Concurrent Chemoradiotherapy Versus Radiotherapy Alone for Biopsy-Only Glioblastoma Multiforme

Transcription

1 Concurrent Chemoradiotherapy Versus Radiotherapy Alone for Biopsy-Only Glioblastoma Multiforme Adam J. Kole, MD, PhD 1 ; Henry S. Park, MD, MPH 1 ; Debra N. Yeboa, MD 1 ; Charles E. Rutter, MD 1 ; Christopher D. Corso, MD, PhD 1 ; Sanjay Aneja, MD 1 ; Nataniel H. Lester-Coll, MD 1 ; Brandon R. Mancini, MD 1 ; Jonathan P. Knisely, MD 2 ; and James B. Yu, MD, MHS 1 BACKGROUND: Combined temozolomide and radiotherapy (RT) is the standard postoperative therapy for glioblastoma multiforme (GBM). However, the clearest benefit of concurrent chemoradiotherapy (CRT) observed in clinical trials has been among patients who undergo surgical resection. Whether the improved survival with CRT extends to patients who undergo biopsy only is less certain. The authors compared overall survival (OS) in a national cohort of patients with GBM who underwent biopsy and received either RT alone or CRT during the temozolomide era. METHODS: The US National Cancer Data Base was used to identify patients with histologically confirmed, biopsy-only GBM who received either RT alone or CRT from 2006 through Demographic and clinicopathologic predictors of treatment were analyzed using the chi-square test, the t test, and multivariable logistic regression. OS was evaluated using the log-rank test, multivariable Cox proportional hazard regression, and propensity score-matched analysis. RESULTS: In total, 1479 patients with biopsy-only GBM were included, among whom 154 (10.4%) received RT alone and 1325 (89.6%) received CRT. The median age at diagnosis was 61 years. CRT was associated with a significant OS benefit compared with RT alone (median, 9.2 vs 5.6 months; hazard ratio [HR], 0.64; 95% confidence interval [CI], ; P <.001). CRT was independently associated with improved OS compared with RT alone on multivariable analysis (HR, 0.71; 95% CI, ; P <.001). A significant OS benefit for CRT persisted in a propensity score-matched analysis (HR, 0.72; 95% CI, ; P 5.009). CONCLUSIONS: The current data suggest that CRT significantly improves OS in patients with GBM who undergo biopsy only compared with RT alone and should remain the standard of care for patients who can tolerate therapy. Cancer 2016;122: VC 2016 American Cancer Society. KEYWORDS: chemoradiotherapy, glioblastoma, National Cancer Data Base (NCDB), radiotherapy. INTRODUCTION Glioblastoma multiforme (GBM) is the most common primary central nervous system malignancy in adults and carries a poor prognosis. 1,2 Currently, optimal treatment for patients with GBM consists of maximal safe surgical resection followed by concurrent chemoradiotherapy (CRT) and adjuvant chemotherapy. Approximately 15% to 20% of patients, however, do not undergo any resection after diagnostic biopsy because of poor performance status, multifocal disease, inaccessible tumor location, or potential surgical morbidity because of critical adjacent brain structures. 3-7 It is known that the extent of surgical resection has a significant impact on patient prognosis, 8-10 but it is less clear whether patients with GBM who only undergo biopsy receive a significant survival benefit from CRT over radiotherapy (RT) alone. The standard of care for patients with newly diagnosed GBM is based on a European Organization for Research and Treatment of Cancer (EORTC)-National Cancer Institute of Canada (NCIC) randomized trial in which it was demonstrated that CRT with temozolomide improved median OS over RT alone from 12.1 to 14.6 months. 7 In that study, patients were included after undergoing gross total resection, subtotal resection, or biopsy only. Accordingly, National Comprehensive Cancer Network guidelines do not vary based on the extent of surgical resection. However, the cohort of patients who underwent biopsy only reported by Stupp et al represented a small, underpowered group for which only a trend toward improved survival was observed (7.8 vs 9.4 months; statistical significance not reached). 11 Thus, definitive data to support treatment with CRT over RT alone for patients with biopsy-only GBM are arguably lacking. 12 Corresponding author: James B. Yu, MD, MHS, Smilow Cancer Hospital, 35 Park Street, Lower Level, New Haven, CT 06511; Fax: (203) ; james.b.yu@ yale.edu 1 Department of Therapeutic Radiology, Yale School of Medicine, New Haven, Connecticut; 2 Department of Radiation Medicine, Northwell Health System and Hofstra Northwell School of Medicine, Lake Success, New York The first 2 authors contributed equally to this work Additional supporting information may be found in the online version of this article. DOI: /cncr.30063, Received: January 24, 2016; Revised: March 9, 2016; Accepted: April 11, 2016, Published online May 12, 2016 in Wiley Online Library (wileyonlinelibrary.com) 2364 Cancer August 1, 2016

2 CRT Versus RT for Biopsy-Only GBM/Kole et al In this study, we used a large national database to examine the magnitude of benefit of CRT over RT alone on OS among patients who had biopsy-only GBM and did not undergo surgical resection. MATERIALS AND METHODS Study Population The National Cancer Data Base (NCDB) is a joint project of the Commission on Cancer of the American College of Surgeons and the American Cancer Society. Established in 1989, the NCDB is a comprehensive, nationwide, facility-based oncology data set that captures approximately 70% of all newly diagnosed malignancies in the United States. The data used in this study are derived from a deidentified NCDB file. 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 for the conclusions drawn, from these data by the investigators. The NCDB was used to identify patients with GBM from 2006 through This cohort was selected because patients were treated in the temozolomide era after the 2005 publication of the EORTC-NCIC trial by Stupp et al. 7 We limited our population to patients with biopsy-only GBM who had a diagnosis made by microscopic histologic confirmation and did not undergo surgical resection of the tumor. Patients without positive histology or those with ambiguous or unknown surgical status were excluded. The NCDB does not specify chemotherapeutic agent names, instead indicating whether chemotherapy was single-agent or multi-agent. Therefore, our study population was limited to patients who received single-agent chemotherapy, because this would be most consistent with the receipt of temozolomide. We excluded patients who received nonstandard treatment, including chemotherapy alone or RT to doses other than from 59.4 to 60.0 Gy in 30 to 33 fractions. Patients were included only if they completed RT within 60 days from delivery of the first RT fraction. Concurrent CRT was defined as the initiation of chemotherapy and RT within 14 days of each other. The study cohort is summarized in Figure 1. To improve the generalizability of our findings, we performed a sensitivity analysis that included patients who received nonstandard RT courses or who may not have finished standard RT as planned. The cohort used for this parallel analysis did not restrict patients based on RT dose, fractionation, or time to RT completion. Additional subgroup analyses were performed among the patients aged 70 years in both cohorts. Figure 1. Flow diagram of the current study detailing the patient cohort and the treatment groups. Statistical Methods Patient demographic and clinical factors were compared by treatment group (RT alone vs CRT) using the chisquare test and the t test for categorical and continuous variables, respectively, to identify factors that influenced the likelihood of receiving each therapy. An adapted Charlson-Deyo comorbidity score 13,14 based on the presence or absence of International Classification of Diseases- Ninth Revision-coded comorbid medical conditions was evaluated for patients in both treatment groups. The NCDB assigns a Charlson-Deyo score of 0, 1, or 2 based on the number and relative severity of reported comorbid conditions. The principal GBM diagnosis is not directly reflected in the Charlson-Deyo score. A multivariable logistic regression model was constructed using a forward stepwise approach with univariable inclusion criteria of P <.10. Kaplan-Meier analyses were used to compare OS between the RT-alone and CRT groups. Clinicopathologic factors that potentially affected survival between the RT-alone and CRT groups were dichotomized, and univariable Cox regression was used to calculate unadjusted hazard ratios (HRs) for survival. Variables trending toward significance on univariable analysis (P <.10) were Cancer August 1,

3 then included in a forward stepwise Cox multivariable regression to identify factors that were significantly associated with improved OS. The proportional hazards assumption was checked graphically using log-log survival plots. Propensity score matching was used to create quasicase/control pairs to attempt to balance measured and unmeasured confounders. 15 Matching was performed after randomly ordering patients using the psmatch2 algorithm in the Stata SE software package (StataCorp, College Station, Tex) using 1-to-1 nearest-neighbor matching without replacement. 16 Matching was performed using clinically relevant variables that were associated with differences in OS and/or the likelihood of receiving CRT versus RT on univariable analyses. Covariate balance was evaluated using standardized differences of means. 17 We then repeated survival analyses in the matched groups as described above for the unmatched cohort, adjusting for propensity quintile. 18 For this paired sample, the robust variance estimator was used for our Cox proportional hazards model. 19 RESULTS The NCDB contained a total of 41,409 patients with histologically confirmed GBM from 2006 to 2011 who fit our initial inclusion criteria. The majority of patients with biopsy-proven GBM underwent surgical resection or had unknown surgical status (n 5 32,628; 78.8%) and were excluded. The remaining 21.2% of patients (n ) did not undergo surgical resection after diagnosis of GBM and were defined as patients with biopsy-only GBM. Patients were excluded if they received supportive care alone (n ), received chemotherapy alone or had unknown chemotherapy status (n 5 447), received a nonstandard RT regimen (n ), received nonsingleagent chemotherapy (n 5 268), and if CRT was not concurrent (n 5 258). Among our final cohort of 1479 patients, 154 patients (10.4%) received RT alone, and 1325 patients (89.6%) received CRT (Fig. 1). The demographic and clinicopathologic features of our patient population are summarized in Table 1. The overall median age at diagnosis was 61 years; 58.3% of patients were men, 80.1% were white, and 76% had no significant comorbidities (Charlson-Deyo score of 0). The overall median time from diagnosis to the initiation of RT was 24 days. RT was completed in a median of 44 days in each treatment group. Limited data were available regarding Karnofsky performance status and O 6 -methylguanine-dna-methyltransferase (MGMT) methylation status, with only 4.7% and 2.3% of total patients respectively, having data recorded on those characteristics. Patients were well balanced with respect to the 2 groups (Table 1). Univariable logistic regression revealed that patients were more likely to receive RT alone if they were aged 70 years (P <.001). Patient sex, race/ethnicity, diagnosis year, median time from diagnosis to RT initiation, and Charlson-Deyo score did not differ significantly between the RT-alone and CRT groups (Table 2). Multivariable logistic regression analysis was not performed, because only 1 covariate met criteria for inclusion into the model. Among patients who received RT alone, we attempted to examine reasons why they did not receive chemotherapy. However, those data were missing for the vast majority of patients (91%). Overall, patients who received CRT had significantly improved OS compared with those who received RT alone (median OS: 9.2 vs 5.6 months; log-rank P <.0001) (Fig. 2A). Statistically significant differences in survival between treatment groups were observed as early as 6 months and persisted through 24 months after diagnosis. Long-term survival at 36 months after diagnosis was poor in both the RT-alone group and the CRT group (Table 3). Adjusting for differences in age, sex, race/ethnicity, diagnosis year, comorbidity score, and time to receipt of RT, CRT retained an independent association with improved OS compared with RT alone (adjusted HR, 0.71; 95% confidence interval [CI], ; P <.001) (Table 4). Age <70 years was significantly associated with improved survival (adjusted HR, 0.58; 95% CI, ; P <.001), and a weaker association also was identified for RT beginning >30 days after diagnosis (adjusted HR, 1.16; 95% CI, ; P 5.014) (Table 4). Subgroup analyses of patients aged >70 years also were performed. The median OS remained superior in patients who received CRT versus RT (6.1 vs 5.2 months; log-rank P <.01). Univariable and multivariable analyses confirmed our finding that CRT was independently associated with improved OS compared with RT alone among elderly patients (adjusted HR, 0.68; 95% CI, ; P 5.004) (Supporting Fig. 1, Supporting Table 1; see online supporting information). In addition, when further restricting the cohort to those aged >70 years who had a Charlson-Deyo score of 1 or 2, the improved survival persisted in the CRT group versus the RT-alone group (HR, 0.57; 95% CI, ; P 5.026; data not shown). Next, we used propensity score matching to balance treatment groups for patient age, race, Charlson-Deyo score, and time from diagnosis to start of RT. Covariates 2366 Cancer August 1, 2016

4 CRT Versus RT for Biopsy-Only GBM/Kole et al TABLE 1. Demographic Characteristics of the Patient Cohort No. of Patients (%) Characteristic RT Alone, n CRT, n Total, n P a Age, y <.001 b Median Categorized (13) 213 (16.1) 233 (15.8) (14.3) 368 (27.8) 390 (26.4) (23.4) 434 (32.8) 470 (31.8) (35.1) 230 (17.4) 284 (19.2) (14.3) 80 (6) 102 (6.9) Sex.830 Men 91 (59.1) 771 (58.2) 862 (58.3) Women 63 (40.9) 554 (41.8) 617 (41.7) Race/ethnicity.470 White 120 (77.9) 1065 (80.4) 1185 (80.1) Black 12 (7.8) 58 (4.4) 70 (4.7) Hispanic <10 54 (4.1) 59 (4) Other/unknown 17 (11) 148 (11.2) 165 (11.2) Diagnosis year (50.7) 589 (44.5) 667 (45.1) (49.4) 736 (55.6) 812 (54.9) Time from diagnosis to start of RT, d.186 Median Categorized <30 99 (64.3) 921 (69.5) 1020 (69) (35.7) 404 (30.5) 459 (31) Charlson-Deyo comorbidity score (78.6) 1003 (75.7) 1124 (76) 1 33 (21.4) 322 (24.3) 355 (24) KPS c d <60 <10 <10 <10 60 <10 61 (4.6) 63 (4.3) Unknown 150 (97.4) 1259 (95) 1409 (95.3) MGMT status c d Methylated <10 18 (1.4) 21 (1.4) Unmethylated <10 12 (0.9) 13 (0.9) Unknown 150 (97.4) 1295 (97.7) 1445 (97.7) Abbreviations: CRT, chemoradiotherapy; KPS, Karnofsky Performance Status; MGMT, O 6 -methylguanine-dna-methyltransferase; RT, radiotherapy. a The reference categories were age <70 years, men, white race, diagnosis years 2006 to 2008, time from diagnosis to RT <30 days, and a Charlson-Deyo score of 0. b This P value indicates a statistically significant difference. c To prevent patient identifiability, the National Cancer Data Base does not allow reporting of any cell sizes <10 patients. d A chi-square test was not performed because of the small sample size. were well balanced on standardized differences of means (Supporting Fig. 2; see online supporting information). The median OS was improved in patients who received CRT versus RT alone (median OS, 8.4 vs 5.7 months; HR, 0.72; 95% CI, ; P 5.009) (Fig. 2B). Finally, we completed a sensitivity analysis in a less selective cohort of patients that was not restricted based on RT dose, fractionation, or time to RT completion (Supporting Fig. 3; see online supporting information). Among the 4033 patients who were included in this parallel analysis, CRT retained its association with improved OS compared with RT alone (median OS, 6.9 vs 3.5 months; log-rank P <.001) (Supporting Fig. 4; see online TABLE 2. Univariable Logistic Regression to Determine Odds of Receiving Chemoradiotherapy Versus Radiotherapy Alone Variable OR (95% CI) P Age:<70 vs 70 y ( ) <.001 a Sex: Men vs women ( ).830 Race/ethnicity: White vs nonwhite ( ).470 Diagnosis year: ( ).144 vs Time from diagnosis to start ( ).186 of RT: <30 vs 30 d Charlson-Deyo Score: 0 vs ( ).430 Abbreviations: CI, confidence interval; OR, odds ratio; RT, radiotherapy. a This P value indicates a statistically significant difference. Cancer August 1,

5 TABLE 3. Overall Survival Over Time by Treatment Group Variable RT Alone, n CRT, n Median OS (95% CI), mo 5.6 ( ) 9.2 ( ) OS (%) at: 3 mo 90.2 ( ) 93.8 ( ) 6 mo 43.6 ( ) 68.1 ( ) 12 mo 22.8 ( ) 37.8 ( ) 18 mo 10.1 ( ) 22.2 ( ) 24 mo 6.7 ( ) 13.7 ( ) 36 mo 5.2 ( ) 7.3 ( ) Abbreviations: CI, confidence interval; CRT, chemoradiotherapy; OR, odds ratio; OS, overall survival; RT, radiotherapy. Figure 2. (A) Overall survival is illustrated for patients with unresected glioblastoma multiforme who received either radiotherapy (RT) alone or chemoradiotherapy (CRT). (B) Overall survival is illustrated between the treatment groups for the propensity-matched analysis cohort. supporting information). Compared with RT alone, CRT was also independently associated with improved OS on multivariable analysis in our parallel cohort (adjusted HR, 0.58; 95% CI, ; P <.001) and in the subgroup of patients aged >70 years in our parallel cohort (adjusted HR 0.62; 95% CI, ; P <.001). DISCUSSION CRT is well established as the current standard of care after surgical resection of GBM. However, patients who undergo biopsy only comprise an understudied and often underpowered group for which there are limited definitive data to support treatment recommendations. Here, using a large national database, we observed a significant survival benefit for CRT over RT alone for patients with biopsy-only GBM. Increasing evidence suggests that the extent of surgical resection after a diagnosis of GBM is among the most important prognostic factors, along with age, performance status, and MGMT methylation status. 8-10,20-22 However, very little is known about the optimal therapeutic approaches for patients with biopsy-only GBM. In the seminal EORTC-NCIC study, a small subgroup (93 of 573 total patients) underwent biopsy only and were randomized to receive either CRT with temozolomide or RT alone. In these patients, a nonstatistically significant trend toward improved OS was observed for CRT. 11 Other single-institution analyses have proposed a possible improvement in survival with CRT, but those studies were been small, nonrandomized, and predated the temozolomide era. 3,23 In our large national study, we demonstrated that CRT was independently associated with significantly improved OS compared with RT alone, with median survival of 9.2 and 5.6 months for CT versus RT alone, respectively. Although the overall prognosis for patients who receive with either RT alone or CRT remains very poor, the observed 3.6-month absolute prolongation of survival represents a substantial relative increase in survival after diagnosis. It is well known that elderly patients have worse OS than the general population of patients with GBM. Consequently, alternative treatment regimens have been proposed for this subgroup of patients However, we observed that CRT remained superior to RT alone when adjusting for age differences on multivariable analysis. In addition, subgroup analysis of patients aged >70 years demonstrated improved survival in those who received CRT compared with RT alone. Although we did observed that CRT was associated with a similar relative survival benefit over RT alone among elderly patients who 2368 Cancer August 1, 2016

6 CRT Versus RT for Biopsy-Only GBM/Kole et al TABLE 4. Cox Proportional Hazards Regression Demonstrating Predictors of Overall Survival Univariable Analysis Multivariable Analysis Variable HR (95% CI) P AHR (95% CI) P Treatment: CRT vs RT alone ( ) < ( ) <.001 a Age:<70 vs 70 y ( ) < ( ) <.001 a Sex: Men vs women ( ).265 Race/ethnicity: White vs nonwhite ( ) ( ).083 Diagnosis year: vs ( ).131 Time from diagnosis to start of RT: <30 vs 30d ( ) ( ).014 a Charlson-Deyo score: 0 vs ( ) ( ).067 Abbreviations: AHR, adjusted hazard ratio; CI, confidence interval; CRT, chemoradiotherapy; HR, hazard ratio; RT, radiotherapy. a This P value indicates a statistically significant difference. underwent biopsy only, the absolute benefit of CRT for these patients was less than 1 month, potentially representing a less clinically meaningful benefit. Our study has several limitations. First, the NCDB does not provide data on salvage therapies after the primary treatment course. Whether the CRT and RT-alone groups were balanced with respect to subsequent therapies (eg, bevacizumab, reirradiation) is unknown. Second, given the retrospective nature of this study, there is a strong potential that selection bias could play a significant role in our findings. Indeed, we note that patients who are elderly are more likely to receive RT alone, and this contributes in part to the worse OS observed in that treatment group. Although we adjusted for this known confounder, the possibility remains that the patients who received RT alone represent a less fit patient population and were selected for this regimen because of other unknown or unmeasured confounders. For example, the NCDB lacks robust data on performance status and MGMT methylation status in the vast majority of patients. It has been demonstrated that these are important prognostic factors for patients with GBM, 20,28,29 and it is unknown whether the 2 treatment groups were well balanced for Karnofsky performance status and MGMT status. Although our data indicate that survival was improved among a national sample of patients who underwent biopsy only and received CRT over RT alone, prospective randomized studies are necessary to rule out selection bias. Finally, in our initial analysis, we excluded patients who did not complete a full course of conventionally fractionated RT to limit the impact of immortal time bias. 30 The subpopulation of patients who survive at least through the duration of therapy may not be representative of all patients who do not undergo surgical resection. In fact, studies have suggested that as few as 50% of patients with biopsy-only GBM receive and complete RT. 3 In a sensitivity analysis, we included a large group of patients who received CRT or RT alone without restricting our cohort to those who completed a standard course of therapy. Although we observed that CRT retained its association with improved survival in this parallel cohort, it is paramount that clinical judgment is used to determine whether each individual patient is likely to tolerate CRT or whether intensive therapy could be detrimental to the patient. Patients with biopsy-only GBM are relatively common in clinical settings. Within the NCDB, since 2006, approximately 20% of patients with newly diagnosed GBM did not undergo surgical resection, consistent with large historic clinical trials. 4-7 Although survival remains particularly poor for this challenging subgroup, our findings support CRT as the continued standard of care for patients with biopsy-only GBM. FUNDING SUPPORT No specific funding was disclosed. CONFLICT OF INTEREST DISCLOSURES Debra N. Yeboa reports travel expenses from Eli Lilly outside the submitted work. Charles E. Rutter, Christopher D. Corso, and Nataniel H. Lester-Coll report personal fees from Elekta AB outside the submitted work. Jonathan P. Knisely reports personal fees from Elekta AB, BrainLab AG, and Cyber Medical Corporation outside the submitted work. James B. Yu reports institutional research funding from 21st Century Oncology outside the submitted work. AUTHOR CONTRIBUTIONS Adam J. Kole, Henry S. Park, and James B. Yu conceptualized the project. Debra N. Yeboa obtained the data used for the study. All authors contributed to data analysis and editing of the article. REFERENCES 1. Ostrom QT, Bauchet L, Davis FG, et al. The epidemiology of glioma in adults: a state of the science review. Neuro Oncol. 2014;16: Cancer August 1,

7 2. Woehrer A, Bauchet L, Barnholtz-Sloan JS. Glioblastoma survival: has it improved? Evidence from population-based studies. Curr Opin Neurol. 2014;27: Balana C, Capellades J, Teixidor P, et al. Clinical course of highgrade glioma patients with a biopsy-only surgical approach: a need for individualised treatment. Clin Transl Oncol. 2007;9: Werner-Wasik M, Scott CB, Nelson DF, et al. Final report of a phase I/II trial of hyperfractionated and accelerated hyperfractionated radiation therapy with carmustine for adults with supratentorial malignant gliomas. Radiation Therapy Oncology Group Study Cancer. 1996;77: Blumenthal DT, Won M, Mehta MP, et al. Short delay in initiation of radiotherapy may not affect outcome of patients with glioblastoma: a secondary analysis from the Radiation Therapy Oncology Group database. J Clin Oncol. 2009;27: Nieder C, Grosu AL, Astner S, Molls M. Treatment of unresectable glioblastoma multiforme. Anticancer Res. 2005;25: Stupp R, Mason WP, van den Bent MJ, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 2005;352: Gorlia T, van den Bent MJ, Hegi ME, et al. Nomograms for predicting survival of patients with newly diagnosed glioblastoma: prognostic factor analysis of EORTC and NCIC trial / CE.3. Lancet Oncol. 2008;9: Kreth FW, Thon N, Simon M, et al. Gross total but not incomplete resection of glioblastoma prolongs survival in the era of radiochemotherapy. Ann Oncol. 2013;24: Sanai N, Polley MY, McDermott MW, Parsa AT, Berger MS. An extent of resection threshold for newly diagnosed glioblastomas. J Neurosurg. 2011;115: Stupp R, Hegi ME, Mason WP, et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol. 2009;10: Knisely JP, Baehring JM. A silver lining on the horizon for glioblastoma. Lancet Oncol. 2009;10: Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40: Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45: Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70: Wang EH, Rutter CE, Corso CD, et al. Patients selected for definitive concurrent chemoradiation at high-volume facilities achieve improved survival in stage III non-small-cell lung cancer. J Thorac Oncol. 2015;10: Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28: D Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17: Lin DY, Wei LJ. The robust inference for the cox proportional hazards model. J Am Stat Assoc. 1989;84: Hegi ME, Diserens AC, Gorlia T, et al. MGMT gene silencing and benefit from temozolomide in glioblastoma. N Engl J Med. 2005; 352: Mirimanoff RO, Gorlia T, Mason W, et al. Radiotherapy and temozolomide for newly diagnosed glioblastoma: recursive partitioning analysis of the EORTC 26981/22981-NCIC CE3 phase III randomized trial. J Clin Oncol. 2006;24: Simpson JR, Horton J, Scott C, et al. Influence of location and extent of surgical resection on survival of patients with glioblastoma multiforme: results of 3 consecutive Radiation Therapy Oncology Group (RTOG) clinical trials. Int J Radiat Oncol Biol Phys. 1993; 26: Fazeny-Dorner B, Wenzel C, Veitl M, et al. Survival and prognostic factors of patients with unresectable glioblastoma multiforme. Anticancer Drugs. 2003;14: Roa W, Brasher PM, Bauman G, et al. Abbreviated course of radiation therapy in older patients with glioblastoma multiforme: a prospective randomized clinical trial. J Clin Oncol. 2004;22: Keime-Guibert F, Chinot O, Taillandier L, et al. Radiotherapy for glioblastoma in the elderly. N Engl J Med. 2007;356: Malmstrom A, Gronberg BH, Marosi C, et al. Temozolomide versus standard 6-week radiotherapy versus hypofractionated radiotherapy in patients older than 60 years with glioblastoma: the Nordic randomised, phase 3 trial. Lancet Oncol. 2012;13: Roa W, Kepka L, Kumar N, et al. International Atomic Energy Agency randomized phase III study of radiation therapy in elderly and/or frail patients with newly diagnosed glioblastoma multiforme [published online ahead of print September 21, 2015]. J Clin Oncol. doi: /JCO Lamborn KR, Chang SM, Prados MD. Prognostic factors for survival of patients with glioblastoma: recursive partitioning analysis. Neuro Oncol. 2004;6: Thon N, Eigenbrod S, Grasbon-Frodl EM, et al. Predominant influence of MGMT methylation in non-resectable glioblastoma after radiotherapy plus temozolomide. J Neurol Neurosurg Psychiatry. 2011;82: Park HS, Gross CP, Makarov DV, Yu JB. Immortal time bias: a frequently unrecognized threat to validity in the evaluation of postoperative radiotherapy. Int J Radiat Oncol Biol Phys. 2012;83: Cancer August 1, 2016