IDEAL monitoring of musical skills during awake craniotomy: From step 1 to step 2

The aim of awake brain surgery is to perform a maximum resection on the one hand, and to preserve cognitive functions, quality of life and personal autonomy on the other hand. Historically, language and sensorimotor functions were most frequently monitored. Over the years other cognitive functions, including music, have entered the operation theatre. Cases about monitoring musical abilities during awake brain surgery are emerging, and a systematic method how to monitor music would be the next step. According to the IDEAL framework for surgical innovations our study aims to present future recommendation based on a systematic literature search (PRISMA) in combination with lessons learned from three case reports from our own clinical practice with professional musicians (n = 3). We plead for structured procedures including individual tailored tasks. By embracing these recommendations, we can both improve clinical care and unravel music functions in the brain.

NnU-Net versus mesh growing algorithm as a tool for the robust and timely segmentation of neurosurgical 3D images in contrast-enhanced T1 MRI scans

PURPOSE

This study evaluates the nnU-Net for segmenting brain, skin, tumors, and ventricles in contrast-enhanced T1 (T1CE) images, benchmarking it against an established mesh growing algorithm (MGA).

METHODS

We used 67 retrospectively collected annotated single-center T1CE brain scans for training models for brain, skin, tumor, and ventricle segmentation. An additional 32 scans from two centers were used test performance compared to that of the MGA. The performance was measured using the Dice-Sørensen coefficient (DSC), intersection over union (IoU), 95th percentile Hausdorff distance (HD95), and average symmetric surface distance (ASSD) metrics, with time to segment also compared.

RESULTS

The nnU-Net models significantly outperformed the MGA (p < 0.0125) with a median brain segmentation DSC of 0.971 [95CI: 0.945-0.979], skin: 0.997 [95CI: 0.984-0.999], tumor: 0.926 [95CI: 0.508-0.968], and ventricles: 0.910 [95CI: 0.812-0.968]. Compared to the MGA's median DSC for brain: 0.936 [95CI: 0.890, 0.958], skin: 0.991 [95CI: 0.964, 0.996], tumor: 0.723 [95CI: 0.000-0.926], and ventricles: 0.856 [95CI: 0.216-0.916]. NnU-Net performance between centers did not significantly differ except for the skin segmentations Additionally, the nnU-Net models were faster (mean: 1139 s [95CI: 685.0-1616]) than the MGA (mean: 2851 s [95CI: 1482-6246]).

CONCLUSIONS

The nnU-Net is a fast, reliable tool for creating automatic deep learning-based segmentation pipelines, reducing the need for extensive manual tuning and iteration. The models are able to achieve this performance despite a modestly sized training set. The ability to create high-quality segmentations in a short timespan can prove invaluable in neurosurgical settings.

Longitudinal characteristics of T2-FLAIR mismatch in IDH-mutant astrocytomas: Relation to grade, histopathology, and overall survival in the GLASS-NL cohort

Background

The T2-FLAIR mismatch sign is defined by signal loss of the T2-weighted hyperintense area with Fluid-Attenuated Inversion Recovery (FLAIR) on magnetic resonance imaging, causing a hypointense region on FLAIR. It is a highly specific diagnostic marker for IDH-mutant astrocytoma and is postulated to be caused by intercellular microcystic change in the tumor tissue. However, not all IDH-mutant astrocytomas show this mismatch sign and some show the phenomenon in only part of the lesion. The aim of the study is to determine whether the T2-FLAIR mismatch phenomenon has any prognostic value beyond initial noninvasive molecular diagnosis.

Methods

Patients initially diagnosed with histologically lower-grade (2 or 3) IDH-mutant astrocytoma and with at least 2 surgical resections were included in the GLASS-NL cohort. T2-FLAIR mismatch was determined, and the growth pattern of the recurrent tumor immediately before the second resection was annotated as invasive or expansive. The relation between the T2-FLAIR mismatch sign and tumor grade, microcystic change, overall survival (OS), and other clinical parameters was investigated both at first and second resection.

Results

The T2-FLAIR mismatch sign was significantly related to Grade 2 (80% vs 51%), longer post-resection median OS (8.3 vs 5.2 years), expansive growth, and lower age at second resection. At first resection, no relation was found between the mismatch sign and OS. Microcystic change was associated with areas of T2-FLAIR mismatch.

Conclusions

T2-FLAIR mismatch in IDH-mutant astrocytomas is correlated with microcystic change in the tumor tissue, favorable prognosis, and Grade 2 tumors at the time of second resection.

The impact of etiology in lesion-symptom mapping - A direct comparison between tumor and stroke

INTRODUCTION

Lesion-symptom mapping is a key tool in understanding the relationship between brain structures and behavior. However, the behavioral consequences of lesions from different etiologies may vary because of how they affect brain tissue and how they are distributed. The inclusion of different etiologies would increase the statistical power but has been critically debated. Meanwhile, findings from lesion studies are a valuable resource for clinicians and used across different etiologies. Therefore, the main objective of the present study was to directly compare lesion-symptom maps for memory and language functions from two populations, a tumor versus a stroke population.

METHODS

Data from two different studies were combined. Both the brain tumor (N = 196) and stroke (N = 147) patient populations underwent neuropsychological testing and an MRI, pre-operatively for the tumor population and within three months after stroke. For this study, we selected two internationally widely used standardized cognitive tasks, the Rey Auditory Verbal Learning Test and the Verbal Fluency Test. We used a state-of-the-art machine learning-based, multivariate voxel-wise approach to produce lesion-symptom maps for these cognitive tasks for both populations separately and combined.

RESULTS

Our lesion-symptom mapping results for the separate patient populations largely followed the expected neuroanatomical pattern based on previous literature. Substantial differences in lesion distribution hindered direct comparison. Still, in brain areas with adequate coverage in both groups, considerable LSM differences between the two populations were present for both memory and fluency tasks. Post-hoc analyses of these locations confirmed that the cognitive consequences of focal brain damage varied between etiologies.

CONCLUSION

The differences in the lesion-symptom maps between the stroke and tumor population could partly be explained by differences in lesion volume and topography. Despite these methodological limitations, both the lesion-symptom mapping results and the post-hoc analyses confirmed that etiology matters when investigating the cognitive consequences of lesions with lesion-symptom mapping. Therefore, caution is advised with generalizing lesion-symptom results across etiologies.

Study protocol of the GLOW study: maximising treatment options for recurrent glioblastoma patients by whole genome sequencing-based diagnostics—a prospective multicenter cohort study

Background

Glioblastoma (GBM), the most common glial primary brain tumour, is without exception lethal. Every year approximately 600 patients are diagnosed with this heterogeneous disease in The Netherlands. Despite neurosurgery, chemo -and radiation therapy, these tumours inevitably recur. Currently, there is no gold standard at time of recurrence and treatment options are limited. Unfortunately, the results of dedicated trials with new drugs have been very disappointing. The goal of the project is to obtain the evidence for changing standard of care (SOC) procedures to include whole genome sequencing (WGS) and consequently adapt care guidelines for this specific patient group with very poor prognosis by offering optimal and timely benefit from novel therapies, even in the absence of traditional registration trials for this small volume cancer indication.

Methods

The GLOW study is a prospective diagnostic cohort study executed through collaboration of the Hartwig Medical Foundation (Hartwig, a non-profit organisation) and twelve Dutch centers that perform neurosurgery and/or treat GBM patients. A total of 200 patients with a first recurrence of a glioblastoma will be included. Dual primary endpoint is the percentage of patients who receive targeted therapy based on the WGS report and overall survival. Secondary endpoints include WGS report success rate and number of targeted treatments available based on WGS reports and number of patients starting a treatment in presence of an actionable variant. At recurrence, study participants will undergo SOC neurosurgical resection. Tumour material will then, together with a blood sample, be sent to Hartwig where it will be analysed by WGS. A diagnostic report with therapy guidance, including potential matching off-label drugs and available clinical trials will then be sent back to the treating physician for discussing of the results in molecular tumour boards and targeted treatment decision making.

Discussion

The GLOW study aims to provide the scientific evidence for changing the SOC diagnostics for patients with a recurrent glioblastoma by investigating complete genome diagnostics to maximize treatment options for this patient group.

Trial registration: ClinicalTrials.gov Identifier: NCT05186064.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12920-022-01343-4.

P01.20.B Peroperative cognitive monitoring in patients with aphasia

Background

Cognitive monitoring during awake glioma surgery provides the opportunity to increase the extent of resection while preserving cognitive functions. Preoperative language functioning often determines whether awake surgery can be performed since some form of communication with the patient is essential. However, as complex as it may be to perform intraoperative monitoring in patients with severe preoperative language disturbances, it is risky to assume that language disturbances are definite without functional perspectives. In addition, there are other cognitive functions that should be preserved. This study investigates to what extent cognitive functions can be monitored during awake surgery in the presence of preoperative language disturbances.

Material and Methods

Glioma data were collected in a single-centre consecutive study cohort of 186 patients who underwent awake brain surgery for either low-grade (WHO grade 1-2) glioma (40 percent) or high-grade (WHO grade 3-4) glioma (60 percent). Preoperative language disturbances were evidenced during preoperative neuropsychological assessment based on performances on the Boston Naming Test (N= 27) and Token Test (N=25). Intraoperative procedures were performed as part of clinical care.

Results

Preliminary results on the Boston Naming Test show that intraoperative testing with the use of at least one language test (picture-naming) is feasible in 80 percent of all patients with language disturbances. In those cases, a functional boundary of language functions is found in 85 percent during awake surgery. Monitoring of other non-language functions, such as working memory, inhibition, speech, motor functions and social cognition, provided functional boundaries in 25 up to 85 percent of patients with language disturbances. Postoperatively, cognitive functioning was improved in a significant number of patients.

Conclusion

Based on these preliminary results we conclude that preoperative language disturbances do not prevent from monitoring language functions as well as other cognitive functions during awake glioma surgery. In a larger cohort, more in-depth analyses will be performed on additional outcome measures (Token test, survival rate, post-surgery treatment) and postoperative functioning will be compared to a control group of patients without preoperative language disturbances. By investigating the potential of awake surgery in patients who are severely cognitively affected, we aim to increase the applicability of this method to improve neuro-oncological treatments and optimally maintain quality of life.

P01.04.A Lesion-symptom mapping based on stroke or glioma: etiology matters!

Background

Lesion-symptom mapping is a key tool in understanding the relationship between brain structures and behavior. However, the behavioral consequences of lesions from different etiologies may vary as a result of how they affect brain tissue, and how they are distributed. The inclusion of different etiologies would increase the statistical power and improve generalizability of results, but has been critically debated. Findings from lesion studies are a resource for clinicians and used across different etiologies. This study directly compared lesion-symptom maps (LSM) between two populations (diffuse glioma versus ischemic stroke) in order to investigate if brain areas with adequate coverage in both groups, show topographical overlap in lesion-symptom associations.

Material and Methods

Data from two studies were combined. Both the glioma (N = 196, WHO grade 2-4) and stroke (N = 147) population underwent neuropsychological testing and an MRI, pre-operatively for the tumor population and within three months after stroke. For the purpose of this study, we selected two widely used cognitive tasks, the Rey Auditory Verbal Learning Test and the verbal fluency test. We used a state-of-the-art machine learning-based, multivariate voxel-wise approach to produce LSM for these cognitive tasks for both populations separately.

Results

For both tumor and stroke, our etiology-specific LSM largely followed the expected neuroanatomical pattern based on previous literature. Nevertheless, for both tasks substantial differences in LSM-results were present between the populations in brain areas with adequate coverage in both groups, though we did find similar white matter tracts involved in memory and semantic fluency performance across etiologies. Post-hoc analyses of these locations confirmed an interaction between lesion presence and etiology for a majority of these regions; damage by a tumor, but not a stroke, was related to worse cognitive performance for these regions.

Conclusion

This study provides the first direct comparison of LSM in a large cohort of patients. Differences in LSM were found between the glioma and stroke group, confirming that etiology matters when investigating the cognitive consequence of lesions. These differences could partly be explained by differences in lesion volume and topography. Nonetheless, the pattern shown by glioma patients on the group level is consistent with localizations found in earlier studies on both stroke and glioma patients using different techniques. While glioma series thus can be used to provide converging evidence about functional localization, we do suggest that findings from LSM studies in neuro-oncological populations should be considered as cause-specific. Findings from functional localization research from non-glioma populations should only be applied to a glioma population with caution.

Investigation of glial fibrillary acidic protein (GFAP) in body fluids as a potential biomarker for glioma: a systematic review and meta-analysis

INTRODUCTION

Liquid biopsies are promising diagnostic tools for glioma. In this quantitative systematic review, we investigate whether the detection of intermediate filaments (IF) in body fluids can be used as a tool for glioma diagnosis and prognosis.

MATERIALS AND METHODS

We included all studies in which IF-levels were determined in patients with glioma and healthy controls. Of the 28 identified eligible studies, 12 focused on levels of GFAP in serum (sGFAP) and were included for metadata analysis.

RESULTS

In all studies combined, 62.7% of all grade IV patients had detectable levels of sGFAP compared to 12.7% of healthy controls. sGFAP did not surpass the limit of detection in lower grade patients or healthy controls, but sGFAP was significantly elevated in grade IV glioma (0.12 ng/mL (0.06 - 0.18), P < 0.001) and showed an average median difference of 0.15 ng/mL (0.04 - 0.25, P < 0.01) compared to healthy controls. sGFAP levels were linked to tumour volume, but not to patient outcome.

CONCLUSION

The presence of sGFAP is indicative of grade IV glioma, but additional studies are necessary to fully determine the usefulness of GFAP in body fluids as a tool for grade IV glioma diagnosis and follow-up.

Timing of glioblastoma surgery and patient outcomes: a multicenter cohort study

Background

The impact of time-to-surgery on clinical outcome for patients with glioblastoma has not been determined. Any delay in treatment is perceived as detrimental, but guidelines do not specify acceptable timings. In this study, we relate the time to glioblastoma surgery with the extent of resection and residual tumor volume, performance change, and survival, and we explore the identification of patients for urgent surgery.

Methods

Adults with first-time surgery in 2012–2013 treated by 12 neuro-oncological teams were included in this study. We defined time-to-surgery as the number of days between the diagnostic MR scan and surgery. The relation between time-to-surgery and patient and tumor characteristics was explored in time-to-event analysis and proportional hazard models. Outcome according to time-to-surgery was analyzed by volumetric measurements, changes in performance status, and survival analysis with patient and tumor characteristics as modifiers.

Results

Included were 1033 patients of whom 729 had a resection and 304 a biopsy. The overall median time-to-surgery was 13 days. Surgery was within 3 days for 235 (23%) patients, and within a month for 889 (86%). The median volumetric doubling time was 22 days. Lower performance status (hazard ratio [HR] 0.942, 95% confidence interval [CI] 0.893–0.994) and larger tumor volume (HR 1.012, 95% CI 1.010–1.014) were independently associated with a shorter time-to-surgery. Extent of resection, residual tumor volume, postoperative performance change, and overall survival were not associated with time-to-surgery.

Conclusions

With current decision-making for urgent surgery in selected patients with glioblastoma and surgery typically within 1 month, we found equal extent of resection, residual tumor volume, performance status, and survival after longer times-to-surgery.

Tinnitus treatment by vagus nerve stimulation: A systematic review

Background

Tinnitus is a phantom sensation of sound, which can have a negative impact on quality of life of those affected. No curative treatments are currently known. Neuromodulation by vagus nerve stimulation has emerged as a new treatment option for tinnitus, though till date the effectiveness remains unclear. Therefore, we aim to review the effect of vagus nerve stimulation on tinnitus distress and tinnitus symptom severity in patients with chronic tinnitus.

Methods

We searched Pubmed, Embase and the Cochrane Library systematically for RCTs, observational studies and case studies on the effect of VNS treatment for tinnitus on October 29, 2019. Studies including adult patients with subjective tinnitus, comparing transcutaneous or implantable VNS to placebo or no treatment or before and after application of VNS treatment on tinnitus distress and tinnitus symptom severity measured with a validated questionnaire were eligible. The risk of bias was assessed with the appropriate tool for each type of study.

Results

Our search identified 9 primary studies of which 2 RCTs, 5 cohort studies and 2 case series or reports. 5 studies used transcutaneous VNS treatment and 4 used implanted VNS treatment. 6 studies combined VNS treatment with sound therapy. There was a serious risk of bias in all studies, especially on confounding. Most studies reported a small decrease in tinnitus distress or tinnitus symptom severity.

Conclusion

Due to methodological limitations and low reporting quality of the included studies, the effect of VNS on tinnitus remains unclear. To draw conclusions for which patient population and to what extent (t)VNS is beneficial in the treatment of tinnitus, a randomised controlled trial should be considered.

Robust Deep Learning-based Segmentation of Glioblastoma on Routine Clinical MRI Scans Using Sparsified Training

PURPOSE

To improve the robustness of deep learning-based glioblastoma segmentation in a clinical setting with sparsified datasets.

MATERIALS AND METHODS

In this retrospective study, preoperative T1-weighted, T2-weighted, T2-weighted fluid-attenuated inversion recovery, and postcontrast T1-weighted MRI from 117 patients (median age, 64 years; interquartile range [IQR], 55-73 years; 76 men) included within the Multimodal Brain Tumor Image Segmentation (BraTS) dataset plus a clinical dataset (2012-2013) with similar imaging modalities of 634 patients (median age, 59 years; IQR, 49-69 years; 382 men) with glioblastoma from six hospitals were used. Expert tumor delineations on the postcontrast images were available, but for various clinical datasets, one or more sequences were missing. The convolutional neural network, DeepMedic, was trained on combinations of complete and incomplete data with and without site-specific data. Sparsified training was introduced, which randomly simulated missing sequences during training. The effects of sparsified training and center-specific training were tested using Wilcoxon signed rank tests for paired measurements.

RESULTS

A model trained exclusively on BraTS data reached a median Dice score of 0.81 for segmentation on BraTS test data but only 0.49 on the clinical data. Sparsified training improved performance (adjusted P < .05), even when excluding test data with missing sequences, to median Dice score of 0.67. Inclusion of site-specific data during sparsified training led to higher model performance Dice scores greater than 0.8, on par with a model based on all complete and incomplete data. For the model using BraTS and clinical training data, inclusion of site-specific data or sparsified training was of no consequence.

CONCLUSION

Accurate and automatic segmentation of glioblastoma on clinical scans is feasible using a model based on large, heterogeneous, and partially incomplete datasets. Sparsified training may boost the performance of a smaller model based on public and site-specific data.Supplemental material is available for this article.Published under a CC BY 4.0 license.

Comparing Glioblastoma Surgery Decisions Between Teams Using Brain Maps of Tumor Locations, Biopsies, and Resections

Purpose

The aim of glioblastoma surgery is to maximize the extent of resection while preserving functional integrity, which depends on the location within the brain. A standard to compare these decisions is lacking. We present a volumetric voxel-wise method for direct comparison between two multidisciplinary teams of glioblastoma surgery decisions throughout the brain.

Methods

Adults undergoing first-time glioblastoma surgery from 2012 to 2013 performed by two neuro-oncologic teams were included. Patients had had a diagnostic biopsy or resection. Preoperative tumors and postoperative residues were segmented on magnetic resonance imaging in three dimensions and registered to standard brain space. Voxel-wise probability maps of tumor location, biopsy, and resection were constructed for each team to compare patient referral bias, indication variation, and treatment variation. To evaluate the quality of care, subgroups of differentially resected brain regions were analyzed for survival and functional outcome.

Results

One team included 101 patients, and the other included 174; 91 tumors were biopsied, and 181 were resected. Patient characteristics were largely comparable between teams. Distributions of tumor locations were dissimilar, suggesting referral bias. Distributions of biopsies were similar, suggesting absence of indication variation. Differentially resected regions were identified in the anterior limb of the right internal capsule and the right caudate nucleus, indicating treatment variation. Patients with (n = 12) and without (n = 6) surgical removal in these regions had similar overall survival and similar permanent neurologic deficits.

Conclusion

Probability maps of tumor location, biopsy, and resection provide additional information that can inform surgical decision making across multidisciplinary teams for patients with glioblastoma.

Radiological differences between subtypes of WHO 2016 grade II-III gliomas: a systematic review and meta-analysis

BACKGROUND

Isocitrate dehydrogenase (IDH) mutation and 1p/19q-codeletion are oncogenetic alterations with a positive prognostic value for diffuse gliomas, especially grade II and III. Some studies have suggested differences in biological behavior as reflected by radiological characteristics. In this paper, the literature regarding radiological characteristics in grade II and III glioma subtypes was systematically evaluated and a meta-analysis was performed.

METHODS

Studies that addressed the relationship between conventional radiological characteristics and IDH mutations and/or 1p/19q-codeletions in newly diagnosed, grade II and III gliomas of adult patients were included. The "3-group analysis" compared radiological characteristics between the WHO 2016 glioma subtypes (IDH-mutant astrocytoma, IDH-wildtype astrocytoma, and oligodendroglioma), and the "2-group analysis" compared radiological characteristics between 1p/19q-codeleted gliomas and 1p/19q-intact gliomas.

RESULTS

Fourteen studies (3-group analysis: 670 cases, 2-group analysis: 1042 cases) were included. IDH-mutated astrocytomas showed more often sharp borders and less frequently contrast enhancement compared to IDH-wildtype astrocytomas. 1p/19q-codeleted gliomas had less frequently sharp borders, but showed a heterogeneous aspect, calcification, cysts, and edema more frequently. For the 1p/19q-codeleted gliomas, a sensitivity of 96% was found for heterogeneity and a specificity of 88.1% for calcification.

CONCLUSIONS

Significant differences in conventional radiological characteristics exist between the WHO 2016 glioma subtypes, which may reflect differences in biological behavior. However, the diagnostic value of the independent radiological characteristics is insufficient to reliably predict the molecular genetic subtype.

Interrelationships between molecular subtype, anatomical location, and extent of resection in diffuse glioma: a systematic review and meta-analysis

Background

The introduction of the 2016 WHO Classification of Tumors of the Central Nervous System has resulted in tumor groupings with improved prognostic value for diffuse glioma patients. Molecular subtype, primarily based on IDH-mutational status and 1p/19q-status, is a strong predictor of survival. It is unclear to what extent this finding may be mediated by differences in anatomical location and surgical resectability among molecular subgroups. Our aim was to elucidate possible correlations between (1) molecular subtype and anatomical location and (2) molecular subtype and extent of resection.

Methods

We performed a systematic review of literature searching for studies on molecular subtype in relation to anatomical location and extent of resection. Only original data concerning adult participants suffering from cerebral diffuse glioma were included. Studies adopting similar outcomes measures were included in our meta-analysis.

Results

In the systematic analysis for research questions 1 and 2, totals of 20 and 9 studies were included, respectively. Study findings demonstrated that IDH-mutant tumors were significantly more frequently located in the frontal lobe and less often in the temporal lobe compared with IDH-wildtype gliomas. Within the IDH-mutant group, 1p/19q-codeleted tumors were associated with more frequent frontal and less frequent temporal localization compared with 1p/19q-intact tumors. In IDH-mutant gliomas, greater extent of resection was achieved than in IDH-wildtype tumors.

Conclusions

Genetic profile of diffuse cerebral glioma influences their anatomical location and seems to affect tumor resectability.

Between-hospital variation in mortality and survival after glioblastoma surgery in the Dutch Quality Registry for Neuro Surgery

Purpose

Standards for surgical decisions are unavailable, hence treatment decisions can be personalized, but also introduce variation in treatment and outcome. National registrations seek to monitor healthcare quality. The goal of the study is to measure between-hospital variation in risk-standardized survival outcome after glioblastoma surgery and to explore the association between survival and hospital characteristics in conjunction with patient-related risk factors.

Methods

Data of 2,409 adults with first-time glioblastoma surgery at 14 hospitals were obtained from a comprehensive, prospective population-based Quality Registry Neuro Surgery in The Netherlands between 2011 and 2014. We compared the observed survival with patient-specific risk-standardized expected early (30-day) mortality and late (2-year) survival, based on age, performance, and treatment year. We analyzed funnel plots, logistic regression and proportional hazards models.

Results

Overall 30-day mortality was 5.2% and overall 2-year survival was 13.5%. Median survival varied between 4.8 and 14.9 months among hospitals, and biopsy percentages ranged between 16 and 73%. One hospital had lower than expected early mortality, and four hospitals had lower than expected late survival. Higher case volume was related with lower early mortality (P = 0.031). Patient-related risk factors (lower age; better performance; more recent years of treatment) were significantly associated with longer overall survival. Of the hospital characteristics, longer overall survival was associated with lower biopsy percentage (HR 2.09, 1.34–3.26, P = 0.001), and not with academic setting, nor with case volume.

Conclusions

Hospitals vary more in late survival than early mortality after glioblastoma surgery. Widely varying biopsy percentages indicate treatment variation. Patient-related factors have a stronger association with overall survival than hospital-related factors.

Electronic supplementary material

The online version of this article (10.1007/s11060-019-03229-5) contains supplementary material, which is available to authorized users.