J Clin Oncol by American Society of Clinical Oncology INTRODUCTION

Transcription

1 JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T Minimally Invasive Lung Cancer Surgery Performed by Thoracic Surgeons as Effective as Thoracotomy Daniel J. Boffa, Andrzej S. Kosinski, Anthony P. Furnary, Sunghee Kim, Mark W. Onaitis, Betty C. Tong, Patricia A. Cowper, Jessica R. Hoag, Jeffrey P. Jacobs, Cameron D. Wright, Joe B. Putnam Jr, and Felix G. Fernandez Author affiliations and support information (if applicable) appear at the end of this article. Published at jco.org on May 23, The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. Corresponding author: Daniel J. Boffa, MD, Yale School of Medicine, PO Box , New Haven, CT ; Daniel.boffa@ yale.edu by American Society of Clinical Oncology X/18/3699-1/$20.00 A B S T R A C T Purpose The prevalence of minimally invasive lung cancer surgery using video-assisted thoracic surgery (VATS) has increased dramatically over the past decade, yet recent studies have suggested that the lymph node evaluation during VATS lobectomy is inadequate. We hypothesized that the minimally invasive approach to lobectomy for stage I lung cancer resulted in a longitudinal outcome that was not inferior to thoracotomy. Patients and Methods Patients. 65 years of age who had undergone lobectomy for stage I lung cancer between 2002 and 2013 were analyzed within the Society of Thoracic Surgeons General Thoracic Surgery Database, which had been linked to Medicare data, as part of a retrospective-cohort, noninferiority study. Results A total of 10,597 patients with clinical stage I lung cancer who underwent lobectomy were evaluated (4,448 patients underwent thoracotomy, and 6,149 underwent VATS). VATS patients had a more favorable distribution of all health-related variables, including pulmonary function (59% of VATS patients had intact spirometry v 51% of thoracotomy patients; P ). Cox proportional hazards models were performed over two eras to account for an evolving practice standard. The mortality risk associated with the VATS approach was not greater than thoracotomy in either the earlier era (2002 to 2008; hazard ratio, 0.97; 95% CI, 0.87 to 1.09; P =.62) or the more recent era (2009 to 2013; hazard ratio, 0.84; 95% CI, 0.75 to 0.93; P ). Kaplan-Meier survival estimates of 2,901 propensity-matched VATS-thoracotomy pairs demonstrated that the 4-year survival associated with VATS (68.6%) was modestly superior to thoracotomy (64.8%; P =.003). The analyses detailed above were replicated in a separate cohort of pathologic stage I patients with similar findings. Conclusion The long-term efficacy of lobectomy for stage I lung cancer performed using the VATS approach by board-certified thoracic surgeons does not seem to be inferior to that of thoracotomy. J Clin Oncol by American Society of Clinical Oncology ASSOCIATED CONTENT Appendix DOI: DOI: INTRODUCTION Over the past two decades, the surgical management of solid organ malignancies has been transformed by the widespread adoption of minimally invasive surgical (MIS) techniques. MIS approaches generally allow patients to recover with less pain, 1 fewer complications, 2,3 and expedited return to their baseline functionality. 4 However, for lung cancer, the nation s leading cancer killer, the MIS approach has recently been challenged as being oncologically inferior to the traditional surgical approach (thoracotomy). 5,6 This controversy has been fueled by several large observational reports of radiographically occult lymph node metastases being less likely to be found in the surgical specimens (nodal upstaging) of patients with lung cancer treated using MIS (videoassisted thoracic surgery [VATS]) compared with those treated using thoracotomy The obvious concern is that radiographically occult nodal metastases are being left behind in VATS patients, presumably because the VATS approach does not allow surgeons to perform as complete a surgical lymph node evaluation. Not only would failure to remove lymph node metastases leave patients with residual cancer (itself a poor prognosticator), 11 but potentially more importantly, patients with 2018 by American Society of Clinical Oncology 1

2 Boffa et al unrecognized nodal metastasis could also miss the opportunity to benefit from adjuvant chemotherapy. 12 Although proponents and critics of VATS agree that a randomized controlled trial would best clarify the oncologic efficacy of the VATS approach, it is highly unlikely that an appropriately powered trial will take place. 13 Putting aside the challenges of surgeon equipoise 14 (VATS has become a standard of care for many thoracic surgical practices),. 5,000 patients would need to be randomly assigned to demonstrate survival differences resulting from differences in nodal upstaging, 8 which many consider to be prohibitive. As an alternative, observational studies of large data registries have the potential power to depict approach-specific survival differences with degree of confidence that approaches clinical trials Unfortunately, currently available databases (eg, National Cancer Database, SEER Database) lack critical data required to properly adjust for important differences in the VATS and thoracotomy cohorts (eg, pulmonary function, performance status, body mass index). 13 In 2002, the Society of Thoracic Surgeons (STS) created a General Thoracic Surgery Database (GTSD) with the sole purpose of improving the quality of patient care. In an effort to enhance observational study, the STS-GTSD evolved to capture an unprecedented level of detail concerning patient health (eg, detailed comorbidities, performance status, pulmonary function), tumor attributes, and surgical procedures. In the current study, the STS-GTSD was linked to longitudinal follow-up and health care utilization data in Medicare to compare VATS and thoracotomy approaches. We hypothesized that the survival of patients with stage I lung cancer managed with the VATS approach by boardcertified thoracic surgeons is not inferior to that of patients who underwent thoracotomy. PATIENTS AND METHODS Databases The STS-GTSD is a prospectively maintained, comprehensive data resource that captures the care of patients who are operated on by boardcertified thoracic surgeons. With. 71,829 records, it is currently among the largest and the most comprehensive thoracic surgical databases in the world, but it does not include longitudinal follow-up. 18 The Centers of Medicare and Medicaid Services (CMS) database captures elements of care for Medicare beneficiaries in the United States $ 65 years of age. The CMS database includes sociodemographic data as well as claims to Medicare and longitudinal follow-up. A subset of patients in the STS-GTSD $ 65 years old were linked to CMS records using a deterministic matching algorithm based on patient age and sex, the hospital in which surgery took place, admission date, and discharge date, as previously described. 19,20 Overall, 69.5% of patients were able to be matched (linked cohort). The most common reason patients were not able to be matched was that patients did not use Medicare as the primary payer (eg, used a health maintenance organization or commercial payer). Comparisons of matched and unmatched patients were performed previously and did not seem to identify significant differences in sociodemographics or tumor attributes that would obviously influence mortality risk. 19 Patients The study included treatment-naïve patients $ 65 years old who underwent a lobectomy for lung cancer in the linked cohort (STS-GTSD linked to CMS). Patients were excluded if they had undergone prior thoracic surgery on the same side, if their surgery was emergent or urgent, or if they had undergone induction chemotherapy or radiation therapy. Patients were also excluded if they were missing data for the covariates of sex, race, forced expiratory volume in 1 second (FEV1), Zubrod score, American Society of Anesthesiologists risk class, or T1 or T2 staging status (defined under Variables). The prevalence of missing data for each variable is listed in Appendix Tables A1 and A2 (online only). Conversions (from VATS to thoracotomy) were coded in the STS- GTSD as having both a VATS procedure and a thoracotomy on the same day. Converted patients were analyzed with the VATS cohort for all of the analyses in the main body of the article, based on their intended approach. Of note, sensitivity analyses were performed analyzing converted patients with thoracotomy patients, and the study conclusions were unchanged (Appendix Table A3, online only). Because neither analysis is purely VATS or thoracotomy, we are comparing regimens (approach v approach plus converted patients). Study Cohorts Clinical stage I cohort. The primary study analyses were performed using a clinical stage I cohort. Pathologic stage I cohort. A set of secondary analyses was performed in a cohort of pathologic stage I tumors to address bias in the clinical stage I cohort by tumor attributes not captured by the STS-GTSD. For example, central tumors are more likely than peripheral tumors to have radiographically occult lymph node metastases detected in surgical specimens (ie, clinical N0 but pathologic N1 or N2) 21,22 and are potentially more likely to be approached by thoracotomy. Examination of the pathologic stage I cohort (all node negative) mitigates this potential bias. Variables The following independent variables were taken from the STS-GTSD and included in multivariable analyses: age; sex; race; year of surgery; body mass index; American Society of Anesthesiologist risk class; Zubrod score (performance status, 0 to 5) 23 ; presence of coronary artery disease, cerebrovascular disease, congestive heart failure, hypertension, diabetes mellitus, corticosteroid use, peripheral vascular disease, or renal insufficiency (creatinine. 2 mg/dl or hemodialysis); FEV 1 (percent predicted); cigarette use; laterality; and clinical and pathologic stage. 24 Annual lobectomy volume was determined as follows: 4 3 (total number of lobectomies for lung cancer for each quarter submitted from a hospital 4 the number of quarters the hospital submitted patients). The study period included a transition between editions of the American Joint Committee on Cancer (AJCC) stage classification for lung cancer (from the sixth to seventh edition). Because of insufficient staging data to forward-stage earlier patients, the majority of the study was performed by back-staging later patients. As a result, for analyses that include all study years, all patients were staged according to the AJCC sixth edition. However, for analyses that are restricted to later years (ie, the propensity-matched analyses), the AJCC seventh edition was used (as indicated in tables). Several variables were abstracted from Medicare outpatient and carrier files (only available for current study between 2008 and 2013; specific codes listed in Appendix Table A4, online only). The clinical staging evaluation in the 4 months before surgery date was characterized by claims for imaging (positron emission tomography [PET] scanning, computed tomography [CT] scanning, brain magnetic resonance imaging, or head CT) and invasive mediastinal staging (mediastinoscopy, endobronchial ultrasound, endoscopic ultrasound, or thoracoscopy with mediastinal or pleural biopsy). The use of adjuvant chemotherapy was determined by a query for chemotherapy claims for agents with activity in lung cancer that first occurred within the 6 months 25 after date of surgery. Operative mortality was determined at 30 and 90 days from date of surgery using both STS-GTSD operative mortality fields and vital status obtained from the Medicare data by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

3 VATS Versus Thoracotomy for Lung Cancer Table 1. Overview of Clinical Stage I Populations Variable Thoracotomy (n = 4,448) VATS or Conversion (n = 6,149) P Median age, years (interquartile range) 73 (69-78) 73 (69-78).3421 Female 2,247 (50.5) 3,455 (56.2) White 4,170 (93.8) 5,798 (94.3).3771 Surgery year ,525 (34.3) 1,049 (17.1) ,923 (65.7) 5,100 (82.9) BMI group, kg/m , (2.3) 131 (2.1) ,312 (29.5) 1,993 (32.4) ,851 (41.6) 2,516 (40.9) (18.1) 1,025 (16.7) (8.5) 484 (7.9) American Society of Anesthesiologist risk class I 23 (0.5) 43 (0.7) II 561 (12.6) 927 (15.1) III 3,180 (71.5) 4,699 (76.4) IV 682 (15.3) 479 (7.8) V 2 (0.0) 1 (0.0) Zubrod score 0 1,918 (43.1) 3,099 (50.4) 1 2,322 (52.2) 2,826 (46.0) (4.7) 224 (3.6) Coronary artery disease (yes) 1,247 (28.0) 1,513 (24.6) Cerebrovascular disease (yes) 457 (10.3) 579 (9.4).1422 Congestive heart failure (yes) 194 (4.4) 179 (2.9) Hypertension (yes) 3,001 (67.5) 4,120 (67.0).6143 Diabetes mellitus (yes) 895 (20.1) 1,130 (18.4).0242 Corticosteroid use (yes) 153 (3.4) 187 (3.0).2505 Peripheral vascular disease (yes) 574 (12.9) 690 (11.2).0083 Chronic kidney disease (Cr. 2 mg/dl or dialysis; yes; set 119 (2.7) 123 (2.0).0217 missing to no) FEV 1, % predicted, (1.7) 106 (1.7) (13.2) 623 (10.1) ,498 (33.7) 1,798 (29.2) $ 80 2,285 (51.4) 3,622 (58.9) Smoking status.0051 Never smoked 558 (12.5) 895 (14.6) Past smoker (stopped. 1 month prior to operation) 3,026 (68.0) 4,147 (67.4) Current smoker 864 (19.4) 1,107 (18.0) Clinical tumor status T1 2,891 (65.0) 4,681 (76.1) T2 1,557 (35.0) 1,468 (23.9) Nodal upstaging (cn0 pn1-3) 650 (14.6) 675 (11.0) Adjuvant chemotherapy (CMS) for surgeries* 530 (16.3) 766 (13.8).0016 Median days from surgery date to drug date (CMS) for 54.5 ( ) 49.0 ( ) surgeries (interquartile range) CT scan of chest (CMS) for surgeries 2,932 (90.1) 5,007 (90.3).74 Brain imaging (CMS) for surgeries 1,136 (34.9) 1,582 (28.5) PET scan (CMS) for surgeries 2,781 (85.5) 4,739 (85.5).98 Invasive mediastinal staging (EBUS, mediastinoscopy, EUS, or 1,087 (33.4) 1,621 (29.2) multiple; CMS) for surgeries Positive surgical margins 36 (3.5) 74 (3).40 NOTE. Data presented as No. (%) unless otherwise indicated. Abbreviations: BMI, body mass index; CMS, Centers for Medicare and Medicaid Services; Cr, creatinine; CT, computed tomography; EBUS, endobronchial ultrasound; EUS, endoscopic ultrasound FEV 1, forced expiratory volume in 1 second; PET, positron emission tomography; VATS, video-assisted thoracic surgery. *Includes chemotherapy given within 180 days of surgery date. These data were only available for patients who underwent surgery between 2008 and Includes staging evaluations performed within the 120 days before surgery date. These data were only available for patients who underwent surgery between 2008 and Only available from Society of Thoracic Surgeons General Thoracic Surgery Database for 2012 and Not included in any adjusted analyses. Statistical Considerations The amount of missing data within the linked data set was low. For body mass index, missing values were imputed using sex-specific median values. For binary risk factors, missing values were considered as indicating absence of the risk factor. Cox proportional hazards models with robust sandwich variance estimates to adjust for clustering within hospitals were built including the variables listed earlier. Violations of proportional hazards assumption were assessed graphically and by testing significance of the correlation between survival time and the scaled Schoenfeld residuals. No statistically jco.org 2018 by American Society of Clinical Oncology 3

4 Boffa et al significant departure from the proportional hazards assumption was noted for the surgical approach (P..25), although the benefit of VATS seemed to be strongest early on. Minor violations for other covariates were noted but not felt to affect adjusted evaluation of approach. For the clinical stage I patient cohort, era (2002 to 2008 v 2009 to 2013) was found to interact with surgical approach (P =.03). An interaction term was included in the Cox proportional hazards model for clinical stage I patients. No such interaction was noted in the pathologic stage I patients. The noninferiority margin was set at 1.10 (if upper limit of 95% CI of hazard ratio [HR] of VATS to thoracotomy was, 1.10, then VATS was not inferior). All Cox models met noninferiority criteria. Separate propensity matches were performed for clinical and pathologic stage I tumors, as described previously. 7 The propensity models were restricted to patients staged with the most recent version of the staging classification (AJCC seventh edition, 2009 to 2013) to achieve greater balance of tumor attributes. More specifically, initial analyses identified important differences in the distribution of Tsubgroups favoring VATS (Table 1). Because the seventh edition expanded the stratification of the T variable (T1a, T1b, T2a, and T2b) compared with the sixth edition (T1 and T2), we elected to focus on the group with the most precise staging data. The propensity matching also included the extent of clinical staging evaluation (eg, use of PET scan, CT scan, mediastinoscopy). Matching variables and standardized differences (all,.1) 26 are shown in Appendix Figure A1 (online only). Kaplan-Meier estimates of survival were created. In addition, univariable Cox model analyses with robust sandwich variance estimates of propensity-matched patients were performed to adjust for clustering within hospitals. Differences in the distribution of variables across approaches were evaluated using Pearson s x 2 test for categorical variables and the Wilcoxon rank sum test for continuous variables. Two-sided P,.05 was considered to be significant. Analyses were performed using the SAS Version 9.4 (SAS Institute, Cary, NC) and R version ( Sensitivity Analyses The study design was intended to offer the least biased view of the relationship between surgical approach and survival. A number of different permutations of the described study design were performed (ie, realignment of conversions, stratification by clinical T stage, stratification by era, landmarking at 90 days to eliminate surgical mortality) to assess their potential to affect the study findings (Appendix Table A3). RESULTS Patients A total of 10,597 patients with clinical stage I non small-cell lung cancer in the STS-GTSD were successfully linked to Medicare claims files, including 4,448 patients who underwent lobectomy using a thoracotomy approach and 6,149 who underwent VATS. Patient sociodemographic and tumor characteristics for the thoracotomy and VATS patients are listed in Table 1. The vast majority of imbalanced variables would favorably affect the long-term survival of the VATS cohort, including lower prevalence of comorbidities, better pulmonary function, better performance status, greater proportion of female patients, greater percentage of earliest stage tumors, and a lower prevalence of nodal upstaging (ie, clinical node-negative patient who had lymph node metastases discovered in surgical specimen). Adjusted Mortality Risk Among Clinical Stage I Patients An adjusted Cox proportional hazards model was created of clinical stage I patients (Table 2). Because of an interaction between Table 2. Cox Robust Sandwich Estimates of Overall Survival in Clinical Stage I Patients Adjusted HR (95% CI) Adjusted P Risk Factor VATS (v thoracotomy) for (0.87 to 1.09).6237 surgeries* VATS (v thoracotomy) for (0.75 to 0.93) surgeries Surgery year (v ) 1.04 (0.93 to 1.17).5126 for thoracotomy Surgery year (v ) 0.89 (0.80 to 0.99).0368 for VATS Volume per year for 10-volume increase 0.99 (0.98 to 1.00).1397 Clinical stage T2 (v T1) 1.32 (1.22 to 1.44) Female sex (v male) 0.71 (0.67 to 0.77) Age, years Reference (1.06 to 1.29) (1.35 to 1.65),. 001 $ (1.92 to 2.42),. 001 ASA risk class I or II Reference III 1.30 (1.16 to 1.47),. 001 IV or V 1.50 (1.28 to 1.75),. 001 FEV 1, % predicted. 80 Reference, (1.15 to 1.84) (1.25 to 1.57) (1.11 to 1.32) BMI, kg/m Reference, (1.42 to 2.12) (0.74 to 0.91) (0.74 to 0.97) (0.80 to 1.13).5872 Coronary artery disease 1.14 (1.05 to 1.24).0022 Congestive heart failure 1.34 (1.13 to 1.59) Smoking status Never smoked cigarettes Reference Former smoker (stopped. 1 month 1.47 (1.27 to 1.69) before operation) Current smoker 1.72 (1.48 to 1.98) Cerebrovascular disease 1.14 (1.01 to 1.28).0280 Diabetes mellitus 1.13 (1.03 to 1.24).0089 Hypertension 1.02 (0.93 to 1.12).6424 Peripheral vascular disease 1.25 (1.12 to 1.40) Chronic kidney disease (creatinine 1.59 (1.32 to 1.91). 2 mg/dl or dialysis) Corticosteroid use 1.44 (1.21 to 1.72) Race White Reference Black/African American 1.00 (0.88 to 1.14).9773 Other 1.23 (0.76 to 2.00).3992 Nodal upstaging 2.13 (1.92 to 2.36) Zubrod score 0 Reference (1.12 to 1.31) (1.43 to 1.94) Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; FEV 1, forced expiratory volume in 1 second; HR, hazard ratio; VATS, videoassisted thoracic surgery. *An interaction was noted between surgical approach and era (2002 to 2008 v 2009 to 2013). An interaction term was inserted for this purpose. Nodal upstaging refers to cn0 pn1, pn2, or pn3. surgical approach and era, the relationship between surgical approach and mortality risk was examined in two eras (2002 to 2008 and 2009 to 2013). Compared with thoracotomy, VATS lobectomy was not associated with an increased risk of mortality in the earlier by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

5 VATS Versus Thoracotomy for Lung Cancer A Survival Rate (%) VATS Thoracotomy HR, 0.84; 95% CI, 0.74 to 0.94; P =.003 B Survival Rate (%) VATS Thoracotomy HR, 0.91; 95% CI, 0.78 to 1.05; P = Time After Surgery (years) No. of patients/survival rate (%): VATS 2,901 2,143/91.3 1,476/ / / /65.5 Thoracotomy 2,901 2,119/89.3 1,476/ / / / Time After Surgery (years) No. of patients/survival rate (%): VATS 2,414 1,810/91.9 1,264/ / / /67.8 Thoracotomy 2,414 1,814/91.0 1,300/ / / /63.8 Fig 1. Kaplan-Meier curves are shown for propensity-matched cohorts of (A) clinical stage I patients and (B) pathologic stage I patients. Univariable Cox models were also performed, and the hazard ratio (HR) of video-assisted thoracic surgery (VATS) to thoracotomy is shown within the figure with 95% CI and P value. era (HR, 0.97; 95% CI, 0.87 to 1.09; P =.62). In the more recent era, VATS lobectomy seemed to convey better survival compared with thoracotomy lobectomy (HR, 0.84; 95% CI, 0.75 to 0.93; P ). Propensity-Matched Clinical Stage I Patients A propensity model was built using patients from the most recent era (2009 to 2013) that included the extent of clinical staging evaluation (ie, imaging and invasive mediastinal staging) as well as a more precise version of the stage classification system (AJCC seventh edition). Propensity matching was performed, resulting in 2,901 extremely well-balanced pairs of VATS and thoracotomy patients (Appendix Fig A1A). A Kaplan-Meier survival analysis was performed with a median follow-up of 2.31 years for surviving patients. The 4-year survival after VATS lobectomy (68.6%) was modestly superior to that after thoracotomy lobectomy (64.8%; P =.003; Fig 1A). Pathologic Stage I Patients In an effort to examine the efficacy of surgical approach from a different perspective, the comparative analyses were repeated in a cohort of pathologic stage I patients (thoracotomy, n = 4,136; VATS, n = 5,651). The sociodemographic and tumor characteristics were similarly skewed in favor of the VATS cohort in the pathologic stage I cohort, as was previously noted for the clinical stage I patients (Appendix Table A5, online only). Cox proportional hazards analysis failed to identify an increased mortality risk associated with the VATS approach compared with thoracotomy (HR, 0.89; 95% CI, 0.80 to 0.98; P =.025;Table 3). Propensity matching was performed in the most recent era (2009 to 2013), resulting in 2,414 extremely wellbalanced pairs (Appendix Fig A1B). Kaplan-Meier survival estimates (median follow-up of 2.38 years for surviving patients) indicated that the 4-year survival of VATS patients (71.3%) was not inferior to that of thoracotomy patients (68.8%; P =.18; Fig 1B). Administration of Adjuvant Chemotherapy Adjuvant chemotherapy use was evaluated in the most recent era (2008 to 2013). Within the clinical stage I subset, the use of adjuvant chemotherapy was more prevalent among patients who had undergone lobectomy via thoracotomy (16.3%) compared with patients who had undergone VATS (13.8%; P ; Table 1). The interval between surgery and the start of adjuvant chemotherapy was shorter for VATS patients (median, 50 days) compared with thoracotomy patients (median, 55 days; P ). DISCUSSION The current study indicates that the minimally invasive approach to lobectomy for non small-cell lung cancer results in a long-term survival that is not inferior to lobectomy performed by thoracotomy. These findings contrast a wave of studies (including a recent study using the same database) 7 suggesting that surgeons using the VATS approach failed to achieve as complete a surgical lymph node evaluation as surgeons using thoracotomy. 8 There are several potential explanations for the discordance between the current study and the multiple studies suggesting VATS results in an inadequate lymph node evaluation. First and foremost is the possibility that unmeasured bias confounds the survival results and obscures an advantage of thoracotomy. A number of observational studies have suggested that survival after VATS is not inferior to thoracotomy 8,27,28 ; however, these studies lacked sufficient data to appropriately balance approach-specific cohorts. 13 The current study has greatly expanded the effort to balance the approach-specific cohorts by including far greater detail on patient health (eg, pulmonary function, performance status), the extent of staging evaluation performed (eg, PET scan, mediastinoscopy), and surgical volume, as well as exploring a number of alternate study designs and variable alignments (corrective measures summarized in Table 4). Finally, it is well established that general surgeons are less likely to comply with lung cancer guidelines and achieve inferior short- and long-term outcomes compared with thoracic surgeons. 29,31-33 By analyzing a cohort managed by board-certified thoracic surgeons, the effect of surgical approach could be disentangled from the confounding effect of surgical training. Although some degree of unmeasured bias likely persists, we feel it is likely to be of minor impact. jco.org 2018 by American Society of Clinical Oncology 5

6 Boffa et al Table 3. Cox Robust Sandwich Estimates of Overall Survival in Pathologic Stage I Patients Risk Factor Adjusted HR (95% CI; n = 9,787) Adjusted P Surgery year (v ) 0.95 (0.86 to 1.06).3405 VATS approach (v thoracotomy) 0.89 (0.80 to 0.98).0246 Volume per year for 10-volume increase 0.99 (0.98 to 1.01).4014 Pathologic stage T2 (v T1) 1.37 (1.26 to 1.48) Female sex (v male) 0.73 (0.68 to 0.79) Age, years Reference (1.01 to 1.23) (1.34 to 1.66) $ (1.83 to 2.31) ASA risk class IorII Reference III 1.36 (1.19 to 1.56) IV or V 1.54 (1.31 to 1.82) FEV 1, % predicted. 80 Reference, (1.19 to 1.78) (1.26 to 1.62) (1.13 to 1.38) BMI, kg/m Reference, (1.47 to 2.35) (0.75 to 0.92) (0.73 to 0.95) (0.79 to 1.12).5195 Coronary artery disease 1.12 (1.02 to 1.23).0165 Congestive heart failure 1.41 (1.21 to 1.63) Smoking status Never smoked cigarettes Reference Former smoker 1.47 (1.26 to 1.72) Current smoker 1.72 (1.46 to 2.03) Cerebrovascular disease 1.17 (1.04 to 1.31).0066 Diabetes mellitus 1.10 (0.99 to 1.23).0821 Hypertension 1.08 (0.97 to 1.19).1597 Peripheral vascular disease 1.34 (1.19 to 1.50) Chronic kidney disease (creatinine. 2 mg/dl or dialysis) 1.66 (1.32 to 2.09) Corticosteroid use 1.42 (1.19 to 1.69) Race White Reference Black/African American 0.97 (0.83 to 1.14).7259 Other 1.12 (0.67 to 1.89).6585 Zubrod score 0 Reference (1.16 to 1.35) (1.45 to 2.00) Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; FEV 1, forced expiratory volume in 1 second; HR, hazard ratio; VATS, video-assisted thoracic surgery. There is also a possibility that VATS provides an inferior oncologic procedure (eg, lymph node assessment), yet the survival difference attributable to missed nodal metastases is too subtle to be apparent in the current study. For stage II patients to actually benefit from the knowledge of accurate staging data, they would need to be treated with adjuvant chemotherapy. The VATS approach was associated with a 2.2% decrease in the prevalence of upstaged patients, and the use of adjuvant chemotherapy affords a 5% increase in 5-year survival. 12 Therefore, the effect of missed opportunities to give chemotherapy on 5-year survival of the overall cohort would be small. The novel data resource used in this study (STS-GTSD linked to claims data) represents a major advance in the observational study of surgically managed lung cancer. The sole purpose of this database was to enhance the care provided by thoracic surgeons. This study serves as a clear example of the potential contributions of a society-wide quality improvement effort, as the linked database greatly informed a timely debate in the field (VATS v thoracotomy for early-stage lung cancer). This study has a number of limitations in addition to those intrinsic to observational study. The current study reports all-cause mortality and not cancer-specific mortality. Although the current study represents the most exhaustive attempt to balance approach-specific cohorts to date, it is possible that important differences in oncologic efficacy were obscured by imbalances in competing mortality risk (ie, health). The population was restricted to patients. 65 years old. More than 60% of patients undergoing lung cancer surgery in the United States are. 65 years old 34 ; however, the findings may not be applicable to younger patients. Histology was not captured by the STS-GTSD for the years of this study. A by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

7 VATS Versus Thoracotomy for Lung Cancer Table 4. Summary of Measures Taken to Reduce Study Bias Potential Bias Direction of Bias (before adjustment) Prior Studies 8,27,28 Adjustment in Current Study Patient performance status Favors VATS (Table 1) Not accounted for Performance status included in adjusted models Pulmonary function Favors VATS (Table 1) Not accounted for FEV 1 included in adjusted models Malnutrition Favors VATS (Table 1) Not accounted for Body mass index included in adjusted models Peripheral location of tumor (better prognosis) 22 Potentially favors VATS Not accounted for Pathologic stage I subset* Annual lobectomy volume Favors VATS 3 Not accounted for Volume included in Cox model Extent of clinical staging evaluation (greater clinical evaluation better outcomes) 30 Unknown Not accounted for 27,28 Invasive mediastinal staging only 8 Included in propensity matching Converted patients (patients who had VATS and thoracotomy on the same day) Evolving standard of care Favors VATS Conversions as VATS 27 Conversions as thoracotomy 8 Not stated 28 Earlier: potentially favor thoracotomy Later: potentially favor VATS Not accounted for 27,28 Adjusted for year 8 Conversions were analyzed with both VATS and thoracotomy in separate analyses Examined relative efficacy in two eras (sensitivity analyses) Adjusted for year Surgeon training (specialty training better Unknown Not accounted for All board-certified thoracic surgeons outcomes) 29 Abbreviations: FEV 1, forced expiratory volume in 1 second; VATS, video-assisted thoracic surgery. *Central tumors were more likely to have radiographically occult lymph node metastases (ie, be upstaged from clinical N0 to pathologic N1 or N2). The pathologic stage I subset minimizes the effect of this bias by eliminating nodal metastases from both cohorts. Conversions can bias irrespective of how they are considered. Conversions are a combination of the following: reactions to challenging intraoperative findings or events (complications); and a planned diagnostic VATS limited resection to confirm cancer, followed by planned thoracotomy for anatomic resection. In the former scenario, adding conversions to the thoracotomy cohort would bias against thoracotomy, because the cancer-specific reasons for conversion (typically larger or more invasive tumors) or conversion for major surgical complications both have the potential to detract from long-term prognosis. In the latter scenario, analyzing converted patients with the VATS cohort would bias against thoracotomy, because only peripheral tumors could be wedged by VATS and peripheral tumors have a better prognosis. Therefore, removing the patients who had undergone a diagnostic VATS wedge from the thoracotomy cohort would reduce the proportion of patients with peripheral tumors (better prognosis) in this cohort. Earlier, thoracotomy was standard of care, and VATS was performed as the exception and results were vulnerable to the learning curve phase of adoption. Currently, themajorityofearly-stage lungcancerismanagedbyvats, andthoracotomyhasbecomethe exception, whichmayalsohaveprognosticimplications (more challenging, central tumors). more recent cohort of pathologic stage I patients in the STS-GTSD (2015 to 2017, not included in current study) found the prevalence of adenocarcinoma to be 64.9% in the thoracotomy subset and 68.2% in the VATS subset. Histology was not a significant prognosticator in previous observational comparisons of surgical approach. 8,27 Furthermore, histology was not a significant prognosticator in the International Association for the Study of Lung Cancer s lung cancer stage classification project 35 ; however, this is still a potential limitation. Although not a randomized trial, we feel the current analytic approach has adequately excluded the possibility that VATS is associated with inferior survival to thoracotomy in early-stage lung cancer, because every Cox model complied with our preset noninferiority margin In conclusion, the survival of patients with stage I lung cancer managed using the VATS approach performed by board-certified thoracic surgeons does not seem to be inferior to that of patients undergoing thoracotomy. AUTHORS DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST Disclosures provided by the authors are available with this article at jco.org. AUTHOR CONTRIBUTIONS Conception and design: Daniel J. Boffa, Anthony P. Furnary, Sunghee Kim, Mark W. Onaitis, Betty C. Tong, Patricia A. Cowper, Jeffrey P. Jacobs, Cameron D. Wright, Joe B. Putnam Jr, Felix G. Fernandez Collection and assembly of data: Daniel J. Boffa Data analysis and interpretation: All authors Manuscript writing: All authors Final approval of manuscript: All authors Accountable for all aspects of the work: All authors REFERENCES 1. Bendixen M, Jørgensen OD, Kronborg C, et al: Postoperative pain and quality of life after lobectomy via video-assisted thoracoscopic surgery or anterolateral thoracotomy for early stage lung cancer: A randomised controlled trial. Lancet Oncol 17: , Delaney CP, Chang E, Senagore AJ, et al: Clinical outcomes and resource utilization associated with laparoscopic and open colectomy using a large national database. Ann Surg 247: , Boffa DJ, Dhamija A, Kosinski AS, et al: Fewer complications result from a video-assisted approach to anatomic resection of clinical stage I lung cancer. J Thorac Cardiovasc Surg 148: , Wang H, Feng M, Tan L, et al: Comparison of the short-term quality of life in patients with esophageal cancer after subtotal esophagectomy via video-assisted thoracoscopic or open surgery. Dis Esophagus 23: , Mathisen DJ: Is video-assisted thoracoscopic lobectomy inferior to open lobectomy oncologically? Ann Thorac Surg 96: , Siegel RL, Miller KD, Jemal A: Cancer statistics, CA Cancer J Clin 67:7-30, Boffa DJ, Kosinski AS, Paul S, et al: Lymph node evaluation by open or video-assisted approaches in 11,500 anatomic lung cancer resections. Ann Thorac Surg 94: , Licht PB, Jørgensen OD, Ladegaard L, et al: A national study of nodal upstaging after thoracoscopic jco.org 2018 by American Society of Clinical Oncology 7

8 Boffa et al versus open lobectomy for clinical stage I lung cancer. Ann Thorac Surg 96: , Medbery RL, Gillespie TW, Liu Y, et al: Nodal upstaging is more common with thoracotomy than with VATS during lobectomy for early-stage lung cancer: An analysis from the National Cancer Data Base. J Thorac Oncol 11: , Martin JT, Durbin EB, Chen L, et al: Nodal upstaging during lung cancer resection is associated with surgical approach. Ann Thorac Surg 101: , Hancock JG, Rosen JE, Antonicelli A, et al: Impact of adjuvant treatment for microscopic residual disease after non-small cell lung cancer surgery. Ann Thorac Surg 99: , Pignon JP, Tribodet H, Scagliotti GV, et al: Lung adjuvant cisplatin evaluation: A pooled analysis by the LACE Collaborative Group. J Clin Oncol 26: , Allen MS: Video-assisted thoracoscopic surgery versus open lobectomy for lung cancer: Time for a randomized trial. Eur J Cardiothorac Surg 51:175, McCulloch P, Taylor I, Sasako M, et al: Randomised trials in surgery: Problems and possible solutions. BMJ 324: , Frakt AB: An observational study goes where randomized clinical trials have not. JAMA 313: , Anglemyer A, Horvath HT, Bero L: Healthcare outcomes assessed with observational study designs compared with those assessed in randomized trials. Cochrane Database Syst Rev 4:MR000034, Ross JS: Randomized clinical trials and observational studies are more often alike than unlike. JAMA Intern Med 174:1557, Boffa DJ, Allen MS, Grab JD, et al: Data from the Society of Thoracic Surgeons General Thoracic Surgery Database: The surgical management of primary lung tumors. J Thorac Cardiovasc Surg 135: , Fernandez FG, Furnary AP, Kosinski AS, et al: Longitudinal follow-up of lung cancer resection from the Society of Thoracic Surgeons General Thoracic Surgery database in patients 65 years and older. Ann Thorac Surg 101: , Jacobs JP, Edwards FH, Shahian DM, et al: Successful linking of the Society of Thoracic Surgeons database to social security data to examine survival after cardiac operations. Ann Thorac Surg 92: 32-39, Lee PC, Port JL, Korst RJ, et al: Risk factors for occult mediastinal metastases in clinical stage I nonsmall cell lung cancer. Ann Thorac Surg 84: , Gao S, Kim AW, Puchalski JT, et al: Indications for invasive mediastinal staging in patients with early non-small cell lung cancer staged with positron emission tomography-computed tomography. Int J Radiat Oncol Biol Phys 98:223, Oken MM, Creech RH, Tormey DC, et al: Toxicity and response criteria of the Eastern Cooperative Oncology Group. Am J Clin Oncol 5: , The Society of Thoracic Surgeons: The Society of Thoracic Surgeons Training Manual for the General Thoracic Surgery Database. sites/default/files/documents/gtsd%20training% 20manual%20_Oct.pdf 25. Salazar MC, Rosen JE, Wang Z, et al: Association of delayed adjuvant chemotherapy with survival after lung cancer surgery. JAMA Oncol 3: , Haukoos JS, Lewis RJ. The propensity score. JAMA 314: , Yang CJ, Kumar A, Klapper JA, et al: A national analysis of long-term survival following thoracoscopic versus open lobectomy for stage I non-small-cell lung cancer. Ann Surg [epub ahead of print on August 9, 2017] 28. Paul S, Isaacs AJ, Treasure T, et al: Long term survival with thoracoscopic versus open lobectomy: Propensity matched comparative analysis using SEER-Medicare database. BMJ 349:g5575, Farjah F, Flum DR, Varghese TK Jr, et al: Surgeon specialty and long-term survival after pulmonary resection for lung cancer. Ann Thorac Surg 87: , Farjah F, Flum DR, Ramsey SD, et al: Multimodality mediastinal staging for lung cancer among Medicare beneficiaries. J Thorac Oncol 4: , Wood DE, Farjah F: Surgeon specialty is associated with better outcomes: The facts speak for themselves. Ann Thorac Surg 88: , Schipper PH, Diggs BS, Ungerleider RM, et al: The influence of surgeon specialty on outcomes in general thoracic surgery: A national sample 1996 to Ann Thorac Surg 88: , Ellis MC, Diggs BS, Vetto JT, et al.: Intraoperative oncologic staging and outcomes for lung cancer resection vary by surgeon specialty. Ann Thorac Surg 92: , Rosen JE, Hancock JG, Kim AW, et al: Predictors of mortality after surgical management of lung cancer in the National Cancer Database. Ann Thorac Surg 98: , Chansky K, Sculier JP, Crowley JJ, et al: The International Association for the Study of Lung Cancer Staging Project: Prognostic factors and pathologic TNM stage in surgically managed nonsmall cell lung cancer. J Thorac Oncol 4: , 2009 Affiliations Daniel J. Boffa and Jessica R. Hoag, Yale School of Medicine, New Haven, CT; Andrzej S. Kosinski, Sunghee Kim, and Patricia A. Cowper, Duke Clinical Research Institute; Betty C. Tong, Duke University School of Medicine, Durham, NC; Anthony P. Furnary, Starr- Wood Cardiac Group, Portland, OR; Mark W. Onaitis, University of California, San Diego School of Medicine, San Diego, CA; Jeffrey P. Jacobs, Johns Hopkins All Children s Heart Institute; Saint Petersburg; Joe B. Putnam Jr, Baptist MD Anderson Cancer Center, Jacksonville, FL; Cameron D. Wright, Massachusetts General Hospital, Boston, MA; and Felix G. Fernandez, Emory University, Atlanta, GA. Support Supported by Grant No. R01 HS from the Agency for Healthcare Research and Quality. nnn by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

9 VATS Versus Thoracotomy for Lung Cancer AUTHORS DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST Minimally Invasive Lung Cancer Surgery Performed by Thoracic Surgeons as Effective as Thoracotomy The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO s conflict of interest policy, please refer to or ascopubs.org/jco/site/ifc. Daniel J. Boffa Research Funding: Epic Sciences Andrzej S. Kosinski No relationship to disclose Anthony P. Furnary Stock or Other Ownership: Edwards Lifesciences Honoraria: Edwards Lifesciences Speakers Bureau: Edwards Lifesciences Patents, Royalties, Other Intellectual Property: Intellectual property rights for The Portland Protocol and multiple patents in the United States, Australia, the United Kingdom, and Canada Travel, Accommodations, Expenses: Edwards Lifesciences Sunghee Kim No relationship to disclose Mark W. Onaitis No relationship to disclose Betty C. Tong Consulting or Advisory Role: Medtronic Patricia A. Cowper Research Funding: AstraZeneca, Eli Lilly, GE Healthcare, Bristol-Myers Squibb, Pfizer, Tenax Therapeutics, Gilead Sciences, Bayer, Novartis Jessica R. Hoag No relationship to disclose Jeffrey P. Jacobs No relationship to disclose Cameron D. Wright No relationship to disclose Joe B. Putnam Jr Consulting or Advisory Role: Merck Travel, Accommodations, Expenses: Merck Felix G. Fernandez No relationship to disclose jco.org 2018 by American Society of Clinical Oncology

10 Boffa et al Appendix A Zubrod score of 2 5 Zubrod score of 1 Volume per year for 10-volume increase Upstaging Surgery year 2013 Surgery year 2012 Surgery year 2011 Surgery year 2010 Current smoker Past smoker Race, other Race, black/african American PET Invasive mediastinal staging Corticosteroid use Chronic kidney disease PVD Hypertension Diabetes CVD CHF CAD BMI > 35.0 kg/m 2 BMI kg/m 2 BMI kg/m 2 BMI < 18.5 kg/m 2 FEV 1 predicted 60%-80% FEV 1 predicted 40%-60% FEV 1 predicted < 40% Female Clinical T2a Clinical T1b CT scan of chest Brain imaging ASA class V ASA class IV ASA class III ASA class II Age 80 years Age years Age years Before matching After matching Standardized Difference (%) Fig A1. The standardized differences between thoracotomy and VATS patients are plotted for each variable for (A) clinical stage I patients and (B) pathologic stage I patients both before (blue triangle) and after (yellow square) propensity matching. ASA, American Society of Anesthesiologists; BMI, body mass index; CAD, coronary artery disease; CHF, congestive heart failure; CT, computed tomography; CVD, cardiovascular disease; FEV 1, forced expiratory volume in 1 second; PET, positron emission tomography; PVD, peripheral vascular disease by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

11 VATS Versus Thoracotomy for Lung Cancer B Zubrod score of 2 5 Zubrod score of 1 Volume per year for 10-volume increase Surgery year 2013 Surgery year 2012 Surgery year 2011 Surgery year 2010 Current smoker Past smoker Race, other Race, black/african American PET Pathologic T2a Pathologic T1b Invasive mediastinal staging Corticosteroid use Chronic kidney disease PVD Hypertension Diabetes CVD CHF CAD BMI > 35.0 kg/m 2 BMI kg/m 2 BMI kg/m 2 BMI < 18.5 kg/m 2 FEV 1 predicted 60%-80% FEV 1 predicted 40%-60% FEV 1 predicted < 40% Female CT scan of chest Brain imaging ASA class V ASA class IV ASA class III ASA class II Age 80 years Age years Age years Before matching After matching Standardized Difference (%) Fig A1. (Continued). jco.org 2018 by American Society of Clinical Oncology

12 Boffa et al Table A1. Detailed Overview of Clinical Stage I Populations No. (%)* Variable Thoracotomy (n = 5,219) VATS (n = 6,071) Conversion (n = 732) P Median age, years (interquartile range) 73.0 ( ) 73.0 ( ) 73.0 ( ).24 Age, years ,468 (28.1) 1,583 (26.1) 197 (26.9) ,552 (29.7) 1,895 (31.2) 233 (31.8) ,284 (24.6) 1,505 (24.8) 186 (25.4) $ (17.5) 1,088 (17.9) 116 (15.8) Sex Male 2,578 (49.4) 2,630 (43.3) 361 (49.3) Female 2,636 (50.5) 3,440 (56.7) 371 (50.7) Missing 5 (0.1) 1 (, 0.1) 0 (0.0) Race.54 White 4,651 (89.1) 5,503 (90.6) 648 (88.5) Black/African American 302 (5.8) 320 (5.3) 43 (5.9) Other 11 (0.2) 16 (0.3) 3 (0.4) Missing 255 (4.9) 232 (3.8) 38 (5.2) Year of surgery (1.5) 12 (0.2) 0 (0.0) (2.6) 5 (0.1) 1 (0.1) (5.1) 24 (0.4) 15 (2.0) (6.2) 71 (1.2) 42 (5.7) (7.2) 146 (2.4) 53 (7.2) (7.8) 281 (4.6) 82 (11.2) (8.1) 405 (6.7) 135 (18.4) (13.3) 619 (10.2) 69 (9.4) (12.5) 817 (13.5) 84 (11.5) (13.1) 1,048 (17.3) 97 (13.3) (11.5) 1,220 (20.1) 75 (10.2) (11.0) 1,423 (23.4) 79 (10.8) BMI, kg/m 2.023, (2.4) 133 (2.2) 14 (1.9) ,575 (30.2) 2,022 (33.3) 221 (30.2) ,160 (41.4) 2,473 (40.7) 303 (41.4) (18.0) 977 (16.1) 135 (18.4) (8.1) 466 (7.7) 59 (8.1) American Society of Anesthesiologists risk class I or II 666 (12.8) 982 (16.2) 103 (14.1) III 3,712 (71.1) 4,639 (76.4) 546 (74.6) IV or V 780 (14.9) 440 (7.2) 77 (10.5) Missing 61 (1.2) 10 (0.2) 6 (0.8) Zubrod score 0 2,212 (42.4) 3,122 (51.4) 294 (40.2) 1 2,700 (51.7) 2,724 (44.9) 393 (53.7) (4.9) 202 (3.3) 42 (5.7) Missing 51 (1.0) 23 (0.4) 3 (0.4) Coronary artery disease No 3,355 (64.3) 4,366 (71.9) 484 (66.1) Yes 1,443 (27.6) 1,449 (23.9) 208 (28.4) Missing 421 (8.1) 256 (4.2) 40 (5.5) Cerebrovascular disease.025 No 4,455 (85.3) 5,319 (87.6) 635 (86.7) Yes 552 (10.6) 555 (9.1) 74 (10.1) Missing 212 (4.1) 197 (3.2) 23 (3.1) Congestive heart failure No 4,560 (87.4) 5,641 (92.9) 668 (91.3) Yes 230 (4.4) 172 (2.8) 23 (3.1) Missing 429 (8.2) 258 (4.2) 41 (5.6) Hypertension No 1,292 (24.8) 1,801 (29.7) 167 (22.8) Yes 3,516 (67.4) 4,034 (66.4) 530 (72.4) Missing 411 (7.9) 236 (3.9) 35 (4.8) Diabetes mellitus.004 No 4,034 (77.3) 4,846 (79.8) 557 (76.1) Yes 1,048 (20.1) 1,102 (18.2) 160 (21.9) Missing 137 (2.6) 123 (2.0) 15 (2.0) Corticosteroid use.009 No 4,552 (87.2) 5,560 (91.6) 658 (89.9) (continued on following page) 2018 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

13 VATS Versus Thoracotomy for Lung Cancer Table A1. Detailed Overview of Clinical Stage I Populations (continued) No. (%)* Variable Thoracotomy (n = 5,219) VATS (n = 6,071) Conversion (n = 732) P Yes 180 (3.4) 170 (2.8) 33 (4.5) Missing 487 (9.3) 341 (5.6) 41 (5.6) Peripheral vascular disease No 4,131 (79.2) 5,184 (85.4) 587 (80.2) Yes 665 (12.7) 632 (10.4) 105 (14.3) Missing 423 (8.1) 255 (4.2) 40 (5.5) Chronic kidney disease (creatinine. 2 mg/dl or dialysis).010 No 4,894 (93.8) 5,759 (94.9) 694 (94.8) Yes 149 (2.9) 121 (2.0) 19 (2.6) Missing 176 (3.4) 191 (3.1) 19 (2.6) Median FEV 1, % predicted (interquartile range) 80.0 ( ) 85.0 ( ) 80.0 ( ) Missing 515 (9.9) 297 (4.9) 74 (10.1) Smoking status Never smoked 681 (13.0) 951 (15.7) 103 (14.1) Past smoker (stopped. 1 month before operation) 3496 (67.0) 4,044 (66.6) 497 (67.9) Current smoker 1,024 (19.6) 1,072 (17.7) 131 (17.9) Missing 18 (0.3) 4 (0.1) 1 (0.1) Operative mortality within 30 days from surgery (composite of CMS and STS operative mortality) No 5,073 (97.2) 5,998 (98.8) 708 (96.7) Yes 146 (2.8) 73 (1.2) 24 (3.3) Operative mortality within 90 days from surgery (composite of CMS and STS operative mortality) No 4,978 (95.4) 5,934 (97.7) 697 (95.2) Yes 241 (4.6) 137 (2.3) 35 (4.8) Adjuvant chemotherapy (CMS) for surgeries.002 No 3,031 (83.5) 4,770 (86.2) 460 (85.3) Yes 597 (16.5) 762 (13.8) 79 (14.7) Median days from surgery date to drug date (CMS) for 55.0 ( ) 50.0 ( ) 49.0 ( ) surgeries (interquartile range) CT scan of chest (CMS) for surgeries.028 No 387 (10.7) 530 (9.6) 69 (12.8) Yes 3,241 (89.3) 5,002 (90.4) 470 (87.2) Brain imaging (CMS) for surgeries No 2,368 (65.3) 3,943 (71.3) 386 (71.6) Yes 1,260 (34.7) 1,589 (28.7) 153 (28.4) PET scan (CMS) for surgeries.021 No 550 (15.2) 802 (14.5) 102 (18.9) Yes 3,078 (84.8) 4,730 (85.5) 437 (81.1) Invasive mediastinal staging (EBUS, mediastinoscopy, EUS, or multiple; CMS) for surgeries No 2,410 (66.4) 3,918 (70.8) 357 (66.2) Yes 1,218 (33.6) 1,614 (29.2) 182 (33.8) Abbreviations: BMI, body mass index; CMS, Centers for Medicare and Medicaid Services; CT, computed tomography; EBUS, endobronchial ultrasound; EUS, endoscopic ultrasound FEV 1, forced expiratory volume in 1 second; PET, positron emission tomography; STS, Society of Thoracic Surgeons; VATS, video-assisted thoracic surgery. *Values are numbers and percentages (in parentheses), unless otherwise noted. P values based on nonmissing data. jco.org 2018 by American Society of Clinical Oncology