Advertisement

Improved Overall Survival with Aggressive Primary Tumor Radiotherapy for Patients with Metastatic Esophageal Cancer

Open ArchivePublished:April 28, 2017DOI:https://doi.org/10.1016/j.jtho.2017.03.026

      Abstract

      Objectives

      The aim of this study was to characterize utilization and survival outcomes associated with primary tumor–directed radiotherapy (PTDRT) in patients with newly diagnosed metastatic esophageal cancer.

      Methods

      We conducted an observational cohort study using the National Cancer Data Base to evaluate patients with newly diagnosed metastatic esophageal cancer between 2004 and 2012. Overall survival outcomes after treatment with chemotherapy plus conventional palliative dose radiotherapy (<5040 cGy), chemotherapy plus definitive dose radiotherapy (≥5040 cGy), or chemotherapy alone were compared by using Cox proportional hazards models with inverse probability of treatment weighting using the propensity score. Potential unmeasured confounding was assessed through sensitivity analyses.

      Results

      The final cohort consisted of 12,683 patients: 57% were treated with chemotherapy alone, 24% were treated with chemotherapy plus palliative dose radiotherapy, and 19% were treated with chemotherapy plus definitive dose radiotherapy. Compared with chemotherapy alone, chemotherapy plus definitive dose radiotherapy was associated with improved survival (median overall survival of 8.3 versus 11.3 months [hazard ratio = 0.72, 95% confidence interval: 0.70–0.74, p ≤ 0.001]), whereas chemotherapy plus palliative dose radiotherapy was associated with slightly inferior outcomes (median overall survival of 8.3 months versus 7.5 months (hazard ratio = 1.10, 95% confidence interval 1.07–1.13, p ≤ 0.001). These findings were robust to potential unmeasured confounding in sensitivity analyses. Additionally, landmark analyses confirmed these findings in patients surviving 12 months or longer.

      Conclusions

      Definitive dose, but not conventional palliative dose, PTDRT is associated with improved overall survival in metastatic esophageal cancer, suggesting that local control may be important to prognosis. These findings support integrating PTDRT into future clinical trials aimed at refining personalized treatment for patients with metastatic esophageal cancer.

      Keywords

      Introduction

      Esophageal cancer is diagnosed in roughly 17,000 U.S. adults annually, and up to half of these patients present with metastatic disease at initial diagnosis. Standard treatment approaches have remained largely unchanged over the past two decades, and the 5-year survival rate in this population is less than 5%.
      • Siegel R.L.
      • Miller K.D.
      • Jemal A.
      Cancer statistics, 2016.

      Surveillance Epidemiology and End Results Program. Previous version: SEER cancer statistics review, 1975-2013. http://seer.cancer.gov/csr/1975_2013/. Accessed November 10, 2016.

      Significant primary tumor progression has been reported to occur in 30% to 40% of patients with metastatic esophageal cancer treated by chemotherapy alone.
      • Hingorani M.
      • Dixit S.
      • Johnson M.
      • et al.
      Palliative radiotherapy in the presence of well-controlled metastatic disease after initial chemotherapy may prolong survival in patients with metastatic esophageal and gastric cancer.
      Furthermore, primary tumor progression is a common cause of morbidity and mortality in this setting, often leading to dysphagia and subsequent malnutrition, chronic bleeding, and/or direct invasion of adjacent vital organs.
      • Riccardi D.
      • Allen K.
      Nutritional management of patients with esophageal and esophagogastric junction cancer.
      • Saito T.
      • Kuwahara A.
      • Shigemitsu Y.
      • et al.
      Factors related to malnutrition in patients with esophageal cancer.
      • Tomaselli F.
      • Maier A.
      • Sankin O.
      • Woltsche M.
      • Pinter H.
      • Smolle-Jüttner F.M.
      Successful endoscopical sealing of malignant esophageotracheal fistulae by using a covered self-expandable stenting system.
      • Maluf-Filho F.
      • Martins B.C.
      • de Lima M.S.
      • et al.
      Etiology, endoscopic management and mortality of upper gastrointestinal bleeding in patients with cancer.
      Frequently, short-course primary tumor–directed radiotherapy (PTDRT) is offered to patients with metastatic esophageal cancer to palliate such symptoms. However, a growing body of evidence suggests that survival benefits may be associated with aggressive therapy directed to primary tumor sites in patients with metastatic cancers.
      • Rusthoven C.G.
      • Jones B.L.
      • Flaig T.W.
      • et al.
      Improved survival with prostate radiation in addition to androgen deprivation therapy for men with newly diagnosed metastatic prostate cancer.
      • Seisen T.
      • Sun M.
      • Leow J.J.
      • et al.
      Efficacy of high-intensity local treatment for metastatic urothelial carcinoma of the bladder: a propensity score-weighted analysis from the National Cancer Data Base.
      • Culp S.H.
      • Schellhammer P.F.
      • Williams M.B.
      Might men diagnosed with metastatic prostate cancer benefit from definitive treatment of the primary tumor? A SEER-based study.
      • Ouyang W.W.
      • Su S.F.
      • Hu Y.X.
      • et al.
      Radiation dose and survival of patients with stage IV non-small cell lung cancer undergoing concurrent chemotherapy and thoracic three-dimensional radiotherapy: reanalysis of the findings of a single-center prospective study.
      • Slotman B.J.
      • van Tinteren H.
      • Praag J.O.
      • et al.
      Use of thoracic radiotherapy for extensive stage small-cell lung cancer: a phase 3 randomised controlled trial.
      • Gomez D.R.
      • Blumenschein Jr., G.R.
      • Lee J.J.
      • et al.
      Local consolidative therapy versus maintenance therapy or observation for patients with oligometastatic non-small-cell lung cancer without progression after first-line systemic therapy: a multicentre, randomised, controlled, phase 2 study.
      We therefore examined recent utilization and survival outcomes associated with various regimens of PTDRT among patients with newly diagnosed metastatic esophageal cancer by using a large national cancer registry. We hypothesized that definitive dose PTDRT would be associated with improved overall survival (OS) in patients with metastatic esophageal cancer.

      Materials and Methods

      Patient Selection

      We conducted an observational cohort study to examine national-level patterns of PTDRT utilization in patients with newly diagnosed metastatic esophageal cancer and to assess its impact on patients' OS. Our study sample was drawn from the National Cancer Data Base (NCDB) from 2004 to 2012. This national database, which is jointly sponsored by the American College of Surgeons and the American Cancer Society, draws on hospital registry data from more than 1500 Commission on Cancer–accredited facilities in the United States and Puerto Rico. The NCDB represents roughly 70% of new cancer diagnoses in this region. This study was granted exemption from the need for review by our institutional review board.
      Inclusion criteria for the analysis included all patients with metastatic primary invasive esophageal squamous cell carcinoma or adenocarcinoma (International Classification of Diseases for Oncology, Third Edition, codes 150–159) treated with primary chemotherapy with or without radiation directed to the esophagus or stomach. Patients in whom radiotherapy was delivered to any other region were excluded. Patients staged as M1a before the update of the seventh edition of the American Joint Committee on Cancer staging manual in 2010 were reclassified as having locally advanced disease and excluded, as were patients without evidence of a primary invasive tumor. Lastly, patients treated with radiation doses less than 20 Gy or greater than 72 Gy were excluded, as these doses are not within the conventional range of palliative or definitive radiotherapy to the esophagus.

      Patient Cohorts and Variables

      Patient cohorts were defined as those treated with chemotherapy alone, those treated with chemotherapy plus conventional palliative dose PTDRT (<5040 cGy), and those treated with chemotherapy plus definitive dose PTDRT (≥5040 cGy). This cutoff was chosen to reflect the dose that constitutes definitive treatment in the curative setting.
      • Minsky B.D.
      • Pajak T.F.
      • Ginsberg R.J.
      • et al.
      INT 0123 (Radiation Therapy Oncology Group 94-05) phase III trial of combined-modality therapy for esophageal cancer: high-dose versus standard-dose radiation therapy.
      • Brower J.V.
      • Chen S.
      • Bassetti M.F.
      • et al.
      Radiation dose escalation in esophageal cancer revisited: a contemporary analysis of the National Cancer Data Base, 2004 to 2012.
      Such a stratification would allow us to better assess the effect of definitive dose therapy, compared with chemotherapy, on outcomes without obscuring our findings by patients treated with palliative doses of radiation. Baseline characteristics of patients from these three cohorts were compared by using chi-square tests. Covariates examined included patient age, sex, race/ethnicity, population density (classified as metropolitan, urban, and rural), facility type (academic versus nonacademic), facility geographic region, primary insurance provider, education level (defined as percentage of the population in the patient’s home zip code not achieving a high school degree), income (the median income in the patient’s home zip code), Charlson/Deyo comorbidity score, histologic subtype, American Joint Committee on Cancer T stage and N stage, and treatment year.
      Although not directly recorded in the NCDB, the average dose per fraction was calculated by dividing the total radiation dose by the number of radiation treatments. Additionally, concurrent chemoradiation was defined as chemotherapy and radiation starting within 5 days of one another.

      Statistical Analysis

      The primary objective was to assess the effect of radiotherapy to the primary tumor site on OS, as defined from the date of diagnosis until death or last follow-up. The secondary objective was to characterize patterns of high-dose radiotherapy utilization in metastatic esophageal cancer across the United States.
      Multivariable logistic regression was used to assess the independent effects of all covariates on the odds of being treated with high-dose radiation relative to low-dose radiation or chemotherapy alone. To determine the association between radiotherapy use and OS, we then developed a Cox proportional hazards model to assess the independent effect of high-dose radiotherapy and low-dose radiotherapy on OS compared with chemotherapy alone. Proportional hazards assumptions were tested by using Schoenfeld residuals tests and were not violated. Landmark analyses in patients surviving at least 3, 6, and 12 months were conducted to control for immortal time bias.
      • Park H.S.
      • Gross C.P.
      • Makarov D.V.
      • Yu J.B.
      Immortal time bias: a frequently unrecognized threat to validity in the evaluation of postoperative radiotherapy.
      We estimated a three-level propensity score to obtain the probabilities of high-dose versus low-dose PTDRT versus chemotherapy only for each patient.
      • Burgette L.F.
      • McCaffrey D.F.
      • Griffin B.A.
      Propensity score estimation with boosted regression.
      • McCaffrey D.F.
      • Griffin B.A.
      • Almirall D.
      • Slaughter M.E.
      • Ramchand R.
      • Burgette L.F.
      A tutorial on propensity score estimation for multiple treatments using generalized boosted models.

      Ridgeway G, McCaffrey D, Griffin BA, and Burgette L. twang: toolkit for weighting and analysis of non-equivalent groups. http://cran.rproject.org/web/packages/twang/vignettes/twang.pdf. Accessed November 23, 2016.

      To accomplish this, we used the twang package in R, which uses generalized boosted models and bootstrapping to produce inverse probability of treatment weights (IPTWs) in studies in which more than two treatments are being compared. Our final Cox model incorporated these weights on the basis of propensity score to account for confounding due to measured confounders.
      • Austin P.C.
      The performance of different propensity score methods for estimating marginal hazard ratios.
      Finally, to assess the robustness of our findings to potential unmeasured confounders, we conducted a regression-based sensitivity analysis,
      • Mitra N.
      • Heitjan D.F.
      Sensitivity of the hazard ratio to nonignorable treatment assignment in an observational study.
      using presence of high-volume metastatic disease as a potential unmeasured confounding variable in our Cox model. We estimated that its prevalence would be 50% in our population and that it would correspond to a hazard ratio (HR) of death between 1.5 and 3.0 on the basis of the published literature.
      • Cunningham D.
      • Starling N.
      • Rao S.
      • et al.
      Capecitabine and oxaliplatin for advanced esophagogastric cancer.
      • Bang Y.J.
      • Van Cutsem E.
      • Feyereislova A.
      • et al.
      Trastuzumab in combination with chemotherapy versus chemotherapy alone for treatment of HER2-positive advanced gastric or gastro-oesophageal junction cancer (ToGA): a phase 3, open-label, randomised controlled trial.
      • Parikh R.B.
      • Cronin A.M.
      • Kozono D.E.
      • et al.
      Definitive primary therapy in patients presenting with oligometastatic non-small cell lung cancer.
      The difference in median time between chemotherapy and radiation in patients treated with sequential therapy was determined by the Wilcoxon rank sum test.
      All analyses were performed in R, version 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria). For all analyses, a p value of 0.05 or less was considered statistically significant.

      Results

      Population and Patterns of Care

      After applying our inclusion and exclusion criteria, our final cohort consisted of 12,683 patients. Of these patients, 7274 (57%) were treated with chemotherapy alone, 2983 (24%) were treated with chemotherapy plus palliative dose PTDRT, and 2426 (19%) were treated with chemotherapy plus definitive dose PTDRT (see Supplementary Fig. 1). In the low-dose arm, dose per fraction spanned 180 to 350 cGy. In contrast, 85% of patients in the high-dose arm were treated in fractions of less than 200 cGy (see Supplementary Fig. 2).
      Most of the patients included were male (85%), non-Hispanic white (86%), resided in metropolitan locations (78%), and were covered by Medicare or commercial insurance (83% combined); 78% were without significant comorbid illness and 75% had tumors of the adenocarcinoma histologic subtype (Table 1). Of the patients treated with radiotherapy, 2333 (43%) received sequential chemotherapy (58% of whom received chemotherapy followed by radiation) and 2772 (51%) received concurrent chemotherapy; for 304 patients (6%), the available data were insufficient to determine sequence. Concurrent chemotherapy was more common in patients treated with high-dose radiotherapy than in those treated with low-dose radiotherapy (59% versus 45%, OR = 1.78, 95% confidence interval [CI]: 1.60–2.00, p < 0.001). Patients treated sequentially were more often treated with chemotherapy first (median number of days between starting radiation and starting chemotherapy = 9 [interquartile range = 74] for the overall population). This time span was longer in the case of patients in the definitive dose PTDRT cohort than in the case of those in the palliative cohort (median 14 versus 6 days [p < 0.001]). Regarding chemotherapy regimens in the population, 81% received multiagent chemotherapy, 10% received single-agent chemotherapy, and the regimen was unknown in 9%.
      Table 1Patient Characteristics
      CharacteristicNo RT (n =7274)RT <5040 cGy (n = 2983)RT ≥5040 cGy (n = 2426)Total (N = 12,683)
      Age, y, n (%)
       ≤49789 (11)317 (11)205 (8)1311 (10)
       50–643286 (45)1283 (43)1048 (43)5617 (44)
       65–792770 (38)1117 (37)993 (41)4880 (38)
       ≥ 80429 (6)266 (9)180 (7)875 (7)
      Sex, n (%)
       Male6194 (85)2454 (82)1963 (81)10611 (84)
       Female1080 (15)529 (18)463 (19)2072 (16)
      Race, n (%)
       Non-Hispanic white6286 (86)2445 (82)1938 (80)10669 (84)
       Non-Hispanic black544 (7)341 (11)312 (13)1197 (9)
       Hispanic257 (4)108 (4)94 (4)459 (4)
       Other187 (3)89 (3)82 (3)358 (3)
      County size, n (%)
       Metropolitan5648 (78)2292 (77)1841 (76)9781 (77)
       Urban1159 (16)499 (17)432 (18)2090 (16)
       Rural144 (2)70 (2)53 (2)267 (2)
       Unknown323 (4)122 (4)100 (4)545 (4)
      Facility type, n (%)
       Non-academic4276 (60)2031 (69)1688 (70)7995 (63)
       Academic2873 (40)914 (31)709 (30)4496 (35)
      Facility location, n (%)
       New England475 (7)179 (6)167 (7)821 (6)
       Mid-Atlantic region1331 (18)409 (14)367 (15)2107 (17)
       South Atlantic region1393 (19)601 (20)470 (19)2464 (19)
       Midwest3002 (41)1254 (42)1006 (41)5262 (41)
       Mountain region286 (4)145 (5)118 (5)549 (4)
       Pacific region662 (9)357 (12)269 (11)1288 (10)
       Unknown125 (2)38 (1)29 (1)192 (2)
      Insurance status, n (%)
       Commercial insurance3075 (42)1129 (38)956 (39)5160 (41)
       Medicare2972 (41)1329 (45)1103 (45)5404 (43)
       Medicaid517 (7)272 (9)192 (8)981 (8)
       Uninsured301 (4)148 (5)91 (4)540 (4)
       Other government insurance71 (1)66 (2)43 (2)180 (1)
       Unknown338 (5)39 (1)41 (2)418 (3)
      Educational attainment, n (%)
      Educational attainment defined as percentage of the population in the patient’s home zip code not achieving a high school degree.
       ≥29%1023 (14)483 (16)386 (16)1892 (15)
       20%–28.9%1573 (22)711 (24)588 (24)2872 (23)
       14%–19.9%1810 (25)737 (25)573 (24)3120 (25)
       <14%2516 (35)933 (31)766 (32)4215 (33)
       Unknown352 (5)119 (4)113 (5)584 (5)
      Income, n (%)
       <$30,000848 (12)398 (13)350 (14)1596 (13)
       $30,000–$35,0001247 (17)549 (18)447 (18)2243 (18)
       $35,000–$45,9991936 (27)878 (29)658 (27)3472 (27)
       >$46,0002891 (40)1039 (35)858 (35)4788 (38)
       Unknown352 (5)119 (4)113 (5)584 (5)
      Charlson/Deyo comorbidity score, n (%)
       05676 (78)2331 (78)1920 (79)9927 (78)
       11241 (17)524 (18)397 (16)2162 (17)
       ≥2357 (5)128 (4)109 (4)594 (5)
      Histologic subtype, n (%)
       Squamous cell carcinoma1353 (19)815 (27)865 (36)3033 (24)
       Adenocarcinoma5460 (75)1982 (66)1463 (60)8905 (70)
       Unknown461 (6)186 (6)98 (4)745 (6)
      T stage
       T1696 (10)182 (6)159 (7)1037 (8)
       T2325 (4)217 (7)226 (9)768 (6)
       T31301 (18)885 (30)859 (35)3045 (24)
       T41094 (15)474 (16)364 (15)1932 (15)
       Unknown3858 (53)1225 (41)818 (34)5901 (47)
      N stage, n (%)
       Node negative1088 (15)464 (16)417 (17)1969 (16)
       Node positive4091 (56)1856 (62)1604 (66)7551 (60)
       Unknown2095 (29)663 (22)405 (17)3163 (25)
      Treatment year, n (%)
       2004574 (8)298 (10)267 (11)1139 (9)
       2005572 (8)288 (10)246 (10)1106 (9)
       2006662 (9)282 (9)290 (12)1234 (10)
       2007763 (10)333 (11)292 (12)1388 (11)
       2008799 (11)328 (11)290 (12)1417 (11)
       2009957 (13)361 (12)269 (11)1587 (13)
       2010931 (13)382 (13)285 (12)1598 (13)
       20111000 (14)323 (11)262 (11)1585 (12)
       20121016 (14)388 (13)225 (9)1629 (13)
      RT, radiotherapy.
      a Educational attainment defined as percentage of the population in the patient’s home zip code not achieving a high school degree.
      On multivariable analysis, clinical stage T3 versus T1 disease predicted for a higher rate of high-dose PTDRT relative to low-dose radiation or chemotherapy only (28% versus 15%, OR = 2.95, 95% CI: 2.53–3.46], p < 0.001), as did being African American rather than non-Hispanic white (26% versus 18%, OR = 1.34, 95% CI: 1.16–1.56, p < 0.001). Additional factors associated with increased utilization of high-dose PTDRT included age older than 80 years, facility location in the Mountain region, and non-Medicare/non-Medicaid government insurance status. Conversely, rates of high-dose PTDRT were lower in patients with the adenocarcinoma histologic subtype relative to squamous cell carcinoma (16% versus 29%, OR = 0.57, 95% CI: 0.51–0.63, p < 0.001). Other factors predicting for less use of high-dose PTDRT included treatment at an academic facility, facility location in the Mid-Atlantic region, and Charlson/Deyo comorbidity index score of 2 or more (Table 2). In addition, there was a statistically significant increase in utilization of chemotherapy alone over time (p trend < 0.001 [Fig. 1]). Factors associated with health care disparity, such as insurance provider, income, and educational attainment, were not distributed significantly differently between treatment groups.
      Table 2Factors Associated with Receipt of High-Dose RT
      CharacteristicPatients Treated with RT ≥5040 cGyUnivariable AnalysisMultivariable Analysis
      OR (95% CI)p ValueOR (95% CI)p Value
      Age, y
       ≤4916%
       50–6419%1.07 (0.95–1.21)0.271.00 (0.88–1.15)0.91
       65–7920%1.15 (1.02–1.30)0.0261.08 (0.91–1.27)0.36
       ≥8021%1.57 (1.32–1.87)<0.0011.55 (1.26–1.92)<0.001
      Sex
       Male18%
       Female22%1.29 (1.17–1.42)<0.0011.1 (0.99–1.22)0.077
      Race
       Non-Hispanic white18%
       Non-Hispanic black26%1.72 (1.53–1.94)<0.0011.34 (1.16–1.56)<0.001
       Hispanic20%1.13 (0.93–1.36)0.211.10 (0.89–1.35)0.41
       Other23%1.31 (1.06–1.62)0.0121.19 (0.94–1.50)0.16
      County size
       Metropolitan19%
       Urban21%1.10 (1.00–1.21)0.0551.01 (0.90–1.12)0.92
       Rural20%1.17 (0.91–1.49)0.211.03 (0.79–1.34)0.86
       Unknown18%0.94 (0.79–1.12)0.481.03 (0.81–1.32)0.76
      Facility type
       Non-academic21%
       Academic16%0.65 (0.60–0.70)<0.0010.63 (0.58–0.68)<0.001
      Facility location
       New England20%
       Mid-Atlantic region17%0.80 (0.68–0.94)0.0080.88 (0.74–1.05)0.17
       South Atlantic region19%1.06 (0.90–1.24)0.510.94 (0.80–1.12)0.56
       Midwest19%1.03 (0.89–1.20)0.661.10 (0.94–1.29)0.20
       Mountain region21%1.26 (1.02–1.57)0.0361.44 (1.14–1.81)0.001
       Pacific region21%1.30 (1.09–1.55)0.0041.25 (1.04–1.51)0.013
       Unknown15%0.74 (0.53–1.02)0.066
      Not analyzed because of collinearity.
      Not analyzed because of collinearity.
      Insurance status
       Commercial insurance19%
       Medicare20%1.21 (1.12–1.30)<0.0011.04 (0.92–1.16)0.60
       Medicaid20%1.32 (1.15–1.52)<0.0011.16 (1.00–1.35)0.056
       Uninsured17%1.17 (0.98–1.40)0.0831.06 (0.87–1.29)0.54
       Other government insurance24%2.26 (1.67–3.08)<0.0012.15 (1.56–2.96)<0.001
       Unknown10%0.35 (0.27–0.45)<0.0010.38 (0.29–0.50)<0.001
      Educational attainment
      Educational attainment defined as percentage of the population in the patient’s home zip code not achieving a high school degree.
       ≥29%20%
       20%–28.9%20%0.97 (0.87–1.09)0.641.07 (0.93–1.23)0.35
       14%–19.9%18%0.85 (0.76–0.96)0.0060.98 (0.84–1.13)0.73
       <14%18%0.79 (0.71–0.89)<0.0010.94 (0.80–1.11)0.466
       Unknown19%0.78 (0.64–0.94)0.0080.84 (0.64–1.1)0.195
      Income
       <$30,00022%
       $30,000–$35,00020%0.91 (0.80–1.03)0.130.95 (0.82–1.10)0.48
       $35,000–$45,99919%0.90 (0.80–1.01)0.0811.00 (0.86–1.17)0.99
       >$46,00018%0.74 (0.66–0.83)<0.0010.92 (0.78–1.09)0.32
       Unknown19%0.75 (0.62–0.91)0.003
      Not analyzed because of collinearity.
      Not analyzed because of collinearity.
      Charlson/Deyo comorbidity score
       019%
       118%0.99 (0.90–1.09)0.850.96 (0.87–1.06)0.42
       ≥218%0.89 (0.75–1.05)0.160.84 (0.70–1.01)0.071
      Histologic subtype
       Squamous cell carcinoma29%
       Adenocarcinoma16%0.51 (0.47–0.55)<0.0010.57 (0.51–0.63)<0.001
       Unknown13%0.50 (0.42–0.58)<0.0010.54 (0.46–0.65)<0.001
      T stage
       T115%
       T229%2.78 (2.30–3.38)<0.0012.77 (2.27–3.39)<0.001
       T328%2.74 (2.36–3.18)<0.0012.95 (2.53–3.46)<0.001
       T419%1.56 (1.34–1.83)<0.0011.47 (1.25–1.74)<0.001
       Unknown14%1.08 (0.94–1.24)0.281.17 (1.01–1.36)0.036
      N stage
       Node negative21%
       Node positive21%1.04 (0.95–1.15)0.300.96 (0.86–1.07)0.40
       Unknown13%0.63 (0.56–0.71)0.0190.73 (0.64–0.83)0.030
      Treatment year
       200423%
       200522%0.95 (0.80–1.12)0.530.91 (0.76–1.08)0.292
       200624%0.88 (0.75–1.03)0.110.83 (0.70–0.99)0.038
       200721%0.83 (0.71–0.97)0.0220.78 (0.66–0.92)0.003
       200820%0.79 (0.67–0.92)0.0030.70 (0.59–0.83)<0.001
       200917%0.67 (0.57–0.78)<0.0010.61 (0.52–0.72)<0.001
       201018%0.73 (0.62–0.85)<0.0010.66 (0.56–0.78)<0.001
       201117%0.59 (0.51–0.69)<0.0010.54 (0.46–0.64)<0.001
       201214%0.61 (0.53–0.71)<0.0010.54 (0.46–0.64)<0.001
      Note: Percentages in column 2 reflect the percentage of each subgroup treated with high-dose radiotherapy as a percentage of all patients in that subgroup.
      RT, radiotherapy; CI, confidence interval.
      a Not analyzed because of collinearity.
      b Educational attainment defined as percentage of the population in the patient’s home zip code not achieving a high school degree.
      Figure thumbnail gr1
      Figure 1Temporal trends in use of chemotherapy alone and chemotherapy with low- or high-dose radiotherapy in metastatic esophageal cancer. The Spearman test for trend was significant for the trend of increasing utilization of chemotherapy alone over time (p trend < 0.001).

      OS

      The median follow-up was 19.8 months (range 0.33–122.5) for patients alive at the end of the study period and 8.4 months (range 0–122.5) overall. On multivariable analysis, receipt of high-dose radiotherapy was associated with a statistically significant improvement in OS relative to no radiotherapy (median OS = 11.2 versus 8.4 months, HR = 0.72, 95% CI: 0.68–0.75, p ≤ 0.001 [Table 3]). In contrast, low-dose radiation was associated with worse OS relative to chemotherapy alone, although the magnitude of the difference was small (median OS = 7.6 months versus 8.4 months, HR = 1.07, 95% CI: 1.02–1.12, p = 0.004). An interaction term was introduced into the model to assess a potential interaction between use of radiation and concurrent versus sequential chemoradiation. In patients who were treated with radiation, neither the use of concurrent chemoradiation (p = 0.08) nor the interaction between concurrent chemoradiation and low-dose radiotherapy (p = 0.44) nor high-dose radiotherapy (p = 0.14) was statistically significant. In addition, our model did not reveal a meaningful interaction between histologic subtype and treatment cohort (HR = 0.95, 95% CI: 0.95–0.99, p = 0.05).
      Table 3Multivariable Cox Proportional Hazards Model for Overall Survival
      CharacteristicCovariate AdjustedPropensity Score–Weighted
      HR (95% CI)p ValueHR (95% CI)p Value
      Radiation dose
       No RT
       <50401.07 (1.02–1.12)0.0041.10 (1.07–1.13)<0.001
       ≥5040 cGy0.72 (0.68–0.75)<0.0010.72 (0.70–0.74)<0.001
      Age, y
       ≤49
       50–641.00 (0.93–1.07)0.99
       65–790.99 (0.91–1.07)0.71
       ≥801.22 (1.10–1.35)<0.001
      Sex
       Male
       Female0.92 (0.87–0.96)<0.001
      Race
       Non-Hispanic white
       Non-Hispanic black0.93 (0.86–1.00)0.044
       Hispanic0.78 (0.70–0.87)<0.001
       Other0.88 (0.78–0.99)0.037
      County size
       Metropolitan
       Urban1.01 (0.96–1.07)0.67
       Rural0.95 (0.83–1.08)0.42
       Unknown1.02 (0.90–1.15)0.77
      Facility type
       Nonacademic
       Academic0.92 (0.88–0.96)<0.001
      Facility location
       New England
       Mid-Atlantic region0.96 (0.88–1.05)0.38
       South Atlantic region1.05 (0.96–1.14)0.29
       Midwest1.05 (0.97–1.13)0.22
       Mountain region1.02 (0.91–1.15)0.71
       Pacific region1.00 (0.91–1.09)0.92
      Insurance status
       Commercial insurance
       Medicare1.09 (1.03–1.16)0.002
       Medicaid1.20 (1.11–1.30)<0.001
       Uninsured1.17 (1.06–1.29)0.001
       Other government insurance1.03 (0.88–1.20)0.76
       Unknown0.95 (0.85–1.06)0.37
      Educational attainment
      Educational attainment defined as percentage of the population in the patient’s home zip code not achieving a high school degree.
       ≥29%
       20%–28.9%0.96 (0.90–1.03)0.28
       14%–19.9%0.97 (0.90–1.05)0.46
       <14%1.00 (0.93–1.08)0.97
       Unknown1.01 (0.89–1.15)0.89
      Income
       <$30,000
       $30,000–$35,0000.97 (0.90–1.05)0.44
       $35,000–$45,9990.96 (0.89–1.04)0.34
       >$46,0000.95 (0.87–1.03)0.21
       Unknown
      Charlson/Deyo comorbidity score
       0
       11.15 (1.09–1.21)<0.001
       ≥21.33 (1.22–1.45)<0.001
      Histologic subtype
       Squamous cell carcinoma
       Adenocarcinoma0.99 (0.94–1.04)0.73
       Unknown1.25 (1.15–1.36)<0.001
      T stage
       T1
       T20.87 (0.79–0.96)0.006
       T30.92 (0.85–0.99)0.023
       T41.13 (1.04–1.22)0.004
       Unknown1.05 (0.98–1.13)0.17
      N stage
       Node negative
       Node positive1.06 (1.01–1.12)0.025
       Unknown1.10 (1.03–1.17)<0.001
      Treatment year
       2004
       20051.01 (0.93–1.10)0.84
       20060.90 (0.83–0.98)0.013
       20070.92 (0.85–0.99)0.037
       20080.89 (0.82–0.96)0.003
       20090.86 (0.79–0.93)<0.001
       20100.84 (0.77–0.91)<0.001
       20110.80 (0.74–0.87)<0.001
       20120.84 (0.78–0.91)<0.001
      Note: For clarity, only propensity score–weighted HRs for the primary outcome are displayed.
      HR, hazard ratio; CI, confidence interval; RT, radiotherapy.
      a Educational attainment defined as percentage of the population in the patient’s home zip code not achieving a high school degree.
      In our second analysis using propensity score weighting, receipt of high-dose radiotherapy remained statistically significantly associated with improved OS relative to chemotherapy alone (IPTW model HR = 0.72, 95% CI: 0.70–0.74, p < 0.001 [see Table 3]). In the IPTW model, median survival was 11.3 months in the high-dose PTDRT cohort and 8.3 months in the chemotherapy-alone cohort. This corresponded to 1- and 2-year OS estimates of 47% versus 34% and 19% versus 12%, respectively. Low-dose radiotherapy again had a marginally detrimental effect on OS (Fig. 2 and Table 3).
      Figure thumbnail gr2
      Figure 2Inverse probability of treatment weighting adjusted overall survival in patients with metastatic esophageal cancer treated with chemotherapy alone (no radiotherapy [RT]), radiation doses less than 5040 cGy, or radiation doses of 5040 cGy or higher. The shaded region surrounding each line represents the upper and lower bounds of the 95% confidence interval. For the comparison, the hazard ratio is 0.72 (95% confidence interval: 0.70–0.74, p ≤ 0.001).
      Landmark analyses of patients surviving a minimum of 3, 6, and 12 months were conducted to control for the effect of immortal time bias. In all circumstances, the effect of high-dose radiotherapy remained significantly associated with improved OS (Table 4).
      Table 4Landmark Analysis, Patients Surviving at Least 3, 6, and 12 Months
      RT and DosePatient Survival Time
      Patients Surviving at Least 3 mo
      Univariable AnalysisMultivariable Analysis
      HR (95% CI)p ValueHR (95% CI)p Value
      No RT
      <5040 cGy1.15 (1.10–1.21)<0.0011.14 (1.09–1.20)<0.001
      ≥5040 cGy0.81 (0.77–0.85)<0.0010.82 (0.77–0.86)<0.001
      Patients surviving at least 6 mo
      Univariable analysisMultivariable analysis
      No RT
      <5040 cGy1.11 (1.05–1.17)<0.0011.11 (1.05–1.18)<0.001
      ≥5040 cGy0.81 (0.77–0.86)<0.0010.83 (0.78–0.88)<0.001
      Patients surviving at least 12 mo
      Univariable analysisMultivariable analysis
      No RT
      <5040 cGy0.99 (0.91–1.08)0.871.03 (0.94–1.13)0.53
      ≥5040 cGy0.80 (0.74–0.87)<0.0010.85 (0.78–0.92)<0.001
      Note: Hazard ratios determined by multivariable Cox proportional hazards regression for patients surviving at least 3 or 6 months.
      HR, hazard ratio; CI, confidence interval; RT, Radiotherapy.
      To evaluate the potential effects of unmeasured confounding in our Cox regression model, we conducted a sensitivity analysis to estimate the effect of a hypothetical confounder on the observed association between high-dose PTDRT on survival (see Supplementary Table 1). In our analysis, a strong confounder (HR = 2.0) would need to be five times more prevalent in the chemotherapy-alone cohort to nullify the significance of our findings. A very strong confounder (HR = 3.0) would still have to be 2.5 times more prevalent in the chemotherapy-alone cohort to render the results no longer statistically significant.

      Discussion

      In this large national observational cohort study of 12,683 patients with metastatic esophageal cancer, we found a significant association between receipt of definitive dose radiotherapy to the primary tumor and OS. Conventional palliative dose radiotherapy, by contrast, was not associated with improved survival. The association remained statistically significant in propensity score–weighted models accounting for a large number of measured confounders, sensitivity analyses assessing the potential effects of unmeasured confounding, and landmark analyses accounting for immortal time bias. These findings may suggest that an aggressive approach combining chemotherapy with definitive doses of radiotherapy in selected patients with metastatic esophageal cancer can lead to improved survival outcomes compared with chemotherapy alone. Finally, our patterns of care analysis indicated that chemotherapy alone continued to remain the predominant treatment strategy in this population—a trend that increased over the study period.
      Our findings support the hypothesis that local control of the primary tumor may play an important role in the survival outcomes of patients with newly diagnosed metastatic esophageal cancer. Mechanistically, this approach is rational given the life-threatening risks posed by an esophageal tumor that are independent of a patient’s systemic disease burden. This concept is consistent with prior data demonstrating an association between treatment of locoregional disease and survival in metastatic esophageal cancer. For example, esophagectomy in patients with metastatic disease has been correlated with better survival outcomes in at least two single-institution retrospective studies.
      • Wang J.
      • Suri J.S.
      • Allen P.K.
      • et al.
      Factors predictive of improved outcomes with multimodality local therapy after palliative chemotherapy for stage IV esophageal cancer.
      • Blank S.
      • Lordick F.
      • Dobritz M.
      • et al.
      A reliable risk score for stage IV esophagogastric cancer.
      The addition of thoracic lymph node dissection in this context was also associated with improved survival in a prior Surveillance, Epidemiology, and End Results analysis.
      • Wu S.G.
      • He Z.Y.
      • Wang Y.
      • et al.
      Lymph node dissection improved survival in patients with metastatic thoracic esophageal cancer: an analysis of 220 patients from the SEER database.
      Additionally, a 60-patient prospective randomized phase II study in metastatic esophageal cancer suggested that radiation to the primary tumor to 5040 cGy may provide an OS benefit compared with chemotherapy alone.
      • Li T.
      • Lv J.
      • Li F.
      • et al.
      Prospective randomized phase 2 study of concurrent chemoradiation therapy (CCRT) versus chemotherapy alone in stage IV esophageal squamous cell carcinoma (ESCC).
      Population-level data from the Surveillance, Epidemiology, and End Results registry have also demonstrated an association between receipt of radiotherapy and OS in metastatic esophageal cancer.
      • Wu S.G.
      • Xie W.H.
      • Zhang Z.Q.
      • et al.
      Surgery combined with radiotherapy improved survival in metastatic esophageal cancer in a Surveillance Epidemiology and End Results population-based study.
      However, this analysis was limited by the lack of data regarding receipt of chemotherapy in the study population as well as by the limited details on radiation treatment, including dose. The current study represents the largest and most statistically rigorous comparison of survival outcomes in patients with metastatic esophageal cancer treated with aggressive primary tumor–directed therapy and it is the first to our knowledge to evaluate the impact of radiation dose in this context.
      Currently, the standard of care in metastatic esophageal cancer involves chemotherapy using multiagent platinum-containing regimens, with palliative radiotherapy as indicated for symptomatic management of dysphagia, pain, bleeding, and fistula. This approach is based on prospective trials conducted within the last 10 years and resulting in median OS times reported at around 9 months.
      • Cunningham D.
      • Starling N.
      • Rao S.
      • et al.
      Capecitabine and oxaliplatin for advanced esophagogastric cancer.
      • Al-Batran S.E.
      • Hartmann J.T.
      • Probst S.
      • et al.
      Phase III trial in metastatic gastroesophageal adenocarcinoma with fluorouracil, leucovorin plus either oxaliplatin or cisplatin: a study of the Arbeitsgemeinschaft Internistische Onkologie.
      • Van Cutsem E.
      • Moiseyenko V.M.
      • Tjulandin S.
      • et al.
      Phase III study of docetaxel and cisplatin plus fluorouracil compared with cisplatin and fluorouracil as first-line therapy for advanced gastric cancer: a report of the V325 Study Group.
      • Al-Batran S.E.
      • Pauligk C.
      • Homann N.
      • et al.
      The feasibility of triple-drug chemotherapy combination in older adult patients with oesophagogastric cancer: a randomised trial of the Arbeitsgemeinschaft Internistische Onkologie (FLOT65+).
      Our survival estimates for patients treated without radiotherapy are consistent with these outcomes (median survival of 8.3 months in the propensity score–adjusted cohort), yet they are improved by an average of 3 months in patients treated with doses of radiotherapy that are consistent with a definitive course.
      • Minsky B.D.
      • Pajak T.F.
      • Ginsberg R.J.
      • et al.
      INT 0123 (Radiation Therapy Oncology Group 94-05) phase III trial of combined-modality therapy for esophageal cancer: high-dose versus standard-dose radiation therapy.
      • Brower J.V.
      • Chen S.
      • Bassetti M.F.
      • et al.
      Radiation dose escalation in esophageal cancer revisited: a contemporary analysis of the National Cancer Data Base, 2004 to 2012.
      By contrast, in our study, lower doses of radiation were associated with worse survival than with chemotherapy alone, although the magnitude of the difference was small. An explanation for this may be evident after comparing radiation fractionation schemes between cohorts. Patients in the lower-dose cohort were more often treated in a manner consistent with pure palliation involving larger fraction sizes to a low cumulative dose—a strategy that expedites treatment and minimizes acute effects, but with less expectation of local control. Therefore, the survival outcomes in these patients overall could have been less favorable owing to disease burden, comorbidity, or some other unmeasured factor. However, we posit that definitive dose PTDRT may play a role in improving survival in well-selected patients—those whose survival is most threatened by failure to achieve local control of the primary tumor. Such patients may include those with large tumors threatening airway invasion, those with severe luminal obstruction, and those with lower systemic burden of metastatic disease whose treatment of the primary tumor might prevent seeding further metastatic spread. As with any aggressive oncologic approach, only patients with the best performance status and the most robust multidisciplinary and social support should be considered for this approach.
      Previous efforts have focused on how best to incorporate radiation therapy into the management of metastatic esophageal cancer. Most recently, the randomized TROG 03.01 phase III trial comparing radiation with concurrent chemoradiation in 220 patients with metastatic esophageal cancer demonstrated no change in dysphagia response between the two approaches, with increased toxicity in the concurrent chemoradiation arm.
      • Penniment M.G.
      • Harvey J.A.
      • Wong R.
      • et al.
      Best practice in advanced esophageal cancer: a report on Trans-Tasman Radiation Oncology Group TROG 03.01 and NCIC CTG ES.2 multinational phase 3 study in advanced esophageal cancer (OC) comparing quality of life (QOL) and palliation of dysphagia in patients treated with radiation therapy (RT) or chemoradiation therapy (CRT).
      The radiation doses used in this study were within a palliative range of 30 to 35 Gy. Although the trial was not powered to specifically assess the effect of these treatments on OS, median survival in both arms was similar (210 days versus 203 days). Similarly in our study, concurrent delivery of chemotherapy and radiation was not associated with improved survival, irrespective of RT dose, suggesting that patients need not undergo the substantial toxicity of concurrent chemoradiation to experience a survival benefit from high-dose PTDRT in clinical practice.
      Our study results suggest that providers did attempt to select patients most likely to benefit from high-dose PTDRT. For example, patients with adenocarcinoma were approximately half as likely to receive RT than were patients with the squamous histologic subtype, which may reflect a bias in perceived radiosensitivity of squamous cell carcinoma compared with adenocarcinoma.
      • van Hagen P.
      • Hulshof M.C.
      • van Lanschot J.J.
      • et al.
      Preoperative chemoradiotherapy for esophageal or junctional cancer.
      Further, patients with higher Charlson/Deyo comorbidity scores were significantly less likely to receive this more aggressive treatment, and although patients with higher T stage were more likely to be treated with high-dose PTDRT, this was not true for the most advanced (T4) cases. Additionally, the use of chemotherapy alone significantly increased over the study period. This could reflect improvements in chemotherapy with the publication of several important trials demonstrating equivalent efficacy with easier and less toxic regimens for metastatic esophageal cancer.
      • Cunningham D.
      • Starling N.
      • Rao S.
      • et al.
      Capecitabine and oxaliplatin for advanced esophagogastric cancer.
      • Al-Batran S.E.
      • Hartmann J.T.
      • Probst S.
      • et al.
      Phase III trial in metastatic gastroesophageal adenocarcinoma with fluorouracil, leucovorin plus either oxaliplatin or cisplatin: a study of the Arbeitsgemeinschaft Internistische Onkologie.
      Lastly, there was interestingly no evidence of significant disparities in use of high-dose PTDRT with respect to insurance provider, educational attainment, or income.
      The primary limitation of this study is that because of a lack of randomization, selection bias may have favored patients who received definitive dose radiotherapy. These patients may have been generally healthier with less extensive metastatic disease, and this source of bias could have affected our observed results. However, we attempted to address this concern by (1) balancing treatment groups with respect to measured confounders through multivariable regression with propensity score weighting and (2) by conducting sensitivity analyses to model the effect of unmeasured confounding factors. With regard to measured confounders, the extent of medical comorbidity (as assessed by the distribution of Charlson/Deyo comorbidity score) was not significantly different in our patient cohorts. Further, we attempted to estimate disease extent by T and N stage; however, information on these factors was missing for a significant number of patients. With regard to unmeasured confounders, such as, hypothetically, systemic burden of metastatic disease and/or overall performance status, our sensitivity analysis demonstrated that our findings are robust; to render our findings nonsignificant, such a factor would need be 2.5 to five times as prevalent in the low-dose arm and confer a twofold to threefold higher risk for death. Our conclusions are also consistent with prior literature and mechanistically plausible, strengthening the observed relationship between high-dose radiotherapy and survival.
      In conclusion, we observed an association between the use of definitive dose radiotherapy and improved OS in patients with newly diagnosed metastatic esophageal cancer treated in the United States. However, the use of PTDRT in general appears to be declining over time. The patient survival outcomes observed in our study after chemotherapy and high-dose PTDRT compare very favorably with patient survival after chemotherapy alone in prior prospective trials.
      • Cunningham D.
      • Starling N.
      • Rao S.
      • et al.
      Capecitabine and oxaliplatin for advanced esophagogastric cancer.
      • Al-Batran S.E.
      • Hartmann J.T.
      • Probst S.
      • et al.
      Phase III trial in metastatic gastroesophageal adenocarcinoma with fluorouracil, leucovorin plus either oxaliplatin or cisplatin: a study of the Arbeitsgemeinschaft Internistische Onkologie.
      • Van Cutsem E.
      • Moiseyenko V.M.
      • Tjulandin S.
      • et al.
      Phase III study of docetaxel and cisplatin plus fluorouracil compared with cisplatin and fluorouracil as first-line therapy for advanced gastric cancer: a report of the V325 Study Group.
      • Al-Batran S.E.
      • Pauligk C.
      • Homann N.
      • et al.
      The feasibility of triple-drug chemotherapy combination in older adult patients with oesophagogastric cancer: a randomised trial of the Arbeitsgemeinschaft Internistische Onkologie (FLOT65+).
      These observations support the conduct of further clinical studies integrating high-dose PTDRT as part of initial treatment in this setting.

      Acknowledgment

      The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology used or the conclusions drawn from these data by the investigator.

      Supplementary Data

      References

        • Siegel R.L.
        • Miller K.D.
        • Jemal A.
        Cancer statistics, 2016.
        CA Cancer J Clin. 2016; 66: 7-30
      1. Surveillance Epidemiology and End Results Program. Previous version: SEER cancer statistics review, 1975-2013. http://seer.cancer.gov/csr/1975_2013/. Accessed November 10, 2016.

        • Hingorani M.
        • Dixit S.
        • Johnson M.
        • et al.
        Palliative radiotherapy in the presence of well-controlled metastatic disease after initial chemotherapy may prolong survival in patients with metastatic esophageal and gastric cancer.
        Cancer Res Treat. 2015; 47: 706-717
        • Riccardi D.
        • Allen K.
        Nutritional management of patients with esophageal and esophagogastric junction cancer.
        Cancer Control. 1999; 6: 64-67
        • Saito T.
        • Kuwahara A.
        • Shigemitsu Y.
        • et al.
        Factors related to malnutrition in patients with esophageal cancer.
        Nutrition. 1991; 7: 117-221
        • Tomaselli F.
        • Maier A.
        • Sankin O.
        • Woltsche M.
        • Pinter H.
        • Smolle-Jüttner F.M.
        Successful endoscopical sealing of malignant esophageotracheal fistulae by using a covered self-expandable stenting system.
        Eur J Cardiothorac Surg. 2001; 20: 734-738
        • Maluf-Filho F.
        • Martins B.C.
        • de Lima M.S.
        • et al.
        Etiology, endoscopic management and mortality of upper gastrointestinal bleeding in patients with cancer.
        United European Gastroenterol J. 2013; 1: 60-67
        • Rusthoven C.G.
        • Jones B.L.
        • Flaig T.W.
        • et al.
        Improved survival with prostate radiation in addition to androgen deprivation therapy for men with newly diagnosed metastatic prostate cancer.
        J Clin Oncol. 2016; 34: 2835-2842
        • Seisen T.
        • Sun M.
        • Leow J.J.
        • et al.
        Efficacy of high-intensity local treatment for metastatic urothelial carcinoma of the bladder: a propensity score-weighted analysis from the National Cancer Data Base.
        J Clin Oncol. 2016; 34: 3529-3536
        • Culp S.H.
        • Schellhammer P.F.
        • Williams M.B.
        Might men diagnosed with metastatic prostate cancer benefit from definitive treatment of the primary tumor? A SEER-based study.
        Eur Urol. 2014; 65: 1058-1066
        • Ouyang W.W.
        • Su S.F.
        • Hu Y.X.
        • et al.
        Radiation dose and survival of patients with stage IV non-small cell lung cancer undergoing concurrent chemotherapy and thoracic three-dimensional radiotherapy: reanalysis of the findings of a single-center prospective study.
        BMC Cancer. 2014; 14: 491
        • Slotman B.J.
        • van Tinteren H.
        • Praag J.O.
        • et al.
        Use of thoracic radiotherapy for extensive stage small-cell lung cancer: a phase 3 randomised controlled trial.
        Lancet. 2015; 385: 36-42
        • Gomez D.R.
        • Blumenschein Jr., G.R.
        • Lee J.J.
        • et al.
        Local consolidative therapy versus maintenance therapy or observation for patients with oligometastatic non-small-cell lung cancer without progression after first-line systemic therapy: a multicentre, randomised, controlled, phase 2 study.
        Lancet Oncol. 2016; 17: 1672-1682
        • Minsky B.D.
        • Pajak T.F.
        • Ginsberg R.J.
        • et al.
        INT 0123 (Radiation Therapy Oncology Group 94-05) phase III trial of combined-modality therapy for esophageal cancer: high-dose versus standard-dose radiation therapy.
        J Clin Oncol. 2002; 20: 1167-1174
        • Brower J.V.
        • Chen S.
        • Bassetti M.F.
        • et al.
        Radiation dose escalation in esophageal cancer revisited: a contemporary analysis of the National Cancer Data Base, 2004 to 2012.
        Int J Radiat Oncol Biol Phys. 2016; 6: 985-993
        • Park H.S.
        • Gross C.P.
        • Makarov D.V.
        • Yu J.B.
        Immortal time bias: a frequently unrecognized threat to validity in the evaluation of postoperative radiotherapy.
        Int J Radiat Oncol Biol Phys. 2012; 83: 1365-1373
        • Burgette L.F.
        • McCaffrey D.F.
        • Griffin B.A.
        Propensity score estimation with boosted regression.
        in: Pan W. Bai H. Propensity Score Analysis: Fundamentals, Developments and Extensions. Guilford, New York2015: 49-73
        • McCaffrey D.F.
        • Griffin B.A.
        • Almirall D.
        • Slaughter M.E.
        • Ramchand R.
        • Burgette L.F.
        A tutorial on propensity score estimation for multiple treatments using generalized boosted models.
        Stat Med. 2013; 32: 3388-3414
      2. Ridgeway G, McCaffrey D, Griffin BA, and Burgette L. twang: toolkit for weighting and analysis of non-equivalent groups. http://cran.rproject.org/web/packages/twang/vignettes/twang.pdf. Accessed November 23, 2016.

        • Austin P.C.
        The performance of different propensity score methods for estimating marginal hazard ratios.
        Stat Med. 2013; 32: 2837-2849
        • Mitra N.
        • Heitjan D.F.
        Sensitivity of the hazard ratio to nonignorable treatment assignment in an observational study.
        Stat Med. 2007; 26: 1398-1414
        • Cunningham D.
        • Starling N.
        • Rao S.
        • et al.
        Capecitabine and oxaliplatin for advanced esophagogastric cancer.
        N Engl J Med. 2008; 358: 36-46
        • Bang Y.J.
        • Van Cutsem E.
        • Feyereislova A.
        • et al.
        Trastuzumab in combination with chemotherapy versus chemotherapy alone for treatment of HER2-positive advanced gastric or gastro-oesophageal junction cancer (ToGA): a phase 3, open-label, randomised controlled trial.
        Lancet. 2010; 376: 687-697
        • Parikh R.B.
        • Cronin A.M.
        • Kozono D.E.
        • et al.
        Definitive primary therapy in patients presenting with oligometastatic non-small cell lung cancer.
        Int J Radiat Oncol Biol Phys. 2014; 89: 880-887
        • Wang J.
        • Suri J.S.
        • Allen P.K.
        • et al.
        Factors predictive of improved outcomes with multimodality local therapy after palliative chemotherapy for stage IV esophageal cancer.
        Am J Clin Oncol. 2016; 39: 228-235
        • Blank S.
        • Lordick F.
        • Dobritz M.
        • et al.
        A reliable risk score for stage IV esophagogastric cancer.
        Eur J Surg Oncol. 2013; 39: 823-830
        • Wu S.G.
        • He Z.Y.
        • Wang Y.
        • et al.
        Lymph node dissection improved survival in patients with metastatic thoracic esophageal cancer: an analysis of 220 patients from the SEER database.
        Int J Surg. 2016; 35: 13-18
        • Li T.
        • Lv J.
        • Li F.
        • et al.
        Prospective randomized phase 2 study of concurrent chemoradiation therapy (CCRT) versus chemotherapy alone in stage IV esophageal squamous cell carcinoma (ESCC).
        Int J Radiat Oncol Biol Phys. 2016; 96: S1
        • Wu S.G.
        • Xie W.H.
        • Zhang Z.Q.
        • et al.
        Surgery combined with radiotherapy improved survival in metastatic esophageal cancer in a Surveillance Epidemiology and End Results population-based study.
        Sci Rep. 2016; 6: 28280
        • Al-Batran S.E.
        • Hartmann J.T.
        • Probst S.
        • et al.
        Phase III trial in metastatic gastroesophageal adenocarcinoma with fluorouracil, leucovorin plus either oxaliplatin or cisplatin: a study of the Arbeitsgemeinschaft Internistische Onkologie.
        J Clin Oncol. 2008; 26: 1435-1442
        • Van Cutsem E.
        • Moiseyenko V.M.
        • Tjulandin S.
        • et al.
        Phase III study of docetaxel and cisplatin plus fluorouracil compared with cisplatin and fluorouracil as first-line therapy for advanced gastric cancer: a report of the V325 Study Group.
        J Clin Oncol. 2006; 24: 4991-4997
        • Al-Batran S.E.
        • Pauligk C.
        • Homann N.
        • et al.
        The feasibility of triple-drug chemotherapy combination in older adult patients with oesophagogastric cancer: a randomised trial of the Arbeitsgemeinschaft Internistische Onkologie (FLOT65+).
        Eur J Cancer. 2013; 49: 835-842
        • Penniment M.G.
        • Harvey J.A.
        • Wong R.
        • et al.
        Best practice in advanced esophageal cancer: a report on Trans-Tasman Radiation Oncology Group TROG 03.01 and NCIC CTG ES.2 multinational phase 3 study in advanced esophageal cancer (OC) comparing quality of life (QOL) and palliation of dysphagia in patients treated with radiation therapy (RT) or chemoradiation therapy (CRT).
        Int J Radiat Oncol Biol Phys. 2014; 90: S3
        • van Hagen P.
        • Hulshof M.C.
        • van Lanschot J.J.
        • et al.
        Preoperative chemoradiotherapy for esophageal or junctional cancer.
        N Engl J Med. 2012; 366: 2074-2084