Abstract
Background/Aim: Standard radiotherapy (RT) for glioblastoma lasts 6 weeks. We aimed to identify patients who would benefit from a hypofractionated approach. Patients and Methods: In 167 patients receiving standard fractionation, 10 factors were analyzed for local control (LC) and overall survival (OS). A survival score was developed and compared to a previous instrument. Results: On multivariate analysis, better LC was significantly associated with the presence of only one lesion and O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation. Better OS was associated with one lesion, better performance status, MGMT promoter methylation, and receipt of chemotherapy. Lesion diameter ≤40 mm and upfront resection were associated with improved OS on univariate analyses. Based on assigning scores to these six factors, three groups, with 32-35, 36-44 and 45-48 points, were designed with 12-month OS-rates of 0%, 56%, and 92%, respectively. Accuracy in predicting death within 12 months and survival ≥12 months was 100% and 92%, respectively, versus 67% and 83% with the previous scoring system. Conclusion: A new survival score with higher accuracy was developed for patients with glioblastoma. Our model can be utilized to individualize RT dose-fractionation recommendations for glioblastoma.
Radiotherapy (RT) is an integral part of the multimodal treatment of glioblastoma, which generally includes maximal safe resection followed by concurrent chemoradiation with temozolomide plus sequential maintenance chemotherapy, with or without addition of tumor-treated fields (1, 2). The standard dose-fractionation regimen of radiotherapy for glioblastoma is 60 Gy in 30 daily fractions of 2.0 Gy over a period of 6 weeks. Another regimen that is used by some centers is 59.4 Gy in 33 daily fractions of 1.8 Gy over 6.5 weeks. The rationale for this regimen is the lower dose per fraction, which is considered associated with less neurocognitive deficits (3, 4). Hypofractionated courses (including doses per fraction >2.0 Gy) have demonstrated equivalent outcomes for elderly patients and those with a poor prognosis. In a randomized trial of elderly patients with glioblastoma, hypofractionated RT with 40 Gy in 15 daily fractions of 2.66 Gy over three weeks was noninferior to standard fractionation (5). More recent studies investigated 52.5 Gy in 15 daily fractions of 3.5 Gy over 3 weeks in elderly or frail patients (6-8). The treatment was shown to be feasible and was associated with favorable outcomes when compared to historical studies. In another randomized trial of elderly (≥65 years) or frail patients, ultra-hypofractionated RT with 25 Gy in 5 daily fractions of 5.0 Gy over 1 week was noninferior to 40 Gy in 15 fractions with respect to overall survival (OS) and progression-free survival (9). Thus, elderly or frail patients may be considered for ultra-hypofractionated RT or moderately hypofractionated RT. Hypofractionated RT over 3 weeks can be combined with temozolomide (10). Patients with more favorable survival prognoses should receive multi-modality treatment including standard fractionation RT (1).
These considerations show that it is important to know a patient’s survival prognosis as precisely as possible when selecting individual treatment. The judgement can be facilitated by prognostic tools. Several survival scores have already been created for patients with glioblastoma (11-17). However, most of these scores are older than 10 years, limited to elderly patients, or did not consider O6-methylguanine-DNA methyltransferase (MGMT) methylation status (11-16). Compared to other existing scores, our previous tool is different, i.e., it is not limited to elderly patients, is more recent, and includes the MGMT promoter methylation status (17). However, it was created from a small sample size of 81 patients, and its accuracy to predict survival for at least 12 months was not optimal. Therefore, we developed a new survival score for patients with glioblastoma from a larger patient cohort and compare the new score to our previous model. We also aimed to identify independent predictors of local control (LC) that may additionally contribute to individualization of RT for glioblastoma, including the identification of patients who may benefit from hypofractionated RT.
Patients and Methods
A total of 167 patients who received standard RT fractionation (5×1.8-2.0 Gy per week) for glioblastoma at the Lübeck Campus of the University Medical Center Schleswig-Holstein between 2012 (installation of new linear accelerators including volumetric-modulated arc therapy) and 2021 were retrospectively evaluated for LC and OS. The study was approved by the responsible Ethics Committee of the University of Lübeck (2022-509).
Ninety-nine patients (59.3%) received 59.4 Gy (5×1.8 Gy per week) and 49 patients (29.3%) 60.0 Gy (5×2.0 Gy per week). The other 19 patients (11.4%) received lower total doses, in 11 due to the proximity of the glioblastoma to organs at risk (e.g., brain stem, optic chiasm) and in eight due to unexpected toxicity during RT or chemoradiation. In 157 patients (94.0%), RT was performed as volumetric-modulated arc therapy. Of the entire cohort, 159 patients (95.2%) received concurrent chemotherapy of temozolomide (1). Maintenance chemotherapy following RT was administered to 132 patients and consisted of six or more courses of temozolomide in 78 of them. Mainly due to toxicity, 54 patients received fewer than six courses of temozolomide, which were followed by procarbazine/lomustine or lomustine alone in eight patients. Upfront gross total resection (GTR) of the glioblastoma was performed in 57 patients (34.1%), and subtotal resection in 70 patients (41.9%). In 40 patients (24.0%), the tumor was considered unresectable and they only underwent a biopsy for histological confirmation of the diagnosis of glioblastoma.
Ten factors (Table I) were analyzed for associations with LC and OS. These factors included the number glioblastoma lesions (1 vs. ≥2), main site of glioblastoma (temporal/parietal vs. frontal/fronto-parietal vs. occipital/parieto-occipital vs. other), maximum diameter of all lesions (median=40 mm; ≤40 vs. >40 mm), Karnofsky performance score (KPS: median=80; ≤80 vs. ≥90), sex (female vs. male), age (median=61 years; ≤60 vs. ≥61 years), MGMT promoter methylation (no vs. yes), extent of resection (GTR vs. subtotal resection vs. biopsy only), addition of chemotherapy (no vs. yes), and total RT dose (≤54 vs. ≥55.8 Gy).
LC and OS were calculated from the day of resection/biopsy. Univariate analyses were performed using the Kaplan–Meier method and the log-rank test. Values of p<0.05 were considered significant. Factors found to be significant in univariate analyses were subsequently analyzed for independence using the Cox proportional hazards model. Again, values of p<0.05 were considered significant. Those factors that were significantly associated with OS on univariate analyses were used to develop the new survival score. For each of these factors, the 12-month OS-rates (as percentages) were divided by 10. The resulting factor scores were added for each patient to obtain the individual patient scores. Based on the 12-month OS rates for the patient scores, three groups were designed for the survival score, representing poor, intermediate and favorable survival prognoses, respectively. In addition, the new survival score was compared to a previous one (17) for accuracy in predicting death within 12 months and survival for at least 12 months, respectively. For the comparisons, we used the positive predictive values (PPVs), which were calculated with the following equations:
PPV for prediction of death within 12 months (least favorable prognostic group):
PPV for prediction of survival for at least 12 months (most favorable prognostic group):
Results
Median follow-up was 16 months (range=2-98 months) for the entire cohort and 19 months (range=3-98 months) in those patients alive at the last contact. On univariate analyses of LC (Table II), improved outcomes were significantly associated with only one lesion (p<0.001), KPS ≥90 (p<0.001), age ≤60 years (p=0.038), MGMT promoter methylation (p=0.002), and GTR (p<0.001). In the multivariate analysis, the number of lesions [hazard ratio (HR)=2.60, 95% confidence interval (CI)=1.51-4.47, p<0.001] and MGMT promoter methylation (HR=0.51, 95% CI=0.35-0.75, p<0.001) remained significant. In contrast, KPS (HR=0.72, 95% CI=0.46-1.14, p=0.16), age (HR=1.38, 95% CI=0.91-2.10, p=0.13), and extent of resection (HR=1.20, 95% CI=0.90-1.60, p=0.21) were not significant in the multivariate analysis of LC.
On univariate analyses of OS (Table III), better outcomes were significantly associated with only one lesion (p<0.001), maximum diameter of ≤40 mm (p=0.040), KPS ≥90 (p<0.001), MGMT promoter methylation (p=0.002), GTR (p<0.001), and addition of chemotherapy (p<0.001). In the multivariate analysis, one lesion (HR=2.53, 95% CI=1.39-4.62, p=0.002), KPS ≥90 (HR=0.50, 95% CI=0.30-0.82, p=0.007), MGMT promoter methylation (HR=0.53, 95% CI=0.34-0.83, p=0.005), and addition of chemotherapy (HR=0.17, 95% CI=0.06-0.47, p<0.001) remained significant. Maximum diameter (HR=1.32, 95% CI=0.83-2.08, p=0.24) and extent of resection (HR=1.16, 95% CI=0.83-1.63, p=0.38) were not significant in the multivariate analysis of OS.
For the new survival score, the six factors, namely the number of lesions, KPS, MGMT promoter methylation status, chemotherapy, maximum lesion diameter, and extent of resection were used. The corresponding points for each factor are shown in Table IV. After addition of the factor scores, the resulting patient scores ranged between 32 and 48 points (Figure 1). Considering the 12-month OS rates associated with the patient scores, three groups were designated, namely 32-35 points (n=6), 36-44 points (n=58), and 45-48 points (n=76). The median survival durations of these groups were 6 months, 15 months, and 30 months, respectively, and the corresponding 12-month OS rates were 0%, 56%, and 92% (Figure 2, p<0.001). The PPVs for correctly predicting death within 12 months and survival for at least 12 months were 100% and 92%, respectively. When applying the previous survival scoring system (17) to the patient cohort of the current study, the 12-month OS rates were 33% (poor prognosis), 74% (intermediate prognosis), and 83% (favorable prognosis), respectively. The PPVs for prediction of death within 12 months and survival for at least 12 months were 67% and 83%, respectively.
Discussion
In addition to new developments of radiotherapy and systemic therapy, outcomes of patients with glioblastoma may be improved by individualization of their treatment (17-22). Regarding RT, individualization includes selection of appropriate RT techniques and dose-fractionation regimens. Standard multi-modality treatment for glioblastoma is based on the randomized trial by Stupp et al. and includes maximal safe resection followed by a 6-week course of concurrent chemoradiation and sequential maintenance chemotherapy (1). To reduce the risk of late sequelae, longer-course RT may be performed with doses per fraction of 1.8 Gy up to 59.4 Gy (3, 4, 17).
A considerable number of patients with glioblastoma have poor survival prognoses. For these patients, a hypofractionated RT course over 3 weeks or an ultra-hypofractionated RT course over 1 week are treatment options. Survival models have been developed to help optimize treatment decisions in terms of RT course (11-17). However, most of the models are older than 10 years and, therefore, do not incorporate the current standard of care (11-15). A more recent score published in 2020 was developed for elderly patients with glioblastoma and may not be applicable to younger patients (16). The majority of the existing scores do not incorporate MGMT promotor methylation status, which is now an important prognostic factor in glioblastoma (11-13, 15). Our group developed a model in 2021 that included MGMT promoter methylation status (17). However, that score also has significant limitations. It was developed from only 81 patients, and its accuracy in identifying patients surviving for at least 12 months was 78% and, thus, not optimal (17). As a result, we developed a new model in the current study with a larger cohort. In the present study, six predictors were identified that were significantly associated with survival on univariate analyses. These factors included the number of glioblastoma lesions, maximum cumulative diameter of the lesions, KPS, MGMT promoter methylation status, extent of upfront resection, and addition of chemotherapy. In comparison to our previous model (17), the new score included two additional factors (maximum cumulative diameter and chemotherapy) and provided weighted factor scores considering the 12-month survival rates (which were divided by 10). The previous system used only 0 points (worse prognosis) or 1 point (better prognosis) as factor scores and was, therefore, less differentiated (17).
The new survival score leads to three prognostic groups, with 32-35 points, 36-44 points, and 45-48 points. Corresponding 12-month survival rates were 0%, 56%, and 92%, respectively, resulting in PPVs of 100% and 92%, respectively, for correct prediction of death within 12 months and survival for at least 12 months. When using the previous score for the present cohort of patients, the corresponding PPVs were 67% and 83%, respectively (17). Thus, the new score was more precise with respect to predicting both death and survival. Considering the poor survival in the group with 32-35 points, these patients may be candidates for ultra-hypofractionated radiotherapy with 5×5 Gy over 1 week (9). Patients within the 36-44 points group should be considered for moderately hypofractionated radiotherapy, for example with 15×2.66 Gy over 3 weeks (5). In selected patients, the dose may be increased to 52.5 Gy in 15 fractions (6-8). Because of the favorable prognosis of the patients within the 45-48 points group, standard fractionation may be most appropriate (1-4). When selecting an individual treatment regimen, the risk of intracerebral progression should also be considered, which can be estimated with the independent prognostic factors of LC identified in this study. These factors, namely the number of glioblastoma lesions and MGMT promoter methylation, were not identified as being associated with LC in one of our previous studies, likely due to the smaller sample size of 91 patients in that study (23). However, when using these prognostic factors and the new survival score developed in this study, one should bear in mind that they are based on retrospective data and require prospective validation.
In conclusion, the new score achieved high accuracy in predicting death within 12 months and survival for at least 12 months in patients receiving multi-modality treatment including standard fractionation RT for glioblastoma. Moreover, the new score was more accurate than our previous model (17). The new survival score can be used to help individualize RT recommendations for patients with glioblastoma.
Acknowledgements
O.Z. received a scholarship from the University of Lübeck within the framework of the emergency aid program for the support of refugee academics from Ukraine.
Footnotes
Authors’ Contributions
The study was designed by all Authors. Data were collected by O.Z. and D.R., and data analyses were performed by D.R. and N.Y.Y. The article was drafted by N.Y.Y. and D.R., and the final version was approved by all Authors.
Conflicts of Interest
The Authors report no conflicts of interest related to this study.
- Received March 16, 2023.
- Revision received March 28, 2023.
- Accepted March 29, 2023.
- Copyright © 2023 International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).