of follow-up so that the study will be adequately powered.
Several strengths and limitations regarding our study are worth noting. Because this was a retrospective study, the use of PORT was not random but, rather, was influenced by patients' and physicians' preferences, baseline characteristics, and practice patterns. Consequently, differences in outcomes among patients who did and did not receive PORT may be explained by confounding by indication. However, we used propensity score methods to balance the study groups and control for all measured covariates, including detailed clinical and tumor characteristics, which are the most important prognostic factors for patients with stage III NSCLC. In addition, IV analyses provide a consistent estimate of the effect of PORT even in the presence of unmeasured confounders. Thus, until data from contemporary RCTs are available, these results are a valuable source of information about the potential benefit of PORT.
Because the SEER-Medicare database is a population-based registry, it is affected less by referral patterns and other sources of bias. Thus, the generalizability of our results should be excellent. The large number of patients with N2 disease in the registry and the extended follow-up ensured that the study was powered to detect relatively small benefits of PORT. However, SEER does not provide data regarding disease recurrence; thus, we were not able to assess whether PORT is associated with other important secondary outcomes, such as increased disease-free survival and/or lower rates of local recurrence. In addition, no data regarding the total radiation dose or fractionation schedule used to treat each patient are provided in the SEER-Medicare database. Thus, we were not able to assess the impact of these factors on lung cancer survival. We excluded patients who received neoadjuvant chemotherapy and did not assess whether the extent of lymph node dissection influenced the potential effect of PORT; these issues should be explored in future studies.
In summary, our findings suggest that PORT is not associated with improved survival in elderly patients with stage III NSCLC who have N2 disease and who underwent surgical resection. Thus, our results are consistent with the conclusion of prior meta-analyses that also found no benefit for PORT in this clinical setting. These data should be important, because the SEER-Medicare trends indicate that a large proportion of elderly patients with N2 disease are currently being treated with PORT despite the lack of definitive evidence from RCTs.
This study was supported by the National Cancer Institute (grant 5R01CA131348-02).
CONFLICT OF INTEREST DISCLOSURES
J.P.W. is a member of the research board of EHE International, has received lecture honorarium from Novartis Pharmaceutical, and was awarded a research grant from GlaxoSmithKline to conduct a study of chronic obstructive pulmonary disease.