Introduction
Glioblastoma, also known as glioblastoma multiforme (GBM), is the most common primary malignant brain tumour characterized by an infiltrative growth pattern and a low survival rate, with only 6.8% of patients surviving five years post-diagnosis according to the 2019 Central Brain Tumour Registry of the United States (CBTRUS) Statistical Report (Ostrom et al.
2019). The reported incidence rate worldwide varies between 0.59 (South Korea) (Lee et al.
2010), 3.19 (USA) (Thakkar et al.
2014), 3.3 (France) (Fabbro-Peray et al.
2019), and 3.8 (Sweden) (Eriksson et al.
2019) per 100,000 persons.
The ‘standard of care’ treatment for GBM is maximal safe surgical resection, followed by radiotherapy with concurrent and adjuvant temozolomide (TMZ) chemotherapy (Al-Holou et al.
2020). The inclusion of TMZ chemotherapy in treatment guidelines was based on the results from a clinical trial by (Stupp et al.
2005). This landmark trial demonstrated that the addition of concomitant and adjuvant TMZ to radiotherapy provided additional survival benefit to patients, with an increase in median overall survival (OS) of 14.6 months compared to 12.1 months in the radiation only arm. The addition of TMZ is considered a major contributor to survival improvement, from the median 6.9 months in those diagnosed during 1995–1996 to 10.3 months in the period 2010–2015, with 2-year survival peaking at 18%, shown by a Swedish single centre study (Eriksson et al.
2019). Several studies from the US compared survival rates in the periods pre- and post-TMZ approval, predominantly based on the Surveillance, Epidemiology and End Results (SEER) Program database (Zhu et al.
2017; Johnson and O'Neill
2012; Darefsky et al.
2012), but also from the Veterans Health Administration (VHA) dataset (Dubrow et al.
2013). These studies confirmed that survival improvement for GBM patients overlapped with the addition of TMZ to standard therapy protocols. Several European studies reported similar findings (Fabbro-Peray et al.
2019; Eriksson et al.
2019; Scoccianti et al.
2010; Ronning et al.
2012; Hansen et al.
2018). However, even with high compliance to therapy regimens, cancer registry-based studies usually report shorter median OS than those shown by clinical trials.
The aim of this study was to evaluate the distribution of treatment modalities after GBM diagnosis, such as surgery, radiotherapy, and chemotherapy, encompassing 16 years (1999–2014) of data collection by the German Cancer Registries, and to analyse their effect on survival. We hypothesized that there would be an improvement in survival due to technological advancements and more effective treatment options for GBM patients between the first and last period.
Methods
Study population
Data were provided by ‘The Centre for Cancer Registry Data’ at the Robert Koch Institute (RKI) (Kraywinkel et al.
2014) that is tasked with annually merging cancer data from 16 German federal states’ individual registries. This data can be freely requested for research purposes. The dataset contained the following variables: date of birth, date of diagnosis, date of death, sex, histology, grading, tumour localization, International Statistical Classification of Diseases (ICD-10), International Classification of Diseases for Oncology, third edition (ICD-O-3) codes, federal-state residence and treatment modalities. Details on administered radiation doses or specific chemotherapy regimens were not available. We analysed only cases (
n = 44,865) with ICD-O-3 codes 9440 (glioblastoma, 98.9%) or 9441 (giant-cell glioblastoma, 1.1%), and excluded patients with other codes (GBM dataset).
Data quality was assessed by checking the proportion of death certificate only (DCO) or autopsy only cases. The recommendation from The European Cancer Registry-based Study on Survival and Care of Cancer Patients (EUROCARE-5 Study) is that registries should have below 13% DCO cases (Rossi et al.
2015). In the GBM dataset, the proportion of DCO cases was 10.8%. The accuracy of the stated diagnosis is likely to be higher if it is based on histological examination (Bray and Parkin
2009). In the current dataset, diagnosis of GBM was histologically verified in 85.7% of cases. Furthermore, the dataset had no missing information on variables deemed important for the assessment of quality and completeness of registry data, such as age at diagnosis, sex and information on federal-state residence (Bray and Parkin
2009).
Cases were excluded if they were diagnosed by means of an autopsy or DCO (n = 4364), less than 18 years old (n = 358) or had missing information on survival status and exact date of death (n = 5). For cases that were recorded as deceased, but had missing exact date of death (n = 1003), a 0.5 months follow-up time was added, assuming that they survived at least 15 days after diagnosis. The final analysed sample contained 40,138 cases.
Definition of periods and study outcome
Cases were divided into three groups according to the year of diagnosis. We defined the period between January 1999 and December 2005 as the pre-TMZ chemotherapy era, January 2006–December 2010 as the period when TMZ chemotherapy became part of the standard therapy protocol (TMZ era), and the period between January 2011–December 2014 as the modern era. Cases were censored at 31st December 2014 (last possible date to be included in the GBM dataset) or after 30 months of follow-up, which ever came first.
The main study outcome was median OS. First, the median OS for all cases was calculated and then evaluated separately for the three defined time periods. Secondly, median OS of cases who received surgical treatment and radiotherapy was compared to patients who received all three treatment modalities, following similar categorization from two previous major clinical trials (Stupp et al.
2005; Perry et al.
2017). Thirdly, age-stratified analysis of median OS was performed for each period. Finally, we analysed overall distribution of each treatment modality.
Statistical analyses
Calculation of median OS was according to the Kaplan–Meier (KM) method. Survival probabilities were graphically presented with KM plots and subgroups were compared with the log-rank test. Cox proportional-hazards models were used to examine survival for each period by computing hazard ratios (HR) with 95% confidence intervals (CI). Models were adjusted for age, sex, and each treatment modality. Information on treatment was available as a dichotomous variable (surgery yes/no, radiotherapy yes/no, chemotherapy yes/no). The significance of three interaction terms (each treatment modality × time period) was also explored in the final model. No competing risk analysis was performed as GBM has characteristic short duration of survival and to avoid bias from erroneous recording of cause of death. Sensitivity analysis was performed to determine the stability of the HRs when federal states with more than 30% missing information on treatment modalities were excluded. Logistic regression models were used to determine if age, sex and time period were predictors for missing information on treatment. Additional information on the statistical analyses is included in the Online Supplemental Resource. All analyses were performed with SAS 9.4 (SAS Institute, Cary, NC, USA).
Discussion
Our results showed that advancements in treatment routines, especially after 2005, led to the improvement of survival based on the Cox model estimates and improved 2-year survival for GBM patients in Germany. First, the median OS was 10 months, which is in line with results from other cancer registry studies (Fabbro-Peray et al.
2019; Eriksson et al.
2019; Zhu et al.
2017; Johnson and O'Neill
2012; Hansen et al.
2018) and the clinical trial in elderly patients by Perry et al. (Perry et al.
2017). Analyzing the age distribution, cases were mostly in the older age categories, which could explain the lower OS in the present study compared to the trial by Stupp et al. (
2005), which had an upper limit (71 years) on age when enrolling participants. Taking into account that clinical trials perform a rigorous selection of patients, with a different distribution of comorbid conditions and better access to treatment, median OS is usually higher than what is observed in population-level studies (Pulte et al.
2019). Secondly, we observed that the median OS, when stratified by age, treatment and time period, improved after 2005. The addition of TMZ chemotherapy to surgical resection and radiotherapy increased median OS for the older age groups, however, it had no to little effect in the younger age groups. The interaction analysis revealed that the effect of treatment modalities changed over time, with the biggest improvement stemming from chemotherapy use and advancements in surgical treatment. We cannot answer based on currently available data whether this is a result of improved precision or the extent of surgery. However, regarding chemotherapy, the only chemotherapy agent approved for first-line treatment of GBM has been TMZ, to which we can attribute most of the improvement. Finally, the absence of cancer-related treatment was more common in the older age groups, while multimodal therapy was more customary in the younger group.
A key factor that influenced survival was the age at diagnosis. The survival disparity between the different age groups was expected since survival declines with increasing age (Laperriere et al.
2013). A recent clinical trial by Perry et al. examined the treatment paradigm for patients older than 65 years, and concluded that the addition of TMZ to short-course radiotherapy resulted in longer survival than radiotherapy alone (Perry et al.
2017). This is relevant for clinical practice, taking into account that 35% of our cases were in the 70 and older age category. Additionally, molecular prognostic factors for GBM are correlated with age. These include O
6-methylguanine-DNA methyltransferase (MGMT) gene promotor methylation and isocitrate dehydrogenase (IDH1) mutations (Weller et al.
2009). About 10% of GBMs are IDH gene mutated and this is associated with a better prognosis, while typically GBMs are IDH wild-type. Elderly patients are more likely to have IDH wild-type, which indicates a shorter survival compared to younger patients (Zhu et al.
2017). Additionally, older patients are burdened with more comorbidities and have more treatment-related side effects (Brandes et al.
2009). However, this group often has methylation of the MGMT promotor, which offers chemosensitivity and increased benefit from TMZ treatment (Perry et al.
2017; Wick et al.
2012; Hegi et al.
2005).
Survival rates for patients with GBM have shown little notable improvement in the last two decades. Our results are comparable to the median OS and 2-year survival rates reported by other population-based registry studies using SEER, VHA, CRN (Cancer Registry of Norway) (Ronning et al.
2012), the FBTDB (French Brain Tumor Database) (Fabbro-Peray et al.
2019) and the Danish Neuro-Oncology Registry (Hansen et al.
2018). According to the study by Dubrow et al. (
2013), addition of TMZ chemotherapy to standard therapy protocols explained the improved survival observed on a population level, even though they showed much lower median OS (6.4 months). Our results are comparable to the study by Zhu et al. (
2017) that reported 10 months median OS and a 2-year survival rate of 16.8%. Other European studies, such as from France, reported 11.2 months (Fabbro-Peray et al.
2019), Norway, 8.3 and 10.1 months in the pre-TMZ and TMZ eras, respectively (Ronning et al.
2012), and Denmark, 11.2 months (Hansen et al.
2018). A clear explanation why there is little change in survival from GBM in Germany during 16 years is difficult. One possible reason could be that cases diagnosed at an older age, whose numbers increased substantially due to demographic changes or improved case ascertainment in the cancer registry, are subjected to less aggressive treatment protocols. A study from Fabbro-Perray et al. (
2019), stated that in ‘real-world’ setting, only 60% of patients initiate standard treatment, and while we are unable to confirm this from our data, the overall use of surgery, radio- and chemotherapy was relatively high and increased in subsequent periods, meaning this could not completely explain the discrepancy.
Glioblastoma multiforme is a costly disease due to hospitalization and life-years lost in the post-diagnosis period, hospice care, and claims for physical, occupational and speech therapy (Aly et al.
2019). Before our analysis, we expected a more substantial improvement in survival of GBM patients, considering advances in surgical techniques and the introduction of postoperative radio-chemotherapy with adjuvant chemotherapy as a standard treatment after 2005. The implementation of neuronavigation and the introduction of intraoperative 5-aminolevulinic acid (5-ALA) fluorescent microscopy have enhanced the possibility to achieve a better gross total tumor resection. Our analyses revealed that surgery is associated with better survival, as also reported by another study (Al-Holou et al.
2020). The effect of radiotherapy advancements was more modest even though there were several novel radiotherapy choices. Intensity-modulated radiation therapy represents an advance option in terms of conformal treatment. Additionally, tumor-treating fields (TTFs) are an important cancer treatment modality with antimitotic effects against rapidly dividing tumor cells (Fabian et al.
2019). This is an upcoming new standard of care for both newly diagnosed and recurrent GBM cases (Kazda et al.
2018). More aggressive treatment of recurrent disease, with the application of stereotactic radiotherapy and improvements in palliative care, might also contribute to better survival. Regarding chemotherapy, the use of TMZ has been consistently associated with improved survival. Co-administration of radiotherapy and TMZ has nearly tripled the 2-year survival of GBM patients in the last two decades from 10 to 27% and quadrupled to 47% in patients with MGMT promotor methylation (Kazda et al.
2018). However, current clinical trials have reported improved chemotherapy benefits only in patients with prognostically positive molecular profiles, and this is important evidence toward the implementation of a personalized treatment strategy (Herrlinger et al.
2019; Bent et al.
2019). Several known and newly discovered chemotherapy and immunotherapy agents have been tested, offering modest or no improvement in median OS (Gilbert et al.
2014; Weller et al.
2017), but no second-line treatment has been established. Regardless of modern state-of-the-art treatment, due to the highly invasive nature of GBM, median OS remains between 10 and 14 months (Seyfried et al.
2019). Additional research is needed to improve the poor prognosis of these patients.
Strengths and limitations
The strength of our study is that it uses a large, nationally representative database of German GBM patients, classified by the ICD-O-3 coding system, with satisfactory data quality and completeness of the information. Registry-based studies are considered to reflect more accurately ‘real-world’ settings. There are also some potential limitations. Cancer registration in Germany was not nationwide before 2009, which could influence the lower number of cases seen in the first period. Additionally, epidemiological cancer registries do not routinely collect information on specific chemotherapy drugs or the applied regimen. It is possible that some cases did not receive the standard TMZ treatment, but a different chemotherapy medication, such as lomustine. In the pre-2005 era, German guidelines recommended the use of carmustine plus teniposide or nimustine plus teniposide based on results from two clinical trials (Weller et al.
2003). However, one study has shown that the most frequent chemotherapy modality after 2005 is TMZ with a median duration of 2.4 months (Aly et al.
2019). Another limitation is that no additional variables associated with survival were collected, such as performance status, the extent of surgery and presence of comorbidities (Eriksson et al.
2019). Due to lack of information on comorbidities we cannot exclude confounding by indication, which is a limitation of observational data, as opposed to randomized controlled trials. In addition, the sensitivity analysis revealed that recent periods had more missing information on treatment modalities and this was exacerbated by older age at diagnosis. Thus, the interpretation of the main effect estimates remains challenging. Finally, data on molecular factors were unavailable.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.