- Email: [email protected]
Survival Comparison in Patients with Stage IV Lung Cancer in Academic versus Community Centers in the United States
ORIGINAL ARTICLE
Survival Comparison in Patients with Stage IV Lung Cancer in Academic versus Community Centers in the United States Sendhilnathan Ramalingam, MD, Michaela A. Dinan, PhD, Jeffrey Crawford, MD* Duke Cancer Institute, Duke University, Durham, North Carolina Received 24 August 2018; revised 28 August 2018; accepted 5 September 2018 Available online - 8 October 2018
ABSTRACT Introduction: Although metastatic NSCLC is widely treated in both academic centers (ACs) and community-based centers (CCs), it is unclear whether outcomes are similar across both settings. A growing variety of chemotherapies and targeted agents for an increasingly histology- and molecularbased treatment strategy could provide an advantage to patients treated in ACs. Using the National Cancer Database, we investigated whether treatment at ACs was associated with a survival advantage in metastatic NSCLC. Methods: We conducted a retrospective analysis of the National Cancer Database after the introduction of novel NSCLC chemotherapy agents from 1998 to 2010. The primary outcome was 2-year survival, which was analyzed by using a multivariable regression model controlling for age, year of diagnosis, sex, primary payer, histologic type, facility type (AC versus CC), and an interaction term allowing a time-based comparison of survival between ACs and CCs. Alpha was set to 0.001 because of the size of the data set. Results: There were 193,279 patients included in this study. The percentage of patients achieving 2-year survival was higher in ACs versus in CCs in 1998 (11.5% versus 9.2% [þ2.3%]), and by 2010, the difference had increased to 17.4% versus 13.1% (þ4.3%). Multivariable analysis confirmed a significant relative increase in 2-year survival associated with ACs versus with CCs from 1998 to 2010 (p ¼ 0.0005). A histology-dependent survival difference was also noted in adenocarcinoma versus in squamous cell carcinoma (10.2% versus 9.9% in 1998 [þ0.3%], increasing to 17.3% versus 10.1% in 2010 [þ7.2%]). Adenocarcinoma survival also varied by treatment facility, with the difference in 2-year survival in ACs versus in CCs increasing from 12.3% versus 9.1% (þ3.2%) in 1998 to 20.5% versus 15.5% (þ5%) in 2010, with a trend toward significance in our multivariable model (p ¼ 0.005). Conclusions: A greater increase in survival was noted in ACs than in CCs over this time period, and it was particularly pronounced among patients with adenocarcinoma
Journal of Thoracic Oncology
Vol. 13 No. 12: 1842-1850
versus in those with squamous cell carcinoma. Given the known advances in adenocarcinoma treatment compared with in squamous cell lung cancer over this time period, our study suggests that potential treatment-related disparities may exist between ACs and CCs. Further study will be needed to validate this disparity in health care and address opportunities to improve survival in patients with stage IV NSCLC across treatment settings. 2018 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. Keywords: Survival; Non-small cell lung cancer; Disparities; Population health
Introduction Lung cancer remains the deadliest malignancy in the United States, accounting for the largest number of cancer-related deaths1 and with roughly half of patients with NSCLC presenting with metastatic disease.2 Treatment for metastatic NSCLC has evolved dramatically over the past 20 years, with increased emphasis on histology- and molecularly driven treatment algorithms. The number of new treatment options that continue to increase the availability of a broader array of treatments began between 1998 and 2010 with the approval
*Corresponding author. Disclosure: The authors declare no conflict of interest. Address for correspondence: Jeffrey Crawford, MD, Duke Cancer Institute, Duke University Medical Center, Department of Medicine, Box 3476, Room 2592, Morris Building, Durham, NC 27710. E-mail: [email protected] ª 2018 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. ISSN: 1556-0864 https://doi.org/10.1016/j.jtho.2018.09.007
December 2018
of cytotoxic agents such as pemetrexed3–5 and chemotherapy doublets, as well as targeted agents against mutations in EGFR6 and bevacizumab.3,7 It was in the nonsquamous subgroup in particular that we recognized that these new therapies are associated with improved outcomes. Before the introduction of these new therapies, advanced NSCLC was treated with platinum-based chemotherapy, with response rates of only about 19% and median overall survival of 7.9 months8,9; however, during our study period national-level database analyses have indicated a trend of increasing survival in advanced NSCLC.10,11 In sum, treatment of advanced NSCLC became more complex in two major ways. First, the histologic type of the tumor became paramount with the addition of pemetrexed in 20043,4 and the realization of its differential advantage in nonsquamous NSCLC.4,5 Second, the introduction of targeted therapies began an era of more complex medical oncology that was dependent on the mutational profile of the tumor.6,12,13 Oncologists increasingly needed to know which profiling to do and which agents were most applicable. A growing body of literature in NSCLC has shown that treatment choices and outcomes in localized NSCLC vary between academic centers (ACs) and community-based centers (CCs).14,15–19 We hypothesized that ACs may have more availability of lung cancer–specific pathologists and availability for molecular testing to manage the growing complexity of systemic therapies for metastatic NSCLC; that this would be manifested as a growing difference in 2-year survival rates between ACs and CC; and that if survival differences exist, there could be disparities of care accounting for these differences.
Methods Data Set and Study Population The National Cancer Database (NCDB) was used to conduct this study. Initially established in 1989, the NCDB currently encompasses approximately 1500 treatment facilities and is estimated to capture approximately 70% of malignancies diagnosed in the United States. Data collected include patient demographics; grade, location, and stage of the relevant malignancy; timing of therapies; types of therapies (including surgery, radiation, and chemotherapy, though not the identity of specific systemic therapies administered); 30-day mortality; and time to last contact. Geographically, the contributing facilities are located in 49 of 50 states as well as in Puerto Rico. About 20% of these are ACs.20 Our data set contained data from 1998 to 2012, allowing for 2-year survival analysis between 1998 and 2010. The NCDB has been used to study outcomes in cancer, including NSCLC. Predominantly, this includes studies of
Stage IV Lung Cancer in ACs versus in CCs
1843
outcomes in surgery and radiation therapy,15,16 as well as chemotherapy in the adjuvant/neodjuvant setting.17–19
Inclusion/Exclusion Criteria The NCDB user file included patients with NSCLC diagnosed between 1998 to 2010. This study was limited to patients with a diagnosis of stage IV NSCLC between 1998 and 2010 at a facility reporting to the NCDB throughout the study period. Patients were excluded if treatment facility type and/or duration of follow-up was missing or treatment was done entirely at another facility. Patients were also excluded if NSCLC was not their first malignancy and if they did not have a treatment decision or did not go through with the treatment prescribed by their physician (treatment was defined as chemotherapy/immunotherapy or the decision not to treat with either). Three types of facility designations were present in the data set: community cancer program, comprehensive community cancer program, and academic/research program (AC). Per the NCDB definition, both community cancer programs and comprehensive community cancer programs provide a range of diagnostic and treatment methods and participate in or refer patients for clinical trials; the distinction between the two designations is a volume-based metric (<500 versus >500 cases/y), as a result of which our data analysis these groups were combined to form the CC designation.21 Academic/ research programs must have more than 500 new cases per year and are associated either with a National Cancer Institute–designated comprehensive cancer center or with residencies in four areas, including internal medicine and general surgery. The primary end point was 2-year survival, allowing us to balance the number of years analyzed in the study while at the same time capturing a large percentage of mortality. Two-year survival was defined as the time between diagnosis and last contact of 24 months or longer. In the NCDB, these data are entered by certified tumor registrars at the relevant treatment site.
Statistical Analysis Multivariable OLS Regression Model. Comparison of the primary end point between ACs and CCs was done by using a multivariable ordinary least squares (OLS) regression model with an interaction term between facility type and year of diagnosis to allow for a continuous, time-based comparison and to help address the effect of baseline differences in the populations of patients in ACs and CCs. Control variables were year of diagnosis, type of treatment facility, age, sex, primary payer status, tumor histologic type, and the interaction term facility type by year of diagnosis. The interaction
1844 Ramalingam et al
term was essential for our analysis; therefore, a linear probability model using OLS regression was favored over logistic regression to allow for generalizable interpretation of interaction coefficients. To evaluate confounders, multivariable models were also used to test changes in several variables over time between ACs and CCs; these variables include age, sex, African American race, histologic type, treatment decision, duration of time to chemotherapy, and distance of travel for treatment (based on zip code of residence and zip code of treating hospital). Predictor variables used in each model are listed in Supplementary Table 1. Charlson-Deyo comorbidity index was not included in the primary outcome model because it was present in the NCDB only after 2003. Presence of brain metastasis at diagnosis could not be included because it was present in the data set only in 2010 and afterward. Determination of baseline differences between ACs and CCs for categorical variables was done with the Pearson chi-square test. Comparison of absolute difference between time points was done with either the two-sample differences in a means test or the difference in proportions test without continuity correction (owing to the size of our sample). Alpha was set to 0.001 for all comparisons because of the large sample size. All analyses were done with R software (R Foundation for Statistical Computing, Vienna, Austria). We applied for and obtained access to the 2012 NCDB participant user file with approval from our institutional review board.
Results Description of Data Set The final data set after satisfaction of the inclusion/ exclusion criteria included 193,279 patients (Supplementary Fig. 1) with clinical or pathologic stage IV NSCLC. All baseline differences between ACs and CCs were statistically significant at p < 0.001 with use of the Pearson chi-square test for all variables except age, for which the two-sample difference in means test was used (Table 1); OLS regression models with interaction terms (facility type year of diagnosis) were used to determine whether differences between ACs and CCs changed over time (see Supplementary Table 1). In ACs versus in CCs, the mean age was 62.1 versus 64.2 years (p < 0.001 [two-sample difference in means test]). The percentage of black patients was higher in ACs than in CCs (17.1% versus 9.6%, respectively [p < 0.001 with use of the Pearson chi-square test). The percentage of patients with Charlson-Deyo comorbidity scores of 2 or higher was lower in ACs than in CCs (7.3% versus 9.3% [p < 0.001 with use of the Pearson chi-square test]). The percentage of patients insured by Medicare was lower in ACs than in
Journal of Thoracic Oncology
Vol. 13 No. 12
Table 1. Demographics Data for the Overall Sample, ACs, and CCs
Characteristic Mean age ± SD, ya Sex, males vs. femalesa Ethnicitya White Black Other NA NSCLC typea Adenocarcinoma BAC Large cell NSCLC NOS Other Squamous Comorbiditiesa CDCC 0 CDCC 1 CDCC 2 Insurance statusa Uninsured Private Medicaid Medicare Other government Unknown
Overall Sample ACs CCs (N ¼ 193,279) (n ¼ 61,874) (n ¼ 131,405) 63.5 ± 11.1 56.7% vs. 43.3%
62.1 ± 11.3 55.2% vs. 44.8%
64.2 ± 11.0 57.4% vs. 42.6%
84.3% 12.0% 3.0% 0.8%
77.6% 17.1% 4.1% 1.3%
87.4% 9.6% 2.4% 0.6%
47.8% 2.1% 6.1% 22.4% 4.2% 17.5%
49.9% 2.4% 4.9% 22.5% 4.2% 16.0%
46.8% 1.9% 6.7% 22.3% 4.2% 18.1%
68.4% 22.9% 8.7%
72.9% 19.8% 7.3%
66.2% 24.4% 9.3%
4.6% 39.7% 6.9% 44.6% 1.0%
5.6% 41.0% 8.9% 37.4% 1.1%
4.2% 39.1% 6.0% 47.9% 1.0%
3.2%
5.9%
1.8%
Note: Academic centers and community centers had some baseline differences, including (1) an older population in community centers, (2) a larger percentage of black patients in academic centers, (3) a larger percentage of Medicare patients in the community, and (4) a higher percentage of higher comorbidity scale patients in the community. a All differences are statistically significant at p < 0.001. AC, academic center; CC, community center; NA, not available; BAC, bronchioloalveolar carcinoma; NOS, not otherwise specified; CDCC, Charlson/Deyo comorbidity score.
CCs (37.4% versus 47.9% (p < 0.001 with use of the Pearson chi-square test]). We tested whether the following variables varied over time by using a multivariable model with the interaction term facility type by year of diagnosis (model descriptions are provided in Supplementary Table 1): age (pictured in Supplemental Fig. 2A), sex (pictured in Supplemental Fig. 2B), percentage of patients of the African American race (pictured in Supplemental Fig. 2C), histologic type, percentage of patients whose primary oncologist decided not to treat with chemotherapy, duration of time to chemotherapy, distance traveled from home to facility, and percentage of Medicare patients. Comorbidity data were available only after 2003, as a result of which time trends were not applicable to our study period. A significant change over time at p less than 0.001 was detected in only two variables.
December 2018
Stage IV Lung Cancer in ACs versus in CCs
The percentage of African Americans was higher in ACs than in CCs in 1998 (15.7% versus 9.6% [þ6.1%]) and increased by 2010 (18.6% versus 10.6% [þ8.0%]). This difference was significant in our multivariable analysis (p < 0.001); however, there was no significant difference when our multivariable model of 2-year survival between ACs and CCs over time was repeated in the African American subpopulation (p ¼ 0.35). The percentage of patients whose primary oncologist decided not to treat with chemotherapy (including targeted therapies) was lower in ACs than in CCs in 1998 (2.8% versus 3.3% [–0.5%]) and increased by 2010 (6.1% versus 8.0% [–1.9%]); this difference was significant in our multivariable analysis (p < 0.001).
Two-Year Survival Is Higher in ACs than in CCs and Increases More Rapidly in ACs In our data set as a whole, there was an absolute increase in 2-year survival rate of 4.7% (from 9.9% in 1998 to 14.6% in 2010 [99.9% confidence interval: 3.4%–6.0%, p < 0.001 with use of the difference in proportions test]) (Fig. 1). Two-year survival in ACs was higher at baseline in 1998, at which time the difference was 2.3% (11.5% in ACs versus 9.2% in CCs). By 2010, the difference had increased to 4.3% (17.4% in ACs versus 13.1% in CCs). This increasing difference over time was significant in our multivariable OLS regression model: 2-year survival increased at a statistically significant rate in ACs compared with in CCs (interaction term coefficient 0.155 per year [p ¼ 0.0005, 99.9% confidence interval: 0.075%–0.0302%]). Factors associated with higher 2-year survival were female sex, later year of diagnosis, lower age, and having insurance (with the exception of Medicaid) (Fig. 2). The association between insurance and survival was previously reported in the literature.22–24 0.25
2-Year Survival Proporon
0.2
0.15
0.1 AC
CC
Overall
0.05
0 1998
1999
2000
2001
2002
2003 2004 2005 Year of Diagnosis
2006
2007
2008
2009
2010
Figure 1. Two-year survival over time in academic centers (ACs) (red), community-based centers (CCs) (blue), and the overall sample (green). Although there is a general trend of increased 2-year survival over the study period, that trend is particularly apparent in ACs.
1845
Comparison of Survival by Histologic Type Differences in overall survival between ACs and CCs were seen predominantly in adenocarcinoma. The 2-year survival rate for adenocarcinoma (including bronchioloalveolar carcinoma) was 10.2% in 1998 versus 17.3 in 2010, and for squamous cell carcinoma it was 9.9% in 1998 versus 10.1% in 2010 (Fig. 3A). The growing difference between adenocarcinoma and squamous carcinoma was statistically significant in a multivariable model: 2-year survival in adenocarcinoma (including bronchioloalveolar carcinoma) outpaced that of squamous cell carcinoma by 0.59% per year or 5.9% per decade (p < 0.001). When the data set was split into early (1998–2000) and late (2008–2010) treatment groups, the survival advantage in ACs became more pronounced, particularly in adenocarcinoma (Fig. 4). Notably, histology-based differences between ACs and CCs were significant in the early and late treatment groups (log-rank test, p < 0.001 with chi-square of 139 and 3 df in the early group and p < 0.001 with chi-squre of 994 and 3 df in the late group). Within the adenocarcinoma group, there was a 1.8% per-decade increase in 2-year survival in ACs versus in CCs, with a trend toward significance (p ¼ 0.005 [Fig. 3B]). This trend was not present in squamous cell carcinoma (p ¼ 0.54 [Fig. 3C]) or NSCLC not otherwise specified (p ¼ 0.14 [Fig. 3D]).
Other Variables not Present in the OLS Model Did Not Confer an Advantage to ACs Several other possible confounders were analyzed and determined not to confer a clear survival advantage to ACs over time: (1) patients whose treatment plan was to not receive chemotherapy, (2) time between diagnosis and treatment, (3) distance traveled, and (4) CharlsonDeyo comorbidity score. All patients in the data set consisting of patients whose treatment plan was to not receive chemotherapy were prescribed a treatment plan by their oncologist. The percentages of patients with a treatment plan that did not include chemotherapy (including targeted therapies) in 1998 versus in 2010 were 3.25% versus 8.00% in CCs and 2.83% versus 6.08% in ACs. In a multivariable model (variables listed in Supplementary Table 1), treatment at an AC was associated with a 0.17% per-year decrease in the proportion of patients not receiving treatment compared with treatment at a CC (p < 0.001 [Supplementary Fig. 3A]); however this was a relatively small percentage of patients, and although interesting, it is not sufficient to contribute to the overall trend in survival in CCs versus in ACs. With regard to the confounder time between diagnosis and treatment, in a multivariable model (see Supplementary Table 1), treatment at CCs was
1846 Ramalingam et al
Journal of Thoracic Oncology
Vol. 13 No. 12
Age* Diagnosis Year* Female* Histology-Adeno vs LC* Histology-Adeno vs NSCLC NOS* Histology-Adeno vs Other* Histology-Adeno vs Squamous* Insurance-Medicaid vs Uninsured Insurance-Medicare vs Uninsured* Insurance-Other Government vs Uninsured* Insurance-Private vs Uninsured* Insurance-Unknown vs Uninsured* -0.1000
-0.0500
0.0000
0.0500
0.1000
Figure 2. Tornado plot of covariates in multivariable model of 2-year survival. Two-year survival of adenocarcinoma outpaced that of all histologic types except bronchioloalveolar carcinoma) (not pictured here). With the exception of receiving Medicaid, having insurance was associated with favorable 2-year survival compared with being uninsured. Later year of diagnosis and female sex were also favorable. Histologic type bronchioloalveolar carcinoma versus adenocarcinoma was excluded for scaling (range 0.143–0.179), Facility type and intercept terms are also not shown, as their values are outside the interpretable range of –1 to 1, and the interaction term is not represented because it is discussed in the text. *Significant at p < 0.001.
consistently associated with shorter time between diagnosis and treatment without a significant change in the difference between ACs and CCs (p ¼ 0.082 [Supplementary Fig. 3B]). The distance traveled to receive treatment (our third possible confounder) did not change significantly between ACs and CCs in a multivariable model (p ¼ 0.25 [Supplementary Fig. 3C]). Notably, with regard to the fourth confounder, namely, Charlson-Deyo comorbidity score, comorbidity data were present in the data set only after 2003. After restricting the data set to patients with comorbidity score present, we ran our multivariable model in each comorbidity strata (0, 1, and 2) and found no statistically significant differences for our interaction term, providing no clear sign that comorbidity score would change the results of our model in the overall sample. We also repeated our multivariable study model after excluding patients with a comorbidity score of 2 or higher, after which the interaction term was still significant (0.0155 [p ¼ 0.00078]).
Discussion We chose to use the NCDB in this study because it encompasses a large breadth of cases with some, though limited, treatment details, while also maintaining hospital identifiers that allowed both a constant sample of hospitals throughout our study and the comparison of ACs versus CCs.25 We observed that patients with metastatic NSCLC treated at ACs and CCs had improving survival from 1998 to 2010 and that this improvement in outcomes occurred to a greater extent in patients treated in academic medical centers. This survival advantage was predominantly seen in patients with the adenocarcinoma histologic type compared with other histologic types. Because of the asymmetric increase in
treatment options for adenocarcinoma and in the absence of data for specific systemic agents in the NCDB, this histology-based difference was consistent with our hypothesis that treatment within ACs has conferred a survival advantage for patients with NSCLC after the introduction of novel targeted therapies that require a molecularly driven, histologically specific approach. To our knowledge, ours is the first analysis comparing outcomes between ACs and CCs broadly across metastatic NSCLC by using a multivariable model to control for confounding factors.26 Other groups have studied SCLC survival and shown positive trends over time. One group used the Surveillance, Epidemiology, and End Results database to show increasing overall survival in metastatic NSCLC between 1990 and 2005 and also noted the beginnings of a survival advantage in adenocarcinoma over squamous cell carcinoma at their later time point.10 Other groups have shown increasing survival in all stages of NSCLC over time across periods spanning the 1980s to the 2000s11,27; however, these studies were limited by their lack of multivariable analyses to control for confounding factors that could have changed over time. These studies also did not conduct analyses based on type of treatment facility, which have largely been limited in the literature to analyses of practice patterns14,17–19,28 and outcomes in focused subgroups of care16,17,19,29 rather than across an entire category of NSCLC as we have done in this study. Our findings of improved survival in stage IV NSCLC are consistent with the literature.10,11,27 Our study design offers several advantages compared with that in prior studies. Morgensztern et al.10 uses a multivariate analysis of survival curves with selection of predictors similar to ours; however, their study did not allow for control of stage migration; by comparing ACs with CCs, our study
December 2018
Stage IV Lung Cancer in ACs versus in CCs
CC
Year of Diagnosis
2010
2009
2008
2006
2007
2004
Year of Diagnosis
D
0.25 0.2 0.15 0.1 0.05
CC
AC
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
0
2-Year Survival Proporon
C 2-Year Survival Proportion
AC
0 2005
2010
2009
2008
2007
2005
2006
2004
2003
2002
2001
1999
2000
0.05 2003
Adeno
2002
Squamous
0
0.1
2000
0.05
0.15
2001
0.1
0.2
1998
0.15
0.25
1999
0.2
2-Year Survival Proporon
B
0.25
1998
2-Year Survival Proporon
A
1847
0.25 0.2 0.15 0.1 0.05
CC
AC
0
Year of Diagnosis
Year of Diagnosis
Figure 3. Two-year survival for the predominant histologic types. (A) Adenocarcinoma (adeno) survival outpaces squamous cell carcinoma survival, which is relatively unchanged over time. (B) Two-year survival in adeno seems to be increasing faster at academic centers (ACs) than in community-based centers (CCs); however, this trend is not apparent in squamous cell carcinoma (C) or NSCLC not otherwise specified (D).
offers an internal control for these effects and also carries the analysis 5 years further. Furthermore, prior data in the NCDB literature has shown a similar increase in stage IV NSCLC diagnoses between 1998 and 2006 across treatment facility type (from 36% to 41.5% and from 34.3% to 39.3% in each of the CC groups and from 34.4% to 38.3% in the AC group),30 making stage migration an unlikely confounder in a comparison between ACs and CCs. Our multivariable analysis was also able to control for confounders for which two other groups11,27 did not control across time. We have further developed and confirmed a survival advantage arising in adenocarcinoma over time and shown a trend toward significance in treatment of adenocarcinoma at ACs versus at CCs. More generally, whereas other studies have highlighted differences between ACs and CCs,14,16–19,28,29 ours is the first to do so across metastatic NSCLC rather than in subgroups of lower-stage malignancy. Quality control was central to our analysis. In our analysis model, we controlled for potential confounding factors that were different between ACs and CCs, including age and female sex, which is associated with a higher prevalence of EGFR mutations.31 Other variables not included in our model also did not change in a way that would confer a survival advantage to ACs, namely, African American race (Supplementary Fig. 2C), time between diagnosis and the start of chemotherapy (see Supplementary Fig. 3B), and distance traveled to receive care (see Supplementary Fig. 3C). African American
patients have been shown in the literature to have decreased survival outcomes compared with white and Hispanic populations,32,33 and the percentage of African American patients actually increased in ACs over time; furthermore, repeating our 2-year survival model in the African American subpopulation did not show a significant change over time. There was an increasing percentage of patients in CCs versus in ACs who had a treatment decision to not treat with systemic therapy, but the absolute percentage difference was negligible, encompassing only 2% of the sample in 2010 (see Supplementary Fig. 3A). Although we could not control for comorbidity score, as this was not recorded in the NCDB until 2003, we did show that when we partitioned post-2003 data by comorbidity score (0, 1, and 2þ), there was no statistically significant difference between ACs and CCs in any strata, and when we removed patients with a comorbidity score of 2 or higher, our multivariable model of 2-year survival maintained significance for our interaction term, indicating that comorbidity score is less likely to be confounding our results.
Limitations Unobserved confounders may bias study results; however, we attempted to mitigate the effect of unmeasured confounders by structuring our analysis as a comparison between ACs and CCs by year over time rather than by an overall comparison in the entire database. Although a stronger trend in adenocarcinoma supports the idea of treatment-related effect, the lack of data regarding specific
1848 Ramalingam et al
Journal of Thoracic Oncology
Vol. 13 No. 12
Figure 4. Survival data for early and late periods in the study, by facility and histologic type: adenocarcinoma (including bronchioloalveolar carcinoma) versus nonadenocarcinoma (including large cell, NSCLC not otherwise specified, squamous cell carcinoma, and other). Red lines indicate nonadenocarcinoma at community-based centers, green lines indicate nonadenocarcinoma at community-based centers, blue lines indicate nonadenocarcinoma at academic centers, and purple lines indicate adenocarcinoma at academic centers. (A) In the group of patients whose cancer was diagnosed between 1998 and 2000, there is a statistically significant difference between the four groups (log-rank test, p < 0.001), but clinically there is not a clear benefit over 60 months. (B) In the group of patients whose cancer was diagnosed between 2008 and 2010, survival of adenocarcinoma in academic centers consistently outpaces that in the other groups over 60 months (log-rank test, p < 0.001).
agents used in treatment (especially in the first line), and the lack of data regarding molecular testing and extent/site of metastatic disease limit the implications of this study. It is possible that there were differences in the percentage of patients treated in ACs with actionable mutations, artificially inflating survival; though mutational status is not recorded in the NCDB, one surrogate (age) did not change significantly between ACs and CCs (Supplementary Fig. 2A).
from 1998 to 2010, during which time there was an increase in both the number of chemotherapy agents approved and complexity of treatment in terms of histologic type and molecular-based treatment strategies. With the more recent introduction of additional targeted agents and immunotherapy, additional study is needed to determine whether this trend has continued. To reduce the disparity in patient outcomes, strategies to accelerate these advances in CCs may be needed.
Conclusion and Further Directions Within the aforementioned limitations, in this study we have shown an increasing survival difference in metastatic NSCLC between ACs and CCs in the period
Supplementary Data Note: To access the supplementary material accompanying this article, visit the online version of the Journal of
December 2018
Thoracic Oncology at www.jto.org and at https://doi. org/10.1016/j.jtho.2018.09.007.
References 1. National Cancer Institute. Surveillance, Epidemiology, and End Results Program. Fast stats. 2017; https://seer. cancer.gov/faststats/. Accessed October 3, 2017. 2. Chen VW, Ruiz BA, Hsieh MC, Wu XC, Ries LA, Lewis DR. Analysis of stage and clinical/prognostic factors for lung cancer from SEER registries: AJCC staging and collaborative stage data collection system. Cancer. 2014;120(suppl 23):3781–3792. 3. National Cancer Institute. Drugs approved for lung cancer. https://www.cancer.gov/about-cancer/treatment/ drugs/lung. Accessed September 9, 2017. 4. Hanna N, Shepherd FA, Fossella FV, et al. Randomized phase III trial of pemetrexed versus docetaxel in patients with non-small-cell lung cancer previously treated with chemotherapy. J Clin Oncol. 2004;22:1589–1597. 5. Scagliotti G, Hanna N, Fossella F, et al. The differential efficacy of pemetrexed according to NSCLC histology: a review of two phase III studies. Oncologist. 2009;14: 253–263. 6. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med. 2004;350:2129–2139. 7. Sandler A, Gray R, Perry MC, et al. Paclitaxel-carboplatin alone or with bevacizumab for non-small-cell lung cancer. N Engl J Med. 2006;355:2542–2550. 8. Schiller JH, Harrington D, Belani CP, et al. Comparison of four chemotherapy regimens for advanced non-smallcell lung cancer. N Engl J Med. 2002;346:92–98. 9. Del Rivero J, Enewold L, Thomas A. Metastatic lung cancer in the age of targeted therapy: improving long-term survival. Transl Lung Cancer Res. 2016;5: 727–730. 10. Morgensztern D, Waqar S, Subramanian J, Gao F, Govindan R. Improving survival for stage IV non-small cell lung cancer: a surveillance, epidemiology, and end results survey from 1990 to 2005. J Thorac Oncol. 2009;4:1524–1529. 11. Xia W, Yu X, Mao Q, et al. Improvement of survival for non-small cell lung cancer over time. Onco Targets Ther. 2017;10:4295–4303. 12. Shepherd FA, Rodrigues Pereira J, Ciuleanu T, et al. Erlotinib in previously treated non-small-cell lung cancer. N Engl J Med. 2005;353:123–132. 13. Maemondo M, Inoue A, Kobayashi K, et al. Gefitinib or chemotherapy for non-small-cell lung cancer with mutated EGFR. N Engl J Med. 2010;362:2380–2388. 14. Koshy M, Malik R, Spiotto M, Mahmood U, Weichselbaum R, Sher D. Disparities in treatment of patients with inoperable stage I non-small cell lung cancer: a population-based analysis. J Thorac Oncol. 2015;10:264–271. 15. Patel AP, Crabtree TD, Bell JM, et al. National patterns of care and outcomes after combined modality therapy for stage IIIA non-small-cell lung cancer. J Thorac Oncol. 2014;9:612–621.
Stage IV Lung Cancer in ACs versus in CCs
1849
16. Wang EH, Rutter CE, Corso CD, et al. Patients Selected for definitive concurrent chemoradiation at high-volume facilities achieve improved survival in stage III nonsmall-cell lung cancer. J Thorac Oncol. 2015;10:937–943. 17. Rajaram R, Paruch JL, Mohanty S, et al. Patterns and predictors of chemotherapy use for resected non-small cell lung cancer. Ann Thorac Surg. 2016;101:533–540. 18. Samson P, Patel A, Crabtree TD, et al. Multidisciplinary treatment for stage IIIA non-small cell lung cancer: does institution type matter? Ann Thorac Surg. 2015;100: 1773–1779. 19. Salazar MC, Rosen JE, Wang Z, et al. Association of delayed adjuvant chemotherapy with survival after lung cancer surgery. JAMA Oncol. 2017;3:610–619. 20. Bilimoria KY, Stewart AK, Winchester DP, Ko CY. The National Cancer Data Base: a powerful initiative to improve cancer care in the United States. Ann Surg Oncol. 2008;15:683–690. 21. American College of Surgeons. About cancer program categories. https://www.facs.org/quality-programs/ cancer/coc /apply/categories. Accessed June 15, 2017. 22. Varlotto JM, Voland R, Mckie K, et al. Population-based differences in the outcome and presentation of lung cancer patients based upon racial, histologic, and economic factors in all lung patients and those with metastatic disease. Cancer Med. 2018;7:1211–1220. 23. Parikh-Patel A, Morris CR, Kizer KW. Disparities in quality of cancer care: the role of health insurance and population demographics. Medicine (Baltimore). 2017;96:e9125. 24. Ellis L, Canchola AJ, Spiegel D, Ladabaum U, Haile R, Gomez SL. Trends in cancer survival by health insurance status in California from 1997 to 2014. JAMA Oncol. 2018;4:317–323. 25. Mohanty S, Bilimoria KY. Comparing national cancer registries: the National Cancer Data Base (NCDB) and the Surveillance, Epidemiology, and End Results (SEER) program. J Surg Oncol. 2014;109:629–630. 26. Ramalingam S, Dinan M, Crawford J. Increasing survival in stage IV NSCLC in academic versus community based centers in the National Cancer Database. Abstract presented at: Multidisciplinary Symposium in Thoracic Oncology. September 15, 2017; Chicago, IL. 27. Wang S, Sun T, Sun H, et al. Survival improvement in patients with non-small cell lung cancer between 1983 and 2012: analysis of the Surveillance, Epidemiology, and End Results database. Tumour Biol. 2017;39, 1010428317691677. 28. Sher DJ, Liptay MJ, Fidler MJ. Prevalence and predictors of neoadjuvant therapy for stage IIIA non-small cell lung cancer in the National Cancer Database: importance of socioeconomic status and treating institution. Int J Radiat Oncol Biol Phys. 2014;89:303–312. 29. Sher DJ, Koshy M, Liptay MJ, Fidler MJ. Influence of conformal radiotherapy technique on survival after chemoradiotherapy for patients with stage III non-small cell lung cancer in the National Cancer Data Base. Cancer. 2014;120:2060–2068. 30. Morgensztern D, Ng SH, Gao F, Govindan R. Trends in stage distribution for patients with non-small cell lung cancer: a National Cancer Database survey. J Thorac Oncol. 2010;5:29–33.
1850 Ramalingam et al 31. Midha A, Dearden S, Mccormack R. EGFR mutation incidence in non-small-cell lung cancer of adenocarcinoma histology: a systematic review and global map by ethnicity (mutMapII). Am J Cancer Res. 2015;5: 2892–2911. 32. Tannenbaum SL, Koru-Sengul T, Zhao W, Miao F, Byrne MM. Survival disparities in non-small cell lung
Journal of Thoracic Oncology
Vol. 13 No. 12
cancer by race, ethnicity, and socioeconomic status. Cancer J. 2014;20:237–245. 33. U.S. Centers for Disease Control and Prevention. United States cancer statistics: data visualizations. Rates of new cancers: lung and bronchus, United States, 2015. https://www.cdc.gov/cancer/lung/statistics/race.htm. Accessed October 8, 2017.
Survival Comparison in Patients with Stage IV Lung Cancer in Academic versus Community Centers in the United States Sendhilnathan Ramalingam, MD, Michaela A. Dinan, PhD, Jeffrey Crawford, MD* Duke Cancer Institute, Duke University, Durham, North Carolina Received 24 August 2018; revised 28 August 2018; accepted 5 September 2018 Available online - 8 October 2018
ABSTRACT Introduction: Although metastatic NSCLC is widely treated in both academic centers (ACs) and community-based centers (CCs), it is unclear whether outcomes are similar across both settings. A growing variety of chemotherapies and targeted agents for an increasingly histology- and molecularbased treatment strategy could provide an advantage to patients treated in ACs. Using the National Cancer Database, we investigated whether treatment at ACs was associated with a survival advantage in metastatic NSCLC. Methods: We conducted a retrospective analysis of the National Cancer Database after the introduction of novel NSCLC chemotherapy agents from 1998 to 2010. The primary outcome was 2-year survival, which was analyzed by using a multivariable regression model controlling for age, year of diagnosis, sex, primary payer, histologic type, facility type (AC versus CC), and an interaction term allowing a time-based comparison of survival between ACs and CCs. Alpha was set to 0.001 because of the size of the data set. Results: There were 193,279 patients included in this study. The percentage of patients achieving 2-year survival was higher in ACs versus in CCs in 1998 (11.5% versus 9.2% [þ2.3%]), and by 2010, the difference had increased to 17.4% versus 13.1% (þ4.3%). Multivariable analysis confirmed a significant relative increase in 2-year survival associated with ACs versus with CCs from 1998 to 2010 (p ¼ 0.0005). A histology-dependent survival difference was also noted in adenocarcinoma versus in squamous cell carcinoma (10.2% versus 9.9% in 1998 [þ0.3%], increasing to 17.3% versus 10.1% in 2010 [þ7.2%]). Adenocarcinoma survival also varied by treatment facility, with the difference in 2-year survival in ACs versus in CCs increasing from 12.3% versus 9.1% (þ3.2%) in 1998 to 20.5% versus 15.5% (þ5%) in 2010, with a trend toward significance in our multivariable model (p ¼ 0.005). Conclusions: A greater increase in survival was noted in ACs than in CCs over this time period, and it was particularly pronounced among patients with adenocarcinoma
Journal of Thoracic Oncology
Vol. 13 No. 12: 1842-1850
versus in those with squamous cell carcinoma. Given the known advances in adenocarcinoma treatment compared with in squamous cell lung cancer over this time period, our study suggests that potential treatment-related disparities may exist between ACs and CCs. Further study will be needed to validate this disparity in health care and address opportunities to improve survival in patients with stage IV NSCLC across treatment settings. 2018 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. Keywords: Survival; Non-small cell lung cancer; Disparities; Population health
Introduction Lung cancer remains the deadliest malignancy in the United States, accounting for the largest number of cancer-related deaths1 and with roughly half of patients with NSCLC presenting with metastatic disease.2 Treatment for metastatic NSCLC has evolved dramatically over the past 20 years, with increased emphasis on histology- and molecularly driven treatment algorithms. The number of new treatment options that continue to increase the availability of a broader array of treatments began between 1998 and 2010 with the approval
*Corresponding author. Disclosure: The authors declare no conflict of interest. Address for correspondence: Jeffrey Crawford, MD, Duke Cancer Institute, Duke University Medical Center, Department of Medicine, Box 3476, Room 2592, Morris Building, Durham, NC 27710. E-mail: [email protected] ª 2018 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. ISSN: 1556-0864 https://doi.org/10.1016/j.jtho.2018.09.007
December 2018
of cytotoxic agents such as pemetrexed3–5 and chemotherapy doublets, as well as targeted agents against mutations in EGFR6 and bevacizumab.3,7 It was in the nonsquamous subgroup in particular that we recognized that these new therapies are associated with improved outcomes. Before the introduction of these new therapies, advanced NSCLC was treated with platinum-based chemotherapy, with response rates of only about 19% and median overall survival of 7.9 months8,9; however, during our study period national-level database analyses have indicated a trend of increasing survival in advanced NSCLC.10,11 In sum, treatment of advanced NSCLC became more complex in two major ways. First, the histologic type of the tumor became paramount with the addition of pemetrexed in 20043,4 and the realization of its differential advantage in nonsquamous NSCLC.4,5 Second, the introduction of targeted therapies began an era of more complex medical oncology that was dependent on the mutational profile of the tumor.6,12,13 Oncologists increasingly needed to know which profiling to do and which agents were most applicable. A growing body of literature in NSCLC has shown that treatment choices and outcomes in localized NSCLC vary between academic centers (ACs) and community-based centers (CCs).14,15–19 We hypothesized that ACs may have more availability of lung cancer–specific pathologists and availability for molecular testing to manage the growing complexity of systemic therapies for metastatic NSCLC; that this would be manifested as a growing difference in 2-year survival rates between ACs and CC; and that if survival differences exist, there could be disparities of care accounting for these differences.
Methods Data Set and Study Population The National Cancer Database (NCDB) was used to conduct this study. Initially established in 1989, the NCDB currently encompasses approximately 1500 treatment facilities and is estimated to capture approximately 70% of malignancies diagnosed in the United States. Data collected include patient demographics; grade, location, and stage of the relevant malignancy; timing of therapies; types of therapies (including surgery, radiation, and chemotherapy, though not the identity of specific systemic therapies administered); 30-day mortality; and time to last contact. Geographically, the contributing facilities are located in 49 of 50 states as well as in Puerto Rico. About 20% of these are ACs.20 Our data set contained data from 1998 to 2012, allowing for 2-year survival analysis between 1998 and 2010. The NCDB has been used to study outcomes in cancer, including NSCLC. Predominantly, this includes studies of
Stage IV Lung Cancer in ACs versus in CCs
1843
outcomes in surgery and radiation therapy,15,16 as well as chemotherapy in the adjuvant/neodjuvant setting.17–19
Inclusion/Exclusion Criteria The NCDB user file included patients with NSCLC diagnosed between 1998 to 2010. This study was limited to patients with a diagnosis of stage IV NSCLC between 1998 and 2010 at a facility reporting to the NCDB throughout the study period. Patients were excluded if treatment facility type and/or duration of follow-up was missing or treatment was done entirely at another facility. Patients were also excluded if NSCLC was not their first malignancy and if they did not have a treatment decision or did not go through with the treatment prescribed by their physician (treatment was defined as chemotherapy/immunotherapy or the decision not to treat with either). Three types of facility designations were present in the data set: community cancer program, comprehensive community cancer program, and academic/research program (AC). Per the NCDB definition, both community cancer programs and comprehensive community cancer programs provide a range of diagnostic and treatment methods and participate in or refer patients for clinical trials; the distinction between the two designations is a volume-based metric (<500 versus >500 cases/y), as a result of which our data analysis these groups were combined to form the CC designation.21 Academic/ research programs must have more than 500 new cases per year and are associated either with a National Cancer Institute–designated comprehensive cancer center or with residencies in four areas, including internal medicine and general surgery. The primary end point was 2-year survival, allowing us to balance the number of years analyzed in the study while at the same time capturing a large percentage of mortality. Two-year survival was defined as the time between diagnosis and last contact of 24 months or longer. In the NCDB, these data are entered by certified tumor registrars at the relevant treatment site.
Statistical Analysis Multivariable OLS Regression Model. Comparison of the primary end point between ACs and CCs was done by using a multivariable ordinary least squares (OLS) regression model with an interaction term between facility type and year of diagnosis to allow for a continuous, time-based comparison and to help address the effect of baseline differences in the populations of patients in ACs and CCs. Control variables were year of diagnosis, type of treatment facility, age, sex, primary payer status, tumor histologic type, and the interaction term facility type by year of diagnosis. The interaction
1844 Ramalingam et al
term was essential for our analysis; therefore, a linear probability model using OLS regression was favored over logistic regression to allow for generalizable interpretation of interaction coefficients. To evaluate confounders, multivariable models were also used to test changes in several variables over time between ACs and CCs; these variables include age, sex, African American race, histologic type, treatment decision, duration of time to chemotherapy, and distance of travel for treatment (based on zip code of residence and zip code of treating hospital). Predictor variables used in each model are listed in Supplementary Table 1. Charlson-Deyo comorbidity index was not included in the primary outcome model because it was present in the NCDB only after 2003. Presence of brain metastasis at diagnosis could not be included because it was present in the data set only in 2010 and afterward. Determination of baseline differences between ACs and CCs for categorical variables was done with the Pearson chi-square test. Comparison of absolute difference between time points was done with either the two-sample differences in a means test or the difference in proportions test without continuity correction (owing to the size of our sample). Alpha was set to 0.001 for all comparisons because of the large sample size. All analyses were done with R software (R Foundation for Statistical Computing, Vienna, Austria). We applied for and obtained access to the 2012 NCDB participant user file with approval from our institutional review board.
Results Description of Data Set The final data set after satisfaction of the inclusion/ exclusion criteria included 193,279 patients (Supplementary Fig. 1) with clinical or pathologic stage IV NSCLC. All baseline differences between ACs and CCs were statistically significant at p < 0.001 with use of the Pearson chi-square test for all variables except age, for which the two-sample difference in means test was used (Table 1); OLS regression models with interaction terms (facility type year of diagnosis) were used to determine whether differences between ACs and CCs changed over time (see Supplementary Table 1). In ACs versus in CCs, the mean age was 62.1 versus 64.2 years (p < 0.001 [two-sample difference in means test]). The percentage of black patients was higher in ACs than in CCs (17.1% versus 9.6%, respectively [p < 0.001 with use of the Pearson chi-square test). The percentage of patients with Charlson-Deyo comorbidity scores of 2 or higher was lower in ACs than in CCs (7.3% versus 9.3% [p < 0.001 with use of the Pearson chi-square test]). The percentage of patients insured by Medicare was lower in ACs than in
Journal of Thoracic Oncology
Vol. 13 No. 12
Table 1. Demographics Data for the Overall Sample, ACs, and CCs
Characteristic Mean age ± SD, ya Sex, males vs. femalesa Ethnicitya White Black Other NA NSCLC typea Adenocarcinoma BAC Large cell NSCLC NOS Other Squamous Comorbiditiesa CDCC 0 CDCC 1 CDCC 2 Insurance statusa Uninsured Private Medicaid Medicare Other government Unknown
Overall Sample ACs CCs (N ¼ 193,279) (n ¼ 61,874) (n ¼ 131,405) 63.5 ± 11.1 56.7% vs. 43.3%
62.1 ± 11.3 55.2% vs. 44.8%
64.2 ± 11.0 57.4% vs. 42.6%
84.3% 12.0% 3.0% 0.8%
77.6% 17.1% 4.1% 1.3%
87.4% 9.6% 2.4% 0.6%
47.8% 2.1% 6.1% 22.4% 4.2% 17.5%
49.9% 2.4% 4.9% 22.5% 4.2% 16.0%
46.8% 1.9% 6.7% 22.3% 4.2% 18.1%
68.4% 22.9% 8.7%
72.9% 19.8% 7.3%
66.2% 24.4% 9.3%
4.6% 39.7% 6.9% 44.6% 1.0%
5.6% 41.0% 8.9% 37.4% 1.1%
4.2% 39.1% 6.0% 47.9% 1.0%
3.2%
5.9%
1.8%
Note: Academic centers and community centers had some baseline differences, including (1) an older population in community centers, (2) a larger percentage of black patients in academic centers, (3) a larger percentage of Medicare patients in the community, and (4) a higher percentage of higher comorbidity scale patients in the community. a All differences are statistically significant at p < 0.001. AC, academic center; CC, community center; NA, not available; BAC, bronchioloalveolar carcinoma; NOS, not otherwise specified; CDCC, Charlson/Deyo comorbidity score.
CCs (37.4% versus 47.9% (p < 0.001 with use of the Pearson chi-square test]). We tested whether the following variables varied over time by using a multivariable model with the interaction term facility type by year of diagnosis (model descriptions are provided in Supplementary Table 1): age (pictured in Supplemental Fig. 2A), sex (pictured in Supplemental Fig. 2B), percentage of patients of the African American race (pictured in Supplemental Fig. 2C), histologic type, percentage of patients whose primary oncologist decided not to treat with chemotherapy, duration of time to chemotherapy, distance traveled from home to facility, and percentage of Medicare patients. Comorbidity data were available only after 2003, as a result of which time trends were not applicable to our study period. A significant change over time at p less than 0.001 was detected in only two variables.
December 2018
Stage IV Lung Cancer in ACs versus in CCs
The percentage of African Americans was higher in ACs than in CCs in 1998 (15.7% versus 9.6% [þ6.1%]) and increased by 2010 (18.6% versus 10.6% [þ8.0%]). This difference was significant in our multivariable analysis (p < 0.001); however, there was no significant difference when our multivariable model of 2-year survival between ACs and CCs over time was repeated in the African American subpopulation (p ¼ 0.35). The percentage of patients whose primary oncologist decided not to treat with chemotherapy (including targeted therapies) was lower in ACs than in CCs in 1998 (2.8% versus 3.3% [–0.5%]) and increased by 2010 (6.1% versus 8.0% [–1.9%]); this difference was significant in our multivariable analysis (p < 0.001).
Two-Year Survival Is Higher in ACs than in CCs and Increases More Rapidly in ACs In our data set as a whole, there was an absolute increase in 2-year survival rate of 4.7% (from 9.9% in 1998 to 14.6% in 2010 [99.9% confidence interval: 3.4%–6.0%, p < 0.001 with use of the difference in proportions test]) (Fig. 1). Two-year survival in ACs was higher at baseline in 1998, at which time the difference was 2.3% (11.5% in ACs versus 9.2% in CCs). By 2010, the difference had increased to 4.3% (17.4% in ACs versus 13.1% in CCs). This increasing difference over time was significant in our multivariable OLS regression model: 2-year survival increased at a statistically significant rate in ACs compared with in CCs (interaction term coefficient 0.155 per year [p ¼ 0.0005, 99.9% confidence interval: 0.075%–0.0302%]). Factors associated with higher 2-year survival were female sex, later year of diagnosis, lower age, and having insurance (with the exception of Medicaid) (Fig. 2). The association between insurance and survival was previously reported in the literature.22–24 0.25
2-Year Survival Proporon
0.2
0.15
0.1 AC
CC
Overall
0.05
0 1998
1999
2000
2001
2002
2003 2004 2005 Year of Diagnosis
2006
2007
2008
2009
2010
Figure 1. Two-year survival over time in academic centers (ACs) (red), community-based centers (CCs) (blue), and the overall sample (green). Although there is a general trend of increased 2-year survival over the study period, that trend is particularly apparent in ACs.
1845
Comparison of Survival by Histologic Type Differences in overall survival between ACs and CCs were seen predominantly in adenocarcinoma. The 2-year survival rate for adenocarcinoma (including bronchioloalveolar carcinoma) was 10.2% in 1998 versus 17.3 in 2010, and for squamous cell carcinoma it was 9.9% in 1998 versus 10.1% in 2010 (Fig. 3A). The growing difference between adenocarcinoma and squamous carcinoma was statistically significant in a multivariable model: 2-year survival in adenocarcinoma (including bronchioloalveolar carcinoma) outpaced that of squamous cell carcinoma by 0.59% per year or 5.9% per decade (p < 0.001). When the data set was split into early (1998–2000) and late (2008–2010) treatment groups, the survival advantage in ACs became more pronounced, particularly in adenocarcinoma (Fig. 4). Notably, histology-based differences between ACs and CCs were significant in the early and late treatment groups (log-rank test, p < 0.001 with chi-square of 139 and 3 df in the early group and p < 0.001 with chi-squre of 994 and 3 df in the late group). Within the adenocarcinoma group, there was a 1.8% per-decade increase in 2-year survival in ACs versus in CCs, with a trend toward significance (p ¼ 0.005 [Fig. 3B]). This trend was not present in squamous cell carcinoma (p ¼ 0.54 [Fig. 3C]) or NSCLC not otherwise specified (p ¼ 0.14 [Fig. 3D]).
Other Variables not Present in the OLS Model Did Not Confer an Advantage to ACs Several other possible confounders were analyzed and determined not to confer a clear survival advantage to ACs over time: (1) patients whose treatment plan was to not receive chemotherapy, (2) time between diagnosis and treatment, (3) distance traveled, and (4) CharlsonDeyo comorbidity score. All patients in the data set consisting of patients whose treatment plan was to not receive chemotherapy were prescribed a treatment plan by their oncologist. The percentages of patients with a treatment plan that did not include chemotherapy (including targeted therapies) in 1998 versus in 2010 were 3.25% versus 8.00% in CCs and 2.83% versus 6.08% in ACs. In a multivariable model (variables listed in Supplementary Table 1), treatment at an AC was associated with a 0.17% per-year decrease in the proportion of patients not receiving treatment compared with treatment at a CC (p < 0.001 [Supplementary Fig. 3A]); however this was a relatively small percentage of patients, and although interesting, it is not sufficient to contribute to the overall trend in survival in CCs versus in ACs. With regard to the confounder time between diagnosis and treatment, in a multivariable model (see Supplementary Table 1), treatment at CCs was
1846 Ramalingam et al
Journal of Thoracic Oncology
Vol. 13 No. 12
Age* Diagnosis Year* Female* Histology-Adeno vs LC* Histology-Adeno vs NSCLC NOS* Histology-Adeno vs Other* Histology-Adeno vs Squamous* Insurance-Medicaid vs Uninsured Insurance-Medicare vs Uninsured* Insurance-Other Government vs Uninsured* Insurance-Private vs Uninsured* Insurance-Unknown vs Uninsured* -0.1000
-0.0500
0.0000
0.0500
0.1000
Figure 2. Tornado plot of covariates in multivariable model of 2-year survival. Two-year survival of adenocarcinoma outpaced that of all histologic types except bronchioloalveolar carcinoma) (not pictured here). With the exception of receiving Medicaid, having insurance was associated with favorable 2-year survival compared with being uninsured. Later year of diagnosis and female sex were also favorable. Histologic type bronchioloalveolar carcinoma versus adenocarcinoma was excluded for scaling (range 0.143–0.179), Facility type and intercept terms are also not shown, as their values are outside the interpretable range of –1 to 1, and the interaction term is not represented because it is discussed in the text. *Significant at p < 0.001.
consistently associated with shorter time between diagnosis and treatment without a significant change in the difference between ACs and CCs (p ¼ 0.082 [Supplementary Fig. 3B]). The distance traveled to receive treatment (our third possible confounder) did not change significantly between ACs and CCs in a multivariable model (p ¼ 0.25 [Supplementary Fig. 3C]). Notably, with regard to the fourth confounder, namely, Charlson-Deyo comorbidity score, comorbidity data were present in the data set only after 2003. After restricting the data set to patients with comorbidity score present, we ran our multivariable model in each comorbidity strata (0, 1, and 2) and found no statistically significant differences for our interaction term, providing no clear sign that comorbidity score would change the results of our model in the overall sample. We also repeated our multivariable study model after excluding patients with a comorbidity score of 2 or higher, after which the interaction term was still significant (0.0155 [p ¼ 0.00078]).
Discussion We chose to use the NCDB in this study because it encompasses a large breadth of cases with some, though limited, treatment details, while also maintaining hospital identifiers that allowed both a constant sample of hospitals throughout our study and the comparison of ACs versus CCs.25 We observed that patients with metastatic NSCLC treated at ACs and CCs had improving survival from 1998 to 2010 and that this improvement in outcomes occurred to a greater extent in patients treated in academic medical centers. This survival advantage was predominantly seen in patients with the adenocarcinoma histologic type compared with other histologic types. Because of the asymmetric increase in
treatment options for adenocarcinoma and in the absence of data for specific systemic agents in the NCDB, this histology-based difference was consistent with our hypothesis that treatment within ACs has conferred a survival advantage for patients with NSCLC after the introduction of novel targeted therapies that require a molecularly driven, histologically specific approach. To our knowledge, ours is the first analysis comparing outcomes between ACs and CCs broadly across metastatic NSCLC by using a multivariable model to control for confounding factors.26 Other groups have studied SCLC survival and shown positive trends over time. One group used the Surveillance, Epidemiology, and End Results database to show increasing overall survival in metastatic NSCLC between 1990 and 2005 and also noted the beginnings of a survival advantage in adenocarcinoma over squamous cell carcinoma at their later time point.10 Other groups have shown increasing survival in all stages of NSCLC over time across periods spanning the 1980s to the 2000s11,27; however, these studies were limited by their lack of multivariable analyses to control for confounding factors that could have changed over time. These studies also did not conduct analyses based on type of treatment facility, which have largely been limited in the literature to analyses of practice patterns14,17–19,28 and outcomes in focused subgroups of care16,17,19,29 rather than across an entire category of NSCLC as we have done in this study. Our findings of improved survival in stage IV NSCLC are consistent with the literature.10,11,27 Our study design offers several advantages compared with that in prior studies. Morgensztern et al.10 uses a multivariate analysis of survival curves with selection of predictors similar to ours; however, their study did not allow for control of stage migration; by comparing ACs with CCs, our study
December 2018
Stage IV Lung Cancer in ACs versus in CCs
CC
Year of Diagnosis
2010
2009
2008
2006
2007
2004
Year of Diagnosis
D
0.25 0.2 0.15 0.1 0.05
CC
AC
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
0
2-Year Survival Proporon
C 2-Year Survival Proportion
AC
0 2005
2010
2009
2008
2007
2005
2006
2004
2003
2002
2001
1999
2000
0.05 2003
Adeno
2002
Squamous
0
0.1
2000
0.05
0.15
2001
0.1
0.2
1998
0.15
0.25
1999
0.2
2-Year Survival Proporon
B
0.25
1998
2-Year Survival Proporon
A
1847
0.25 0.2 0.15 0.1 0.05
CC
AC
0
Year of Diagnosis
Year of Diagnosis
Figure 3. Two-year survival for the predominant histologic types. (A) Adenocarcinoma (adeno) survival outpaces squamous cell carcinoma survival, which is relatively unchanged over time. (B) Two-year survival in adeno seems to be increasing faster at academic centers (ACs) than in community-based centers (CCs); however, this trend is not apparent in squamous cell carcinoma (C) or NSCLC not otherwise specified (D).
offers an internal control for these effects and also carries the analysis 5 years further. Furthermore, prior data in the NCDB literature has shown a similar increase in stage IV NSCLC diagnoses between 1998 and 2006 across treatment facility type (from 36% to 41.5% and from 34.3% to 39.3% in each of the CC groups and from 34.4% to 38.3% in the AC group),30 making stage migration an unlikely confounder in a comparison between ACs and CCs. Our multivariable analysis was also able to control for confounders for which two other groups11,27 did not control across time. We have further developed and confirmed a survival advantage arising in adenocarcinoma over time and shown a trend toward significance in treatment of adenocarcinoma at ACs versus at CCs. More generally, whereas other studies have highlighted differences between ACs and CCs,14,16–19,28,29 ours is the first to do so across metastatic NSCLC rather than in subgroups of lower-stage malignancy. Quality control was central to our analysis. In our analysis model, we controlled for potential confounding factors that were different between ACs and CCs, including age and female sex, which is associated with a higher prevalence of EGFR mutations.31 Other variables not included in our model also did not change in a way that would confer a survival advantage to ACs, namely, African American race (Supplementary Fig. 2C), time between diagnosis and the start of chemotherapy (see Supplementary Fig. 3B), and distance traveled to receive care (see Supplementary Fig. 3C). African American
patients have been shown in the literature to have decreased survival outcomes compared with white and Hispanic populations,32,33 and the percentage of African American patients actually increased in ACs over time; furthermore, repeating our 2-year survival model in the African American subpopulation did not show a significant change over time. There was an increasing percentage of patients in CCs versus in ACs who had a treatment decision to not treat with systemic therapy, but the absolute percentage difference was negligible, encompassing only 2% of the sample in 2010 (see Supplementary Fig. 3A). Although we could not control for comorbidity score, as this was not recorded in the NCDB until 2003, we did show that when we partitioned post-2003 data by comorbidity score (0, 1, and 2þ), there was no statistically significant difference between ACs and CCs in any strata, and when we removed patients with a comorbidity score of 2 or higher, our multivariable model of 2-year survival maintained significance for our interaction term, indicating that comorbidity score is less likely to be confounding our results.
Limitations Unobserved confounders may bias study results; however, we attempted to mitigate the effect of unmeasured confounders by structuring our analysis as a comparison between ACs and CCs by year over time rather than by an overall comparison in the entire database. Although a stronger trend in adenocarcinoma supports the idea of treatment-related effect, the lack of data regarding specific
1848 Ramalingam et al
Journal of Thoracic Oncology
Vol. 13 No. 12
Figure 4. Survival data for early and late periods in the study, by facility and histologic type: adenocarcinoma (including bronchioloalveolar carcinoma) versus nonadenocarcinoma (including large cell, NSCLC not otherwise specified, squamous cell carcinoma, and other). Red lines indicate nonadenocarcinoma at community-based centers, green lines indicate nonadenocarcinoma at community-based centers, blue lines indicate nonadenocarcinoma at academic centers, and purple lines indicate adenocarcinoma at academic centers. (A) In the group of patients whose cancer was diagnosed between 1998 and 2000, there is a statistically significant difference between the four groups (log-rank test, p < 0.001), but clinically there is not a clear benefit over 60 months. (B) In the group of patients whose cancer was diagnosed between 2008 and 2010, survival of adenocarcinoma in academic centers consistently outpaces that in the other groups over 60 months (log-rank test, p < 0.001).
agents used in treatment (especially in the first line), and the lack of data regarding molecular testing and extent/site of metastatic disease limit the implications of this study. It is possible that there were differences in the percentage of patients treated in ACs with actionable mutations, artificially inflating survival; though mutational status is not recorded in the NCDB, one surrogate (age) did not change significantly between ACs and CCs (Supplementary Fig. 2A).
from 1998 to 2010, during which time there was an increase in both the number of chemotherapy agents approved and complexity of treatment in terms of histologic type and molecular-based treatment strategies. With the more recent introduction of additional targeted agents and immunotherapy, additional study is needed to determine whether this trend has continued. To reduce the disparity in patient outcomes, strategies to accelerate these advances in CCs may be needed.
Conclusion and Further Directions Within the aforementioned limitations, in this study we have shown an increasing survival difference in metastatic NSCLC between ACs and CCs in the period
Supplementary Data Note: To access the supplementary material accompanying this article, visit the online version of the Journal of
December 2018
Thoracic Oncology at www.jto.org and at https://doi. org/10.1016/j.jtho.2018.09.007.
References 1. National Cancer Institute. Surveillance, Epidemiology, and End Results Program. Fast stats. 2017; https://seer. cancer.gov/faststats/. Accessed October 3, 2017. 2. Chen VW, Ruiz BA, Hsieh MC, Wu XC, Ries LA, Lewis DR. Analysis of stage and clinical/prognostic factors for lung cancer from SEER registries: AJCC staging and collaborative stage data collection system. Cancer. 2014;120(suppl 23):3781–3792. 3. National Cancer Institute. Drugs approved for lung cancer. https://www.cancer.gov/about-cancer/treatment/ drugs/lung. Accessed September 9, 2017. 4. Hanna N, Shepherd FA, Fossella FV, et al. Randomized phase III trial of pemetrexed versus docetaxel in patients with non-small-cell lung cancer previously treated with chemotherapy. J Clin Oncol. 2004;22:1589–1597. 5. Scagliotti G, Hanna N, Fossella F, et al. The differential efficacy of pemetrexed according to NSCLC histology: a review of two phase III studies. Oncologist. 2009;14: 253–263. 6. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med. 2004;350:2129–2139. 7. Sandler A, Gray R, Perry MC, et al. Paclitaxel-carboplatin alone or with bevacizumab for non-small-cell lung cancer. N Engl J Med. 2006;355:2542–2550. 8. Schiller JH, Harrington D, Belani CP, et al. Comparison of four chemotherapy regimens for advanced non-smallcell lung cancer. N Engl J Med. 2002;346:92–98. 9. Del Rivero J, Enewold L, Thomas A. Metastatic lung cancer in the age of targeted therapy: improving long-term survival. Transl Lung Cancer Res. 2016;5: 727–730. 10. Morgensztern D, Waqar S, Subramanian J, Gao F, Govindan R. Improving survival for stage IV non-small cell lung cancer: a surveillance, epidemiology, and end results survey from 1990 to 2005. J Thorac Oncol. 2009;4:1524–1529. 11. Xia W, Yu X, Mao Q, et al. Improvement of survival for non-small cell lung cancer over time. Onco Targets Ther. 2017;10:4295–4303. 12. Shepherd FA, Rodrigues Pereira J, Ciuleanu T, et al. Erlotinib in previously treated non-small-cell lung cancer. N Engl J Med. 2005;353:123–132. 13. Maemondo M, Inoue A, Kobayashi K, et al. Gefitinib or chemotherapy for non-small-cell lung cancer with mutated EGFR. N Engl J Med. 2010;362:2380–2388. 14. Koshy M, Malik R, Spiotto M, Mahmood U, Weichselbaum R, Sher D. Disparities in treatment of patients with inoperable stage I non-small cell lung cancer: a population-based analysis. J Thorac Oncol. 2015;10:264–271. 15. Patel AP, Crabtree TD, Bell JM, et al. National patterns of care and outcomes after combined modality therapy for stage IIIA non-small-cell lung cancer. J Thorac Oncol. 2014;9:612–621.
Stage IV Lung Cancer in ACs versus in CCs
1849
16. Wang EH, Rutter CE, Corso CD, et al. Patients Selected for definitive concurrent chemoradiation at high-volume facilities achieve improved survival in stage III nonsmall-cell lung cancer. J Thorac Oncol. 2015;10:937–943. 17. Rajaram R, Paruch JL, Mohanty S, et al. Patterns and predictors of chemotherapy use for resected non-small cell lung cancer. Ann Thorac Surg. 2016;101:533–540. 18. Samson P, Patel A, Crabtree TD, et al. Multidisciplinary treatment for stage IIIA non-small cell lung cancer: does institution type matter? Ann Thorac Surg. 2015;100: 1773–1779. 19. Salazar MC, Rosen JE, Wang Z, et al. Association of delayed adjuvant chemotherapy with survival after lung cancer surgery. JAMA Oncol. 2017;3:610–619. 20. Bilimoria KY, Stewart AK, Winchester DP, Ko CY. The National Cancer Data Base: a powerful initiative to improve cancer care in the United States. Ann Surg Oncol. 2008;15:683–690. 21. American College of Surgeons. About cancer program categories. https://www.facs.org/quality-programs/ cancer/coc /apply/categories. Accessed June 15, 2017. 22. Varlotto JM, Voland R, Mckie K, et al. Population-based differences in the outcome and presentation of lung cancer patients based upon racial, histologic, and economic factors in all lung patients and those with metastatic disease. Cancer Med. 2018;7:1211–1220. 23. Parikh-Patel A, Morris CR, Kizer KW. Disparities in quality of cancer care: the role of health insurance and population demographics. Medicine (Baltimore). 2017;96:e9125. 24. Ellis L, Canchola AJ, Spiegel D, Ladabaum U, Haile R, Gomez SL. Trends in cancer survival by health insurance status in California from 1997 to 2014. JAMA Oncol. 2018;4:317–323. 25. Mohanty S, Bilimoria KY. Comparing national cancer registries: the National Cancer Data Base (NCDB) and the Surveillance, Epidemiology, and End Results (SEER) program. J Surg Oncol. 2014;109:629–630. 26. Ramalingam S, Dinan M, Crawford J. Increasing survival in stage IV NSCLC in academic versus community based centers in the National Cancer Database. Abstract presented at: Multidisciplinary Symposium in Thoracic Oncology. September 15, 2017; Chicago, IL. 27. Wang S, Sun T, Sun H, et al. Survival improvement in patients with non-small cell lung cancer between 1983 and 2012: analysis of the Surveillance, Epidemiology, and End Results database. Tumour Biol. 2017;39, 1010428317691677. 28. Sher DJ, Liptay MJ, Fidler MJ. Prevalence and predictors of neoadjuvant therapy for stage IIIA non-small cell lung cancer in the National Cancer Database: importance of socioeconomic status and treating institution. Int J Radiat Oncol Biol Phys. 2014;89:303–312. 29. Sher DJ, Koshy M, Liptay MJ, Fidler MJ. Influence of conformal radiotherapy technique on survival after chemoradiotherapy for patients with stage III non-small cell lung cancer in the National Cancer Data Base. Cancer. 2014;120:2060–2068. 30. Morgensztern D, Ng SH, Gao F, Govindan R. Trends in stage distribution for patients with non-small cell lung cancer: a National Cancer Database survey. J Thorac Oncol. 2010;5:29–33.
1850 Ramalingam et al 31. Midha A, Dearden S, Mccormack R. EGFR mutation incidence in non-small-cell lung cancer of adenocarcinoma histology: a systematic review and global map by ethnicity (mutMapII). Am J Cancer Res. 2015;5: 2892–2911. 32. Tannenbaum SL, Koru-Sengul T, Zhao W, Miao F, Byrne MM. Survival disparities in non-small cell lung
Journal of Thoracic Oncology
Vol. 13 No. 12
cancer by race, ethnicity, and socioeconomic status. Cancer J. 2014;20:237–245. 33. U.S. Centers for Disease Control and Prevention. United States cancer statistics: data visualizations. Rates of new cancers: lung and bronchus, United States, 2015. https://www.cdc.gov/cancer/lung/statistics/race.htm. Accessed October 8, 2017.