The impact of radiation therapy variables on pediatric high-grade glioma outcomes: A National Cancer Database analysis

Purpose: Thepurposeofthisanalysisistoreportpatternsofcareforpediatricpatients with high-grade glioma (pHGG) and evaluate the impact of radiotherapy (RT) variables on outcomes using the National Cancer Database (NCDB). Methods: Eligibility criteria included age < 22 years, histologically diagnosed WHO grade III-IV gliomas treated with ≥ 50 Gy and < 76 Gy RT between 2004 and 2013, and RT initiation within 90 days of diagnosis. RT variables including RT dose, RT timing, andRTmodalitywereanalyzedalongwithbaselinedemographic,tumor,andtreatment variables to assess the impact on overall survival. Results: A total of 498 pHGG patients were included. The median age was 15 years (range, 0-21), common diagnoses were astrocytoma (55%) and glioblastoma (30%), 73.5% underwent surgical resection and 90.2% received chemotherapy. The median RT dose was 59.4 Gy (SD 2.9 Gy) starting at a median of 4.4 weeks from diagnosis (SD 2.5 weeks). Fourteen patients were treated with proton therapy. Median follow-up was 19.6 months with 1-and 3-year overall survival of 78.4% and 40.4%, respectively. On multivariable analysis, female gender, older age, and RT delay of ≥ 6 weeks were significantly associated with a lower rate of death; glioblastoma histology, no surgical resection/biopsy only, and earlier RT initiation < 6 weeks from diagnosis were associated with a higher rate of death. There was no relationship between RT dose or proton versus photon therapy and overall survival. Conclusions: Outcomes for pHGG are poor. There was no benefit to early RT timing when RT is initiated within 90 days of diagnosis or higher RT dose in this dataset.


INTRODUCTION
In the United States, brain tumors are the most common pediatric solid tumor and the leading cause of cancer-related death in children less than 14 years old. 1 Pediatric high-grade gliomas (pHGG), including most commonly anaplastic astrocytoma, diffuse astrocytoma, and glioblastoma are rare, aggressive glial cell tumors that account for less than 20% of brain tumors in the pediatric population. 2 These tumor types are classified by the World Health Organization as grade III or IV, signifying them as highly malignant with tumor characteristics of hypercellularity, nuclear atypia, and highly metabolic activity. 3diatric HGGs are biologically distinct from adult glioblastoma yet share the same poor prognosis.In adult patients, maximal safe surgical resection followed by radiotherapy (RT) with concurrent and adjuvant temozolomide remains the current standard of care. 4In pHGG, no clear standard exists, though a similar management strategy is often employed. 5An initial prospective, randomized study dating back to 1989 showed an increased five-year progression-free survival and overall survival for patients receiving RT with the addition of chemotherapy compared with RT alone. 6Subsequent studies have further investigated a variety of post-surgical chemotherapy regimens in combination with RT without a clear benefit, and ongoing protocols continue to test novel systemic agents in molecular-defined cohorts of pHGG. 2 Despite the varying systemic therapies used, RT has remained a standard of care in the definitive management of pHGG, though data addressing the impact of varying RT variables such as dose, technique, and timing are lacking.The purpose of this analysis is to evaluate the impact of RT variables on outcomes of pediatric patients with HGG treated with definitive RT in the National Cancer Database.

Patient selection
The National Cancer Database (NCDB) is a clinical oncology database supported by the American College of Surgeons and the American Cancer Society.

Independent variables
Radiation treatment parameters used as independent variables for outcomes analysis included radiation timing in weeks from diagnosis until RT initiation, radiation dose, and radiation modality.Radiation modality was defined as either proton therapy, IMRT photon RT, 3D conformal photon RT, or photon RT not otherwise specified (NOS).
Radiation timing was categorized as < 4 weeks, between 4 and 6 weeks, and greater than 6 weeks from diagnosis.The cumulative radiation dose including the sum of any dose described as "regional" or "boost" was analyzed as a continuous variable and defined as escalated (> 59.4 Gy) versus nonescalated (≤ 59.4 Gy).
Additional treatment variables, including surgical resection, receipt of chemotherapy, whether chemotherapy was single agent or multiagent, and chemotherapy sequence with radiation therapy, were also evaluated.Surgical resection was classified as yes (surgical resection was performed/unknown extent of resection) or no (biopsy only) as information regarding the extent of surgical resection was missing from approximately half of the cases.All demographic and tumor characteristics listed in Tables 2 and 3 were also evaluated as independent variables for univariable and multivariable analyses.

Patient population
A total of 498 pHGG patients were included in this analysis.Demographic and tumor characteristics and treatment variables are shown in Tables 2 and 3, respectively.The median age at diagnosis was 15 years (range, 0-21), and 54% of patients were male.The majority of patients were white (79%) and living in a metro area (80%).The most common histologies were astrocytoma (55%) and glioblastoma (30%).Surgical resection was performed in 73.5% of patients and the remaining 26.5% of patients underwent biopsy only for the following reasons: surgery not performed because not part of the planned first-course treatment (23.7%), surgery not recommended/performed due to contraindications (2.4%), and surgery recommended but not performed with no reason noted in the database (0.4%).Chemotherapy was received by 90.2% of patients.Chemotherapy was initiated either before or during radiation in 80.3% of patients, and after RT was completed in 4.6% of patients.The median RT dose was 59.4 Gy (SD 2.9 Gy) starting at a median of 4.4 weeks from diagnosis (SD 2.5 weeks).

Overall survival
The median follow-up was 19.S1 and S2, respectively.There was no association on either univariate or multivariable analysis between overall survival and Hispanic origin, urban versus rural residence, median income, year of diagnosis, tumor grade, size of tumor, or whether chemotherapy was received.

Impact of RT parameters
As previously described, patients treated with earlier RT timing of < 6 weeks from diagnosis demonstrated a significantly increased risk of death on both univariate and multivariate analyses (Table 4; Figure 1B).To evaluate whether this finding could be attributed to patients not undergoing surgical resection receiving earlier RT, the impact of RT timing was also evaluated in the subgroup of patients who underwent surgical resection only.There was no statistically significant association between radiation dose and overall survival on either univariate or multivariate analysis, when analyzed as a continuous variable (HR 0.97 (0.93-1.00),P = .053on univariate analysis) or as a dichotomous variable RT dose ≥ 59.4 Gy versus RT dose < 59.4 Gy (HR 0.80 (0.62-1.04),P = .093on univariate analysis, HR 0.99 (0.74-1.32),P = .942on multivariate analysis).
Patients who received escalated dose RT were more likely to have glioblastoma histology (P = .004),be of greater age (P < .001),and less likely to have a Charles-Deyo score of ≥2 (P = .012)(results not shown in tables).There was also no association between overall survival and RT modality (Protons vs IMRT vs 3D conformal therapy or external beam/photons NOS).

DISCUSSION
Pediatric HGG are a heterogeneous group of patients that comprise a small fraction of pediatric cancer but account for a majority of pediatric cancer deaths under the age of 20 years old. 1 Over the past 50 years, multiple treatment options have been investigated including gross

F I G U R E 1 Kaplan-Meier analysis of overall survival according to tumor histology (A) and timing of RT initiation from diagnosis (B).
TA B L E 4 Univariate and multivariable analyses with overall survival.tumor resection, radiation therapy, and biological and chemotherapy agents.However, the overall prognosis for pHGG patients remains poor with outcomes highly dependent on the tumor histology and patients with glioblastoma demonstrating the worst overall survival.

Covariate
In this study, most pHGG patients received trimodality therapy, including surgical resection, chemotherapy, and radiation therapy, in accordance with current standard practice.Tumor resection was performed in 73.5% of patients and was associated with improved overall survival on multivariable analysis.The Children's Cancer Study Group 945 study has previously demonstrated the importance of surgical resection.Patients with greater than 90% gross tumor resection had improved five-year progression-free survival of 35% compared with only 17% for those with less than 90% tumor resection. 5,8Additionally, complete resection remains difficult for pHGG due to the tumor volumes being ill-defined, irregularly shaped, invasive, or in situations in which the surgeon cannot complete a safe resection (e.g., midline tumors). 9,10Although the occurrence of surgical resection was an important prognostic variable in this analysis, the extent of the resection was not available for the majority of patients in this dataset and so was not analyzed here, therefore presenting a limitation to the analysis.
RT is an essential part of the treatment for pHGG.Between 1966 and 1975, multiple adult Brain Tumor Study Group protocols evaluated the efficacy of RT.Patients who received no RT were compared with groups receiving escalating doses ranging from 45 to 60 Gy.
Patients who received no RT had a median survival of 18.0 weeks, while those who received 50 to 60 Gy showed a median survival of 28.0 and 42.0 weeks, respectively. 11Though different dose fractionations have been investigated, the use of hyperfractionation or accelerated fractionation RT dose delivery has not improved outcomes in either adult gliomas or pediatric brainstem gliomas, 12 and therefore, the conventional dose of 54−60 Gy total dose in 1.8-2.0Gy per day for approximately 6 weeks remains the current standard of care for pHGG.
More recently, RT dose escalation using advanced IMRT or proton therapy techniques to deliver a simultaneous integrated boost of 75 Gy has been studied in adult glioblastoma patients with favorable outcomes demonstrated in small single-institution studies. 13,14This concept is currently being compared with conventional RT doses in a randomized fashion for adult glioblastoma patients receiving concurrent and adjuvant temozolamide in the NRG Oncology BN001 trial (NCT02179086).We sought to evaluate whether an RT dose response could be detected among pHGG patients in this analysis.Patients in this study receive a median RT dose of 59.4 Gy.Though a trend with improved overall survival and higher radiation dose was seen in univariate analysis, the results were not statistically significant, and no association was found in multivariable analysis.Further, more detailed analyses would be required to evaluate the potential impact of RT dose escalation on tumor control and patterns of failure in pHGG, which is not feasible in this population-based dataset.
The results of this study demonstrate that initiation of RT earlier than 6 weeks after diagnosis is associated with a higher rate of death, a finding that was true among all patients and when limited to the subcohort of patients who underwent surgical resection.It should be noted that all patients received RT within 90 days of diagnosis, as initiation of treatment later than that was considered an exclusion criterion.This is a novel finding that, to our knowledge, has not been previously identified among pHGG, though recent studies in adult glioblastoma using the NCDB found similar results.The aforementioned NCDB analysis showed patients with adult glioblastoma who started chemoradiation within 0-1 weeks (HR 1.18), < 1-2 weeks (HR 1.23), and > 2-3 weeks (HR 1.11) from surgery demonstrated worse overall survival compared with initiating treatment after week 5 from surgery. 15This NCDB study, like ours, is limited in its retrospective design and the finding could be attributed to selection bias with patients demonstrating more aggressive histologies, more concerning imaging findings, or more aggressive clinical course initiating RT earlier.
We attempted to minimize potential bias by evaluating the impact of RT timing among the subgroup of patients receiving surgical resection only, and the same results were found.However, the extent of surgical resection is unknown, and patients with subtotal resection and a higher burden of disease remaining after surgery may have poorer outcomes and may also be more likely to receive earlier initiation of RT.Additional prognostic variables, such as tumor molecular characteristics, are unknown and may have impacted the study results in unknown ways.We found that patients initiating RT later were more often Black, less often white, and more likely to initiate chemotherapy neoadjuvant or adjuvant to RT versus during RT.It is unlikely that the lower percentage of white patients contributed to better overall survival outcomes with delayed RT initiation, as white race was associated with improved overall survival on multivariable analysis.Similarly, neither the use of chemotherapy nor the sequence of chemotherapy with RT was associated with overall survival or univariate or multivariate analysis.Other variables such as duration of RT treatment have been correlated with worse clinical outcomes in other pediatric brain tumors 16 and were not analyzed here.It is unknown if this variable is associated with poor overall survival in pHGG or whether there was any difference in RT duration among patients with earlier vs later RT initiation.Although there is no clear clinical hypothesis to explain why RT initiation at > 6 weeks from diagnosis would lead to improved outcomes as compared with 4-6 weeks from diagnosis in and of itself, it is reasonable to conclude that RT timing of > 6 weeks may not be detrimental for some pHGG patients.
No significant difference in outcomes was seen between patients treated with photons versus proton therapy, though these findings should be interpreted with caution as the proton sample size was limited with 2.8% of patients in this cohort receiving proton therapy.Proton therapy is becoming increasingly available and utilized for pediatric brain tumor patients in the United States.Outcome data for pHGG patients treated with proton therapy are limited, and further analyses are warranted.

CONCLUSIONS
Outcomes for pHGG are poor, and novel treatment approaches are needed to improve outcomes.Glioblastoma histology and absence of surgical resection were significantly associated with an increased risk The database includes hospital registry data aggre- gliomas (astrocytomas, glioblastomas, astroblastomas, and oligodendrogliomas) diagnosed between 2004 and 2013, receipt of RT within 90 days of diagnosis, and RT doses of ≥50 Gy and < 76 Gy.RT doses of ≥50 Gy and < 76 Gy were set as dose-range limits to exclude patients who were thought not to have likely received full treatment (< 50 Gy) and doses that were likely listed in error (≥76 Gy).This identified 208,559 patients, and 208,061 patients were excluded for the reasons detailed in Table1.The final cohort included 498 patients.

TA B L E 1 Patient selection and exclusion criteria. Selection and exclusion criteria Sample size Excluded
Abbreviations: NCDB, National Cancer Database; PUF, participant user file; RT, radiotherapy.
Patient demographic and tumor characteristics.Treatment characteristics.