Evaluation of a cohort of long-term surviving patients with glioblastoma

Summary

Aim: The aim of this study is to analyze a cohort of patients with glioblastoma multiforme treated at the neurosurgery department in Ústí nad Labem, whose overall survival time exceeded 3 years. Patients and methods: In a prospectively conducted group of 22 patients, we confirmed glioblastoma IDH wildtype in 21 patients, including the Ki-67 proliferation index, including examination of samples after repeated resections. We describe the therapeutic procedure of patients in the cohort: Stupp regimen including subsequent administration of chemotherapy beyond 6 cycles, number of reoperations, and use of stereotactic radiosurgery using a gamma knife. In addition to comparison with other published cohorts, we evaluate the preoperative characteristics of tumors on MRI, namely the presence of midline shift greater than 10 mm, involvement of the right or left hemisphere, and involvement of the cortex or periventricular region. Results: The median overall survival, excluding survival estimates for 11 living patients according to Kaplan-Meier, is 53 months (4.4 years) ranging from 38 months to 228 months (19 years). In the analysis of the entire patient population (including censoring of surviving patients), we observed a median overall survival of 142 months (11.8 years) at a median follow-up of 74 months, with a 95% confidence interval for the median survival time ranging from 46 to 142 (not reached) months. Survival time > 4 years was observed in 15 patients (71%), > 5 years in 10 patients (48%), and 10 years or more in 6 patients (29%). Conclusion: The effort to achieve the most radical resection and completion of radiochemotherapy according to the Stupp protocol were the common denominators in our cohort of 21 patients with glioblastoma multiforme IDH wildtype with overall survival > 3 years. We cannot confirm the adverse effect of the primary characteristics of MRI of the tumor –⁠ the initial midline overpressure and the involvement of the periventricular region.

This is an unauthorised machine translation into English made using the DeepL Translate Pro translator. The editors do not guarantee that the content of the article corresponds fully to the original language version.

 

Introduction

Despite advances in comprehensive multimodal treatment of multiform glioblastomas (GBM), treatment outcomes remain unsatisfactory and are practically comparable to the data presented in the original study by Stupp et al. from 2005 [1], on the basis of which the current standard of postoperative oncological treatment was defined. In our retrospective analysis of a cohort of 101 GBM patients operated on between 2015 and 2017, we observed a median overall survival (mOS) of 8.8 months at a median follow-up of 22 months [2]. Patients eligible for Stupp's regimen had significantly longer overall survival than patients who did not receive this treatment (mOS 22.6 vs. 4.3 months). There was also a significant difference when comparing overall survival according to the extent of surgery (mOS 12.4 months after resection vs. 4 months after biopsy) and age. Younger patients (≤ 60 years) had significantly longer overall survival (mOS 19.4 vs. 5.1 months). Very similar OS observations in a group of 50 patients with GBM, including volumetric assessment of residual disease after resection, were described in Czech literature by Halaj et al. [3], where patients after radical resection (gross total resection; GTR) and completion of the Stupp regimen achieved an OS prolongation with an average of 19.6 months (mOS 14 months). The large differences observed in the above-mentioned patient categories led us to approach the issue from the opposite perspective (focusing on follow-up) and to describe in greater detail the group of patients who achieve longer overall survival, defined for the purposes of this study as survival longer than 3 years. Historically, the phenomenon of longer survival has been associated more with the diagnosis of so-called secondary glioblastoma arising from the gradual dedifferentiation of a glioma of originally lower grade. However, under the current WHO classification from 2021 [4], this term is no longer used, and these gliomas are classified as IDH (isocitrate dehydrogenase) mutated astrocytomas, WHO grade 4. In Czech literature, the current view on histogenetic diagnosis is clearly presented in an article by Brno authors Hendrych et al. [5] clearly presents the current view of histogenetic diagnosis in Czech literature. GBM is considered to be a diffusely growing glioma with no mutations in the IDH1/2 genes or histone 3 genes (H3F3A or HIST1H3B/C) and the presence of one or more morphological or genetic features, which are: necrosis, microvascular proliferation, and telomerase reverse transcriptase (TERT) promoter mutation, epidermal growth factor receptor (EGFR) amplification, or combined gain of chromosome 7 and loss of chromosome 10 (CNA +7/–10).

The aim of this analysis of a prospectively maintained database of patients with GBM (i.e., patients without IDH mutation) and survival longer than 3 years is to map the basic clinical and pathological characteristics of this patient group as a necessary starting point for further follow-up research aimed at identifying biological markers/changes responsible for longer survival in patients with GBM.

 

Material and methodology

Patient cohort

Patients were enrolled in a prospectively maintained database of GBM patients with survival longer than 3 years since 2016 by neurosurgeons from the Neurosurgery Clinic of J. E. Purkyně University, Masaryk Hospital, KZ a. s. in Ústí nad Labem during outpatient check-ups involving clinical examination and MRI. These were not only newly diagnosed patients (diagnosed after 2015), but also patients who had already been diagnosed for 3 or more years in 2016. The histology of these patients was verified by a second reading by two independent pathologists. IDH status was examined in sections from formalin-fixed paraffin blocks immunohistochemically with the IDH1 antibody (clone H09, Dianova [Boden, Sweden]), and in patients younger than 55 years, IDH gene mutation analysis was performed using the Real-Time polymerase chain reaction (PCR) (Idylla IDH1-2 Mutation Assay Kit [RUO] from Biocartis NV, Mechen, Belgium) and Sanger sequencing in two patients). The database recorded the age at diagnosis, patient gender, date of primary surgery, IDH1 status, number of surgeries and their highest radicality, completion of conventional radiotherapy, number of cycles and chemotherapy used, and use of gamma knife stereotactic radiosurgery (SRS) boost. The database was continuously updated (Bartoš), with information on progression and OS (overall survival) added. For the purposes of this analysis, the database was closed on June 30, 2025. Based on a literature review, findings on MR T1-weighted images with contrast agent from the period immediately before the first operation were retrospectively evaluated in a post hoc analysis in all patients to assess the initial midline shift and the distance of the enhancing tumor margin from the ventricular system (involvement of the periventricular area). In both cases, 10 mm (inclusive) was the limit for dichotomous division of the sample into two subgroups. In the above-mentioned MR sequence, we also assessed the possibility of multifocal localization of the enhancing lesion. On T2/fluid attenuated inversion recovery (FLAIR) weighted images, we evaluated the presence of a significant cystic component of the tumor.

 

Statistical analysis

Basic descriptive characteristics were used to describe quantitative (median and range) and qualitative data (frequency and percentage representation). Kaplan-Meier estimation was used to estimate overall survival, and in the subgroup of patients who had already died (i.e., without the need for censoring), the simple median was also calculated.

 

Results

Between 2016 and 2024, we performed primary resection surgery for GBM in 255 patients, an average of 28 patients per year. In our cohort, we recorded a total of 22 patients with GBM with OS longer than 3 years (of this group of patients, a total of 13 patients were operated on after January 1, 2016). The others were operated on earlier, with the patient with the longest interval having been operated on as early as 2008. These patients included 14 women and 8 men; the median age at diagnosis was 58 years (range 35–72 years). The IDH1 status was verified in these patients, in those younger than 55 years of age by PCR or Sanger sequencing; one patient with focal IDH1 positivity and a detected IDH1 gene mutation in codon 132 was subsequently excluded. Thus, the total number of patients in the cohort with IDH wild-type (IDHwt) GBM is 21, of which two patients had invalid PCR results. GBM was multifocal in two patients (9.5%), and a cystic component was present in four patients (19%). Significant midline pressure ≥ 10 mm was present in seven patients (33%), and the distance between the enhancing edge of the tumor and the ventricle was ≤ 10 mm in 12 patients (57%). Only 50% of patients with tumors exhibiting both characteristics (large shift and proximity to the ventricle) were consistent. In one patient, the histopathologist described a giant cell GBM, while in another, a small-cell GBM.

In 14 patients, the tumor was located in the left hemisphere (67%) and in 7 in the right (33%). The exact location is not clear, as in 10 patients more than one lobe was affected by the tumor. However, the occipital lobe was affected most frequently (in 10 patients), followed by the parietal lobe (in 8 patients) and the frontal and temporal lobes (7 each). All tumors affected the cortical surface, although in three cases the temporobasal surface and in one case each the frontobasal and falcine surfaces were affected, but predominantly, in 16 cases (76%), it was the surface of the cerebral hemisphere convexity.

The median overall survival without taking into account survival estimates for 11 living patients according to Kaplan-Meier is 53 months (4.4 years) ranging from 38 months to 228 months (19 years). In the analysis of the entire patient population (including censoring of surviving patients), we observed a median overall survival of 142 months (11.8 years) with a median follow-up of 74 months and a 95% confidence interval for the median survival time ranging from 46 to 142 (not reached) months. Survival time > 4 years was observed in 15 patients (71%), > 5 years in 10 patients (48%), and 10 years or more in 6 patients (29%). Currently, 11 patients are alive. The characteristics of the cohort and the treatment modalities used are summarized in Table 1 and Figure 1.

Discussion

In this analysis of a prospective cohort of 21 long-term survivors with GBM, we observed a median overall survival of 4.4 years and, when censoring 11 patients still alive, 11.8 years. Our patients can be compared with the cohort published by Krex et al. [6], who recorded 55 patients with OS > 3 years. These were patients from a joint database of seven German clinical centers (German Glioma Network). Apart from younger age, good Karnofsky Performance Status (KPS) at admission, and a higher prevalence of hypermethylation of the methylated-DNA-protein-cysteine-methyltransferase (MGMTp) promoter (74% of 36 examined), no factors predisposing patients diagnosed with GBM to long-term survival could be identified. Recurrence and reoperation do not rule out the possibility of OS > 3 years; reoperation for recurrence was performed in 25 patients (45%) in the Krex cohort, 12 patients (22%) were reoperated twice, and 4 patients (7%) were even reoperated three times. In our cohort, the frequency of reoperations is even higher, with 12 out of 21 patients (57%) undergoing reoperation, and one patient undergoing reoperation 4 times. In the same number of patients (12 out of 21), following a joint indication at the Department of Stereotactic and Radiation Neurosurgery at Na Homolce Hospital, stereotactic radiosurgery using a gamma knife boost was also performed during the course of the disease, ranging from 1 to 3 procedures (Fig. 2). Of the possible techniques for re-irradiation in GBM recurrence, SRS has low toxicity in terms of radionecrosis and can also be targeted using metabolic imaging, such as PET CT or MR [7,8]. However, the data sets differ in this respect, as in the cited Krex data set, SRS was used in only three patients (5.45%) during treatment. What they have in common is that all patients completed conventional photon radiotherapy, and in the German cohort, which was closed at the time of the standard introduction of the Stupp protocol into clinical practice, 37 patients (67%) underwent chemotherapy, most commonly with nimustine (ACNU). In our cohort, which is from a later period than Krex's above-mentioned cohort, chemotherapy was already administered as standard to all patients, in the range of six or more cycles (Stupp protocol with temozolomide [TMZ] in 95%). In two patients, alternative chemotherapy with lomustine was also used as a second line. The average number of chemotherapy cycles for all 21 patients was 24, and the maximum number of cycles achieved by one patient was 100. A study by Chinese authors [9] is devoted exclusively to this topic. In a group of 27 patients treated with six cycles of TMZ, the median OS was 19.4 months and PFS (progression-free survival) was 15 months, compared to 26 patients treated with more than six (7–41, average 10) cycles of TMZ, who had a median OS of 25.6 months and PFS of 20.1 months. Treatment with more than six cycles of TMZ was a statistically significant factor for 2-year survival (66% vs. 36%) and PFS. However, in multivariate analysis, it did not have a positive effect on OS. The authors note, however, that long-term use of TMZ does not increase the likelihood of neutropenia and thrombocytopenia compared to six cycles, so from this perspective, the above-standard number of cycles in our patients is justified.

A significant cohort of long-term GBM survivors, moreover with OS > 5 years, has recently been described in a multicenter (20 centers) study ETERNITY (EORTC 1419) [10]. It evaluates 189 IDHwt GBMs with a median survival of 9.9 (7.9–11.9) years, and surprisingly, this cohort does not show the standard prognostic factors such as age, baseline KPS, radicality of resection, and MGMTp hypermethylation. The only prognostic factor is the absence of tumor recurrence. Patients without tumor recurrence, who numbered 41 (23.3%), achieved a longer median survival time (unclosed at the time of study completion) compared to 135 patients (71.4%) with one or more recurrences (8.92 years; p < 0.001). Surprisingly, patients without tumor recurrence had a higher frequency of MGMTp unmethylated tumors (48.8%). A certain limitation of the study is the fact that in 13 patients (6.8%), the presence or absence of recurrence was not recorded. When evaluating our nine patients with OS > 5 years, four of them did not experience recurrence and their average OS is 7.1 years compared to patients with one or more recurrences, who achieved an average OS of 7.5 years. Therefore, the absence of recurrence cannot be considered a prognostically favorable sign. The ETERNITY study cohort also includes a group of 80 astrocytomas, IDH-mutated, CNS WHO grade 4, which are no longer considered GBM according to the 2021 classification [4,5]. These patients achieved a median survival of 10.7 years (9.9–11.2). One of our patients with only focal IDH1 positivity but a proven IDH1 mutation did not achieve a significantly above-average survival time compared to other patients in our cohort, but we excluded her from the database because, according to the 2021 classification, she has an IDH-mutated astrocytoma, CNS WHO grade 4 (Fig. 3).

Italian neurosurgeons from Treviso retrospectively evaluated 33 patients with OS > 3 years (long-time survivors; LTS) and 7 patients with OS > 10 years, whom they referred to as very long-time survivors (VLTS), from a cohort of 660 GBM patients [11]. Although younger age (< 50 years), Ki-67 value < 10%, IDH1/2 and ATRX mutations, non-TERTp mutations, and MGMTp hypermethylation were associated with a higher probability of LTS and VLTS, the study concluded that the only independent prognostic factors for VLTS are younger age at diagnosis and MGMTp hypermethylation. In our cohort, we recorded six patients assessed as VLTS, three of whom are alive. The longest survival time is 19 years for a patient who underwent surgery in 2008. The Treviso cohort is therefore one of the few examples of the possibility of survival beyond 10 years. In the discussion, the authors state that by 2018, only 37 patients (in series by 26 authors) with survival of more than 10 years had been published. However, this data is inaccurate, as in 2018 Tykocki T. et al. [12] published a review article evaluating a set of 36 series from 1950 to 2014. They found 162 patients with OS > 10 years, with the longest reported survival with GBM being 34 years. Progression-free survival (PFS) was a very important factor in their analysis, enabling VLTS. However, the authors note the unpredictability of this disease, as they recorded two patients with only 3 months of PFS who achieved OS of 13 and 21 years.

In the Czech Republic, we would like to draw attention to a case from 1961, when a 10-year-old boy diagnosed with GBM (recently verified by a second reading, fortunately very rare in childhood with an incidence of 0.8 per 100,000 under the age of 19) underwent surgery at the neurosurgery department in Pilsen. It was an expansive process located in the posterior part of the basal temporal lobe, which was imaged using angiography. The patient had bilateral optic nerve papilla atrophy at admission, so he was almost blind his entire life. His OS to date is 64 years (personal communication and consent –⁠ Doc. Jan Mraček, Ph.D.). According to the new classification, this is a diffuse high-grade glioma of the pediatric type (pHGG), H3-wildtype and IDHwt, with the most favorable variant being pHGG RTK2 with EGFR amplification and TERT promoter mutation [13].

Regarding the Ki-67 proliferation index in repeated resection in our 12 patients undergoing reoperation, Ki-67 decreased in eight (67%) of them (the trend is unknown in one). This decrease in six of them predicted further survival longer than 12 months after the last reoperation. The prognostic value of Ki-67 in the first resected histological material was also evaluated by Indian authors [14], who in their cohort of 129 patients identified 14 patients with OS > 3 years (10.8%), where the median Ki-67 index was 15% in LTS compared to 25% in the STS group. This value is very low compared to our group of 20 LTS patients (the value is not available for one patient). In our patients, the average Ki-67 value in the primary sample was 35% (15–70%).

In conclusion to the first part of the discussion, we cannot overlook the results of an analysis of the US National Cancer Database (NCDB) from 2004–2013 from the Mayo Clinic in Rochester [15]. Of 88,919 patients with GBM, 8,757 (9.8%) survived for more than 3 years. This analysis represents the most robust cohort of patients we have recorded. In addition to the usual favorable prognostic signs (younger age, fewer comorbidities, and MGMTp hypermethylation), another favorable factor for LTS was private health insurance (64.5% vs. 42.6%) and the associated availability of trimodal therapy. Fortunately, this is a factor that patients and doctors in our healthcare system do not yet have to deal with. In any case, even in this extensive database, there has been a gradual increase in 3-year OS of 2.5% between 2004 and 2013, which the authors explain by the increase in the frequency of trimodal treatment (especially chemotherapy by 17.3%, followed by radiotherapy by 5% and surgery by 4.8%).

However, during further literature research, we noted some unusual markers of longer survival related to the initial MR image of the tumor. This was the described dependence of prognosis on initial midline pressure in the work of Johannes Wach et al. [16]. If the septum pellucidum deviation was ≥ 1 cm, it was a statistically significant adverse prognostic sign for patient OS. The authors from Bonn observed a median OS of 18 months in patients with "small pressure" compared to 9 months in the group of patients with "large pressure." We cannot unequivocally confirm this conclusion, as we found seven patients (33%) with a preoperative midline shift ≥ 10 mm and OS > 3 years, who achieved a median OS of 94 months (7.8 years). A total of 14 patients (67%) underwent initial surgery with "small overpressure" and achieved a median OS of 79 months (6.6 years). In our cohort, 1/3 of patients initially had a tumor causing significant midline shift (Fig. 4). We see a possible bias in the cited cohort of German authors in the disproportion between the two groups, with "small overpressure" present in 173 patients (87%) at baseline and "large overpressure" in only 25 patients (13%). However, the initial clinical status (CPS) in both Bonn groups was similar, and the authors partially explain the negative effect of an initial midline shift ≥ 10 mm on OS by a more significant gradual decrease in CPS (< 70%) in this group, immediately after discharge and then at 3 and 6 months of follow-up. It can therefore be said that the initial midline shift also increased the risk of resection. At the same time, it should be noted that the comparison with our cohort is subject to selection bias, as our cohort only included patients with a survival of at least 3 years. In the discussion, the authors of the above-mentioned study draw attention to an interesting hypothesis by radiologists from Cleveland [17], who consider the lateralization of glioblastoma in the right hemisphere with an increase in Mass Effect Deformation Heterogeneity (MEDH) in eloquent structures to be an unfavorable prognostic sign. This results in longer asymptomatic tumor growth in the non-dominant hemisphere with subsequent higher compression of eloquent structures. Indeed, 67% of our patients with OS > 3 years had a tumor located in the left (dominant) hemisphere. Another negative prognostic sign according to the literature may be involvement of the periventricular area. The work of Sebastian Adeberg et al. from Heidelberg [18] compares tumor localization in 50 patients with OS > 3 years (LTS) and 50 patients with OS < 1 year (short-term survivors; STS). They evaluated whether the lesion reached the cortical surface or not, and conversely, whether the deep edge of the enhancement extended ≤ 10 mm to the edge of the ventricular system or not. The tumor in 44 STS (88%) and 33 LTS (68.8%) reached ≤ 10 mm to the edge of the ventricular system, and conversely, in 21 LTS (43.8%) the tumor reached the cortex compared to 16 STS (32%). In our cohort, 12 patients (57%) had a tumor ≤ 10 mm from the edge of the ventricular system; however, this was not only the frontal horn of the lateral ventricle, but in seven patients it was the ventricle atrium or temporal horn. Therefore, we cannot unequivocally support either of the above ideas, except for the fact that in all our patients with OS > 3 years, glioblastoma also infiltrated the surface of the hemisphere cortex, although in 24% (5/21) not on its convexity, and only in one patient (4.7%) did it infiltrate the deep structures of the brain –⁠ the basal ganglia. This patient with a very long survival time (11.8 years) demonstrates the unpredictability of the behavior of treated GBM (Fig. 5). However, we consider the absence of cortical infiltration to be unusual in our routine practice, so we verified this possibility in our cohort of 124 GBM patients operated on between 2015 and 2018 (over 4 years), and only in three patients before the first GBM surgery did the enhancing part of the tumor not reach the surface of the cerebral cortex (2.4%). Another study presenting unusual results is a study by French authors [19], who evaluated the values of a simple blood count (BC) before surgery in a cohort of 85 patients. A higher neutrophil-to-lymphocyte ratio (NLR; NLR > 2.42), thrombocytosis (tr > 236×10⁹/l), and a lower erythrocyte count (ery < 4.59×10¹²/l) were found to be prognostically unfavorable and statistically independent factors. The problem was the exclusion of a relatively large number of patients who were already receiving corticosteroids at the time of the KO examination. In our 11 surviving patients, we observed prognostically "good values" in the initial KO in only four cases (36%); in one patient, we were unable to find the values (he was examined at another hospital), but in six patients, the initial values of erythrocytes/thrombocytes/both elements were "unfavorable" (in two patients each). A more interesting variable may be the NLR (neutrophil-to-lymphocyte ratio). The study mentions in its discussion the consistency of its results with 11 other studies, which also explain the adverse effect of a high neutrophil count in the blood count. The authors explain that this ratio is also present locally in the GBM area (high tumor infiltration of neutrophils compared to low incidence of CD3-positive T lymphocytes) and that these neutrophils contribute to increased proliferation of glioma stem cells (GSCs) through upregulation of S100A4 expression. Furthermore, neurotrophil-induced ferroptosis (accumulation of peroxidized lipids) causes necrosis within GBM and is thus associated with increased infiltrative invasiveness of the tumor.

The limitations of our data set include the relatively small number of patients included and the indisputable fact that it does not include all patients with OS > 3 years who underwent surgery at our facility. We are aware of one patient who, despite surviving for more than 4 years, refused clinical and imaging follow-up and, moreover, underwent surgery without subsequent radiochemotherapy.

 

Conclusion

In our work, we draw attention to the need for long-term monitoring of survival even in patients with such an unfavorable oncological diagnosis as glioblastoma (GBM IDHwt). We recorded 21 patients with survival longer than 3 years, the median current survival without including the survival estimate of 11 living patients according to Kaplan-Meier is 53 months (4.4 years) ranging from 38 months to 228 months (19 years). In the analysis of the entire patient cohort (including censoring of surviving patients), we observed a median overall survival of 142 months (11.8 years) at a median follow-up of 74 months, with a 95% confidence interval for the median survival time ranging from 46 to 142 months (not reached). According to retrospective studies reported in the literature, survival longer than 3 years cannot be predicted except in cases of younger age, good initial clinical condition, and MGMTp hypermethylation. Based on our experience and a follow-up cohort of 21 patients with IDHwt GBM and OS > 3 years, achieving maximum radical resection is appropriate for expressing hope for a better prognosis in a given patient. Radical resection was achieved in 19 of 21 patients (90%), but in some cases it was not the first resection. It is essential to complete radiochemotherapy according to the Stupp protocol. Regarding the number of procedures and modalities of further treatment, especially the continuation of chemotherapy, it is not entirely clear whether this treatment directly leads to long-term survival or whether these patients have a specific subtype of glioblastoma with a better prognosis. Based on our work, we cannot confirm the adverse effect of the initial characteristics of the MR image of the tumor –⁠ initial midline shift and periventricular involvement by the tumor. The initial blood count at the time of detection of suspected GBM, and in particular the determination of NLR, can be easily included in the examination protocol.

 

Ethical aspects

The work was carried out in accordance with the Helsinki Declaration of 1975 and its revisions in 2004 and 2008. The work was approved by the Ethics Committee of Masaryk Hospital on September 23, 2016.

 

Grant support

The work was supported by the Ministry of Health of the Czech Republic –⁠ RVO (MOU, 00209805) and by AZV ČR grants No. NU23-08-00307 and NU23-03-00100.

 

Conflict of interest

The authors declare that they have no conflict of interest in relation to the subject of the study.

 

Table 1. Characteristics of the sample.

 

Characteristics of the sample (n = 22)
Age (years)
Median	58
range	35–72
gender
men	8 (38%)
IDH1 status
wild type	21 (95%)
focally positive	1 (5%)
number of operations
1	9 (41%)
2	9 (41%)
≥ 3	4 (18%)
greatest radicalism
radical	19 (86%)
partial	3 (14%)
completion of conventional EBRT 60Gy	22 (100%)
Completion of 6 cycles of temozolomide chemotherapy	21
Alternative chemotherapy (lomustine, CCNU)	2
Completion of 6 cycles of chemotherapy	22
Number of chemotherapy cycles	6–100 (average 24 cycles)
gamma knife boost	13 (59%), 1–3 procedures

 

CCNU –⁠ ceenu lomustine; n –⁠ number