Five-year survival in patients with Glioblastoma is overestimated in registry data - A nationwide population-based Swedish survey during 1958 - 1999

Background Some glioblastoma (GBM) patients survive more than ve years with hitherto no clearly established epidemiological or molecular causes. Since varying rates of GBM ve-year survival have been reported our aim was to assess the true prevalence of ve-year survivors in Sweden from 1958 to 1999, before the introduction of concomitant temozolomide treatment. Methods After screening the Swedish Cancer Registry and the Cause of Death Register 736 out of 12765 patients with high-grade glioma were dened as ve-year survivors. The full text pathology report was reviewed in 585 patients. Data on epidemiology and treatment were retrieved from the medical records of 556 patients. Results 77 ve-year survivors with primary GBM were identied which corresponded to 0.60 % of the initial population. During 1990 to 1999 GBM ve-year survival was 0. 90%. Younger age, Karnofsky score > 70 and non-eloquent tumour location were found in most but not all ve-year survivors. Conclusion GBM ve-year survival was exceedingly rare in Sweden until 2004, comprising less than 1 % of registered HGGs. Relying on registry data without reviewing the pathology report will overestimate the accurate number of ve-year survivors. To our knowledge, this is the only nationwide population-based study of ve-year survival in GBM patients.


Introduction
Glioblastoma (GBM), the most common malignant brain tumour, carries the worst prognosis of all human tumours, with even shorter survival than in pancreatic cancer and mesothelioma [1]. GBM usually grows aggressively with excessive seeding of tumour cells outside the bulk of the tumour, often in ltrating vital functional brain areas making radical surgical resection impossible [2][3][4]. The biology of GBM is overly complex with genetic and molecular heterogeneity both between patients but also within the same patient making targeted oncological therapy challenging [5][6][7]. The current standard treatment for newly diagnosed GBM patients is maximal safe surgery followed by radiotherapy with concomitant temozolomide and then adjuvant temozolomide [8]. The median survival in patients receiving this regimen is about 14 months [9]. However, due to that not all patients receive standard therapy, median survival for all diagnosed GBM patients is considerably less, being reported from 5.8 to 11.2 months [10][11][12][13]. Certain clinical and molecular characteristics such as younger age, good performance status and gross total surgical resection have been associated with longer survival together with O6-methylguanine methyltransferase (MGMT) promoter methylation status [14][15][16][17][18][19][20][21]. Recently, studies have indicated a remarkable gain in overall survival in subgroups of GBM patients [22][23][24], but even in the most favourable conditions the majority succumb to the disease within 2 years [25]. However, a subset of GBM patients seem more amenable to treatment, having longer progression free and overall survival and are called Long Term Survivors (LTS). The de nition of LTS varies but is commonly de ned as survival >36 months [26]. An even smaller minority survives ve years or more after diagnosis.
There is also a concern about the true number of ve-year survivors due to misdiagnosis and the varying de nition of anaplastic astrocytoma, oligodendroglioma and oligoastrocytoma depending on time and place [28,38,39]. Although the reported numbers need to be interpreted cautiously, some GBM patients have undoubtedly bene tted from the last decade's treatment algorithms with a prolonged overall survival, especially concerning two and three-year survival but not necessarily ve-year survival [75]. In the post temozolomide era, studies investigating molecular factors in GBM LTS patients has been somewhat con icting indicating MGMT promoter methylation as a strong predictor for long term survival whereas the role of isocitrate dehydrogenase (IDH) mutations are less clear [15,[40][41][42]. Previous case reports have implied an association between postoperative infections and LTS [43] but recent retrospective reports have not been able to con rm this [44,45].
Autoimmunity has been reported to prolong survival in glioma [46] but this has not been veri ed in other series [47].
Nevertheless, GBM patients with extended survival may display important prognostic factors for longer survival, hopefully enabling new targets for treatments in the future.
Historically there has been a gradual implementation of new surgical techniques and oncological treatments for GBM since 1960 but the impact of these methods has not been estimated in population-based studies. It is important to assess these changes to de ne the survival status before the introduction of temozolomide in the concomitant and adjuvant settings. The aim of this study was therefore to identify GBM ve-year survivors in Sweden during 1958 -1999, before the introduction of Patients 12765 patients with a diagnosis of HGG were identi ed. 736 of these were de ned as ve-year survivors. The aim was to review the full pathology report, neurosurgical and oncological charts in these patients from the six neurooncological regions in Sweden (South, Southeast, Stockholm, West, Middle and North). 585 full pathology reports and 556 medical charts were found. 151 pathology reports and 180 charts were irretrievable ( g 1). The lost charts were mainly from the earlier decades (1958-1969; 1970-1979 -145 patients, 80%). However, all full pathology reports from patients in the last decade (1990-1999) were retrieved. Data regarding previous disease, presenting symptoms, functional status, previous history of trauma, infection and tumor, radiological exams, neurosurgical operations, re-operations, diagnosis, oncological treatment, and overall survival was extracted from 556 patients with complete medical charts.
Surgical resection was scored as gross total resection (GTR), partial resection or biopsy. Since postoperative MRI was not performed routinely in Sweden until the beginning of the 21st century, resection grade was estimated after reviewing the neurosurgical operative report, i.e. the surgeon´s report. Epidemiological data was collected into a database (File Maker Pro and Excel) for graphics and further statistical analysis.

Statistical methods
Statistical analyses were performed with the free statistical software R-project (R core team (2017). R: A language and environment for statistical computing. R Foundation for Statistical Computing, https://www.R-project.org/) Standard descriptive statistics, tables and diagrams were used to display trends and epidemiological data. Differences in overall survival between different age groups and time periods were compared with the Kaplan-Meier log rank test.

Ethical approval
Ethical approval was obtained from the Regional Ethical Review board in Lund, Sweden (Dnr 617/2004). All research was performed in accordance with the Declaration of Helsinki. Informed consent was obtained from alive adult patients and legal guardians from patients < 18 years.

Results
Survival in patients with high-grade glioma was related to time period and age In the population of 12765 HGG patients, demographics showed a peak in incidence in the sixth and seventh decade ( g 2).

Symptoms and signs
Medical charts from four patients where the full pathology report had veri ed GBM diagnosis could not be retrieved. These with full radiotherapy (56-60 Gy) whereas 4 patients (5 %) did not. 38 patients (52%) did also receive chemotherapy (table 3).

Sample evaluation
The full neuropathology reports were reviewed by the authors as described above. Thirty-three patient samples had been reexamined during the clinical course due to an unexpectedly long survival, warranting a new examination. In additionally twenty-four patients, the diagnosis of GBM was con rmed through iterated surgeries. In sixteen patients, no new evaluation had been made. Hence, in 57 out of 73 cases the GBM diagnosis had been reexamined and veri ed. Additional microscopic sample examinations were not performed by the authors.

Overall Survival
Overall survival in GBM ve-year survivors ranged between 5-55 years with a median survival of 3251 days or approximately 9 years (8 years, 11 months) ( g 5b). Ten patients were still alive in January 2020. Nineteen patients (24%) survived ve but less than six years after diagnosis, representing the largest cluster in the cohort. Thirty-ve patients (0.27%) survived more than ten years.

Discussion
Most of the registered HGG ve-year survivors had an actual diagnosis of anaplastic astrocytoma, astrocytoma grade 2 and other non GBM diagnoses. Of the 585 HGG patients with a veri ed diagnosis, only 77 had a diagnosis of primary GBM which constituted 0.60 % of the total population, illustrating the potential aws when extracting data from registries without verifying the diagnosis [28,39]. Considering the 151 reports that could not be retrieved the true number of ve-year survivors may be slightly higher but supposedly less than 1% as the 1990-1999 cohort had a ve-year survival of 0,9 % (table 1). This number is below the previously reported gures from before year 2000 of 1-2 % [28, 29]. Recent population-based studies estimated GBM ve-year survival between 2.0-2.4 % [10,13]. However, in the American study encompassing over 100000 GBM cases only patient data from accredited cancer centers were included thus possibly overestimating the actual ve-year survival [10]. Furthermore, no review of pathology was made adding additional uncertainty of the actual numbers of veyear survivors. Given that completeness of registry data is less in older patients for brain tumors our estimate might still be too high [50]. Notably, thirty-ve patients (0.27%) lived more than ten years after diagnosis which is in proximity to data from a systematic review on ten-year survivors of 0.71 % [52].
Headache was the most common presenting symptom in 73 GBM ve-year survivors, followed by focal neurological signs, nausea and seizure, correlating with data from other GBM studies [4,53,54].
Some studies have reported longer survival in diffuse glioma patients with a history of allergy and autoimmune disease whereas this relationship was unclear in GBM patients [55]. Only six (8%) patients in our study had a history of allergy, asthma or autoimmune disease, which does not support the theory that previous allergy or autoimmune disease have a protective or treatment promoting effect once GBM has been established. However, our ndings do not contradict the notion of an inverse association between glioma and allergic conditions, protecting the host from developing glioma [56-61].
Five patients (7%) had been treated for prior tumours, but none had a history of other brain tumours which is congruent with About half of the 73 patients (49%) were operated with GTR, the rest with either partial resection (44%) or biopsy (7%).
Twenty-four (33 %) had repeated surgery. This data from the pre-MRI period needs to be interpreted cautiously, being based on surgeon´s report, with a reported low accuracy of approximately 30% [69]. As the extent of resection (EOR) plays a role in GBM survival [14,[22][23][24][70][71][72][73] the percentage of GTR among GBM ve-year survivors in our study seems surprisingly low.
Almost all patients (95%) were treated with radiotherapy whereas a small majority (52%) received chemotherapy, mainly BCNU during the latter decades. The current standard therapy with concomitant radiochemotherapy followed by adjuvant temozolomide have had a signi cant impact on overall survival in clinical trials [9,25,37] and also in population-based reports, especially regarding two-and three-year survival [1, 10-13, 29, 74,75]. However, some of these reports are based on data from the SEER database, a registry that comprises approximately 35 % of the US population with an uneven geographical distribution (https://seer.cancer.gov/data/). Full concomitant radiochemotherapy is mainly given to younger patients deemed clinically t and only a minority of elderly patients above 65 years receives this regimen. The overall impact of temozolomide on GBM ve-year survival is less clear. In a recent meta-analysis ve-year survival was not overly affected before and after 2005 when temozolomide treatment was initiated in clinical practice [75]. Similarly, a newly published population based French study reported a ve-year survival of 2, 4 % [13]. Other studies report a wide variety of GBM veyear survival, ranging from 1-10 % [1, [10][11][12][27][28][29][30][31][32][33][34][35][36]. This variation may be caused by selection bias of treated patients and controls but possibly also due to publication bias, e.g. reporting of lower survival might be less frequent.
There has been an increasing focus on molecular markers in GBM LTS, especially MGMT promoter methylation and IDH mutations. MGMT promotor methylation status is well described as a predictive marker of response to temozolomide in GBM. Several studies have shown a high percentage rate of MGMT promoter methylation in GBM LTS, particularly in patients surviving > 3 years [26, 76-79] but not necessarily in ve-year survivors [21]. Mutation in IDH occurs frequently in low grade and anaplastic astrocytoma and is associated with better prognosis. However, this association seems to be weaker and sometimes lacking in GBM LTS (40-42, 78, 79]. Some, especially younger GBM patients with IDH wt together with ATRX mutations, have been reported to have longer overall survival [80] but this has not been corroborated by others [81]. In summary, no single factor has so far been able to explain ve-year survival in GBM patients either before or after the introduction of concomitant temozolomide.

Limitations
This study has several limitations. It is retrospective, comprising more than 40 years of the Swedish Cancer Registry. Different brain tumor classi cations, errors in registration and lack of completeness are common pitfalls when extracting data from a registry with an obvious risk of gathering and interpreting incorrect data. It is not known how many of the 12765 HGG patients that were true primary GBM and survived less than ve years since it would be unfeasible to review the pathology reports for all patients. However, it is probable that non-GBM diagnosis was much less frequent in patients with a survival less than ve years. Furthermore, the diagnosis was based on previous pathology assessment and then interpreted according to the 1993 WHO classi cation. Errors when interpreting the pathology specimen leading to wrong diagnosis has been reported [35,36]. However, fty-seven of the GBM ve-year survivors (75 %) had been reevaluated and hence reviewed by another neuropathologist verifying the diagnosis. Irretrievable charts from approximately 25 % of the initial study population of 736 patients and the lack of postoperative radiology for de nitive estimation of tumour residual are also important limitations.

Conclusions
The actual GBM ve-year survival in Sweden was 0, 60% during 1958-1999 and 0.9 % during 1990-1999, prior to implementation of concomitant temozolomide treatment. These numbers are well below reported ve-year survival in clinical trials and most population-based studies. Younger age, good preoperative functional status, tumor location and super cially located tumors appears to be the most important clinical prognostic factors for GBM long-term survival but do not explain all ve-year survivors. To our knowledge, this is the only nationwide population-based study of ve-year survival in GBM patients.