Optimal Timing of Radiotherapy Following Gross Total or Subtotal Resection of Glioblastoma: A Real-World Assessment using the National Cancer Database

Treatment for glioblastoma (GBM) includes surgical resection and adjuvant radiotherapy (RT) and chemotherapy. The optimal time interval between surgery and RT remains unclear. The National Cancer Database (NCDB) was queried for patients with GBM. Overall survival (OS) was estimated using Kaplan-Meier and log-rank tests. Univariate (UVA) and multivariable Cox regression (MVA) modeling was used to determine predictors of OS. A total of 45,942 patients were included. On MVA: younger age, female gender, black ethnicity, higher KPS, obtaining a gross total resection (GTR), MGMT promoter-methylated gene status, unifocal disease, higher RT dose, and RT delay of 4–8 weeks had improved OS. Patients who underwent a subtotal resection (STR) had worsened survival with RT delay ≤4 weeks and patients with GTR had worsened survival when RT was delayed >8 weeks. This analysis suggests that an interval of 4–8 weeks between resection and RT results in better survival. Delays >8 weeks in patients with a GTR and delays <4 weeks in patients with a STR/biopsy resulted in worse survival. This impact of time delay from surgery to RT, in conjunction with extent of resection, should be considered in the clinical management of patients and future designs of clinical trials.

the American College of Surgeons and the American Cancer Society. It is a hospital-based registry that captures approximately 70% of newly-diagnosed cancer cases in the United States and Puerto Rico and draws data from >1500 commission-accredited cancer programs. This program originated in 1989 and now contains approximately 34 million records. Data registries contain patient characteristics, cancer staging and tumor histological characteristics, type of first course treatment administered, and outcomes. The American College of Surgeons and the CoC have not verified and are not responsible for the analytic or statistical methodology used, or for the conclusions drawn from these data by the investigator. This NCDB analysis was approved by the institutional review board at MD Anderson Cancer Center and all methods were performed in accordance with the relevant guidelines and regulations. Additionally, a waiver of informed consent was obtained as the information in the Commission on Cancer's NCDB is de-identified.
We queried the NCDB User File for adult patients with primary glioblastoma treated between 2004-2015 using the International Classification of Diseases for Oncology histology codes 9440 and site-specific codes C720-3. Summary of cohort selection is detailed in Fig. 1. We only included patients with World Health Organization (WHO) Grade IV disease, and histologic confirmation of glioblastoma, NOS, giant cell glioblastoma, and gliofibroma. We excluded patients with gliosarcoma or choroid glioma, as well as other non-GBM histologies. We calculated the time between surgical resection and initiation of radiation as the subtraction between the variables: 'Time from diagnosis to radiation' and 'Time from diagnosis to surgical resection' . We excluded all negative and zero day time points to ensure that radiotherapy was performed following surgery, as well as time intervals greater than 6 months as these patients were likely treated with radiotherapy for progressive disease. Time interval between surgery and the start of RT were grouped into ≤4 weeks, 4.1-6 weeks, 6.1-8 weeks, and >8 weeks. These time intervals were selected for their prevalence in the published literature, and for their ease of clinical application. Resection status analysis was performed on patients with a reported gross total resection (GTR) or subtotal resection (STR)/biopsy, only.

Key covariates.
Covariates incorporated into our analysis were selected a priori and included age, KPS, gender, ethnicity, surgical resection type, MGMT promoter-methylation status, unifocal/multifocal, total radiation dose, survival status, and follow-up time.
Statistical analyses. Statistical analyses were performed using Stata/MP statistical software (version 15.1; Stata, College Station, TX). Univariate (UVA) and multivariable (MVA) modeling with Cox regression analysis was used to determine predictors of overall survival. Statistically significant predictor variables on UVA were included in MVA models. Overall survival was estimated using Kaplan-Meier and log-rank test methods. Statistical tests were based on a 2-sided significance level, and a P value ≤ 0.05 was considered to be statistically significant.

Results population characteristics.
A total of 45,942 patients met inclusion criteria and were included in the analysis. The mean age at diagnosis was 61 years (range 18-90 years). The majority of patients were male (59%), of white ethnicity (91%), and had an unknown Karnofsky Performance Status (KPS; 55%). Only 7% of patients had a reported KPS > 70 in this patient cohort.
tumor and treatment characteristics. A total of 11,470 patients underwent a gross total resection (GTR) and 13,594 underwent a subtotal resection or biopsy (STR). The median time interval from resection to RT was 29 days (range 1-179 days). Almost half of the cohort (47%) were initiated on RT at <4 weeks from surgical resection, 38% had a delay of 4.1-6 weeks from surgery, 10% had a delay of 6.1-8 weeks, and 5% had a delay of over 8 weeks from resection before beginning RT. Cohort characteristics by RT delay groupings are summarized in Table 1.

Discussion
The time delay from surgical resection to the start of adjuvant RT for patients with newly diagnosed GBM is a clinically adjustable factor that can be acted upon in clinical practice and in clinical trial design. The impact of this time interval between surgery and the start of RT has been studied in a number of retrospective series and population analyses (Table 3, Fig. 2). These studies have included a few hundred to a few thousand patients treated between 1974 and 2015 with surgical resection, followed by adjuvant radiation, with or without adjuvant Temozolomide (TMZ . Prior analyses included patients with a median RT delay ranging from 12 to 47 days and the median reported OS from these reviews ranged from 7.4 to 26 months. A total of six studies: four retrospective studies, one prospective, and one focused analysis of the NCDB only included patients treated with adjuvant radiation and TMZ 14,21,24,25,27,28 .
Our large retrospective analysis of patients with newly diagnosed GBM from the National Cancer Database found that patients who have at least a 4 week interval between resection and the start of adjuvant RT have improved OS, consistent with the majority of prior published series (Tables 2-3  Additionally, five studies found that a delay of less than 5-6 weeks improves survival; but they did not report significance at the 4 week cut-off point. Glinski  In contrast to these studies favoring an interval of several weeks between surgery and subsequent RT, two studies published in the early 2000's reported worsened OS with any delay in RT. A retrospective analysis of 182 patients treated from 1979-1995 with an average interval between surgery and RT of 26 days reported no significant difference in outcomes associated with delay of RT following surgery or biopsy, but they found that each additional day of RT delay from initial presentation increased the risk of death by 2% 8 . In this study, Do et al. excluded 31 patients who had small tumors, complete resections, or were treated with radiosurgery, as well as those that received adjuvant chemotherapy. The exclusion of patients with these potentially favorable prognostic factors such as smaller tumor size and complete resections as well as adjuvant chemotherapy likely influenced the findings of this study. Irwin et al. published a retrospective review of 172 patients treated from 1993-2003 with an average RT delay of 35 days. In this manuscript, the authors state that each additional week of RT delay increases the risk of death by 8.9%, but this calculation was made using estimated survival curves for a hypothetical population that was derived from a proportional hazards model and not from the study population evaluated in the analysis 9 . In our study, further analysis revealed an association between the extent of surgical resection and the impact of time between surgery and the start of RT on outcome. We found that delay of RT > 8 weeks after GTR or <4 weeks after STR was associated with worse OS, and the best survival outcomes were seen when radiotherapy was initiated 4-8 weeks following surgical resection. Specifically, in patients achieving a GTR, those initiated on www.nature.com/scientificreports www.nature.com/scientificreports/ RT ≤ 8 weeks from surgery had a median OS of 16.9 months, and those initiated >8 weeks had a median OS of 15.2 months (a difference of 1.7 months). In patients in whom a STR was performed, initiating RT ≤ 4 weeks from surgery resulted in a median OS of 12.9 months; whereas, starting RT > 4 weeks from surgery had a median OS of 13.8 months (a difference of 0.9 months). Prior studies that reported improved survival with an RT delay of at least 4 weeks may have had a substantial proportion of patients with a STR based on our analyses.
The mechanism of improved survival with a 4-8 week RT delay following surgery is unknown. Several hypotheses have been suggested in the literature, although no data is available as evidence of the pathophysiology. We hypothesize that some possible mechanisms might include a bias of clinicians to expedite treatment in patients who undergo a STR/biopsy or have more aggressive disease courses, rushing them to start treatment sooner. This may have contributed to the worse survival seen in patients with shorter delay between surgery and RT. In our study population, there were similar proportions of resection types within each time interval from surgery to RT. Prior studies that reported shorter survival with an interval of less than 4 weeks between surgery and RT also suggested increased complications as a result of this short time delay from surgery and the start of RT. Additionally, we have seen in our practice that when RT is initiated before 4 weeks, contrast enhancement on MRI or CT is difficult to interpret in the immediate post-operative setting. Additionally, the surgical cavity can undergo more dramatic changes within the first 4 weeks following surgical resection such that the radiation plan  generated based on early post-operative imaging may not fully encompass the extent of disease throughout the course of RT [31][32][33] . For these reasons, allowing a delay following surgery before acquiring the RT planning imaging may facilitate better definition of the target volume and more stability of the target volume through the course of RT. Ongoing studies of serial imaging during radiotherapy may identify the impact of surgical cavity changes on RT coverage and resulting survival outcomes. Finally, one possible hypothesis to explain our finding of worsened survival when RT was delayed >8 weeks follow a GTR could be a result of re-growth of disease during this extended time delay prior to the start of RT to the point that these delayed cases of GTR represent similar tumor bulk within the radiation volume as STR cases and therefore have worse survival; although there is currently no published evidence of imaging changes following resection of GBM > 8 weeks to further evaluate this idea.
The NCDB provides a valuable tool to study large real-world cohorts and to seek answers to questions that are unlikely to be studied in a prospective randomized clinical trial. However, our study has inherent limitations due to the nature of the NCDB, which can have potential miscoding of variables, selection bias that we cannot detect including the location of the tumor in the brain, the size of the tumor, and any complications from surgery, and missing data for some variables, specifically KPS, use of TMZ, and MGMT promoter-methylation status. These three variables are likely significantly considered when treatment recommendations, including when to start radiotherapy, are being made, and it is unfortunate that more information was not provided for analysis. Secondly, the NCDB only provides all-cause survival without data on progression-free survival or cancer-specific survival.
Furthermore, numerous other prognostic factors have been found to influence survival in patients with GBM, which are not included in the NCDB database. One such factor, the influence of tumor contact with the lateral ventricles, has been shown to decrease survival in patients with GBM on a prior meta-analysis 34 . Additionally, hyperglycemia has been shown to worsen survival in GBM patients treated with surgery, RT, and TMZ 35 . Specifically, the authors found that a glucose level of >6.3 mmol/L portended for worse survival in their retrospective analysis. Finally, numerous imaging characteristics have been evaluated to determine their influence on patient survival, and the degree of necrosis and enhancement on preoperative magnetic resonance imaging studies was found to be correlative 4 .
In conclusion, this analysis of the NCDB suggests that for patients with newly diagnosed GBM, an RT delay of 4 to 8 weeks following resection is associated with better overall survival. The impact of time between surgery and RT was dependent on the extent of resection. Particularly in patients with GTR, delays of longer than 8 weeks was associated with worse survival, making them comparable to patients who receive a STR. This impact of time delay from surgery to the start of RT, in conjunction with the extent of surgery, should be considered in clinical practice and possibly in the design of future clinical trials, particularly those that require additional molecular and genetic analyses that can lead to time delays to start adjuvant treatment.