Deep neural networks allow expert-level brain meningioma segmentation and present potential for improvement of clinical practice

Accurate brain meningioma segmentation and volumetric assessment are critical for serial patient follow-up, surgical planning and monitoring response to treatment. Current gold standard of manual labeling is a time-consuming process, subject to inter-user variability. Fully-automated algorithms for meningioma segmentation have the potential to bring volumetric analysis into clinical and research workflows by increasing accuracy and efficiency, reducing inter-user variability and saving time. Previous research has focused solely on segmentation tasks without assessment of impact and usability of deep learning solutions in clinical practice. Herein, we demonstrate a three-dimensional convolutional neural network (3D-CNN) that performs expert-level, automated meningioma segmentation and volume estimation on MRI scans. A 3D-CNN was initially trained by segmenting entire brain volumes using a dataset of 10,099 healthy brain MRIs. Using transfer learning, the network was then specifically trained on meningioma segmentation using 806 expert-labeled MRIs. The final model achieved a median performance of 88.2% reaching the spectrum of current inter-expert variability (82.6–91.6%). We demonstrate in a simulated clinical scenario that a deep learning approach to meningioma segmentation is feasible, highly accurate and has the potential to improve current clinical practice.

Survival Prediction After Neurosurgical Resection of Brain Metastases: A Machine Learning Approach

BACKGROUND

Current prognostic models for brain metastases (BMs) have been constructed and validated almost entirely with data from patients receiving up-front radiotherapy, leaving uncertainty about surgical patients.

OBJECTIVE

To build and validate a model predicting 6-month survival after BM resection using different machine learning algorithms.

METHODS

An institutional database of 1062 patients who underwent resection for BM was split into an 80:20 training and testing set. Seven different machine learning algorithms were trained and assessed for performance; an established prognostic model for patients with BM undergoing radiotherapy, the diagnosis-specific graded prognostic assessment, was also evaluated. Model performance was assessed using area under the curve (AUC) and calibration.

RESULTS

The logistic regression showed the best performance with an AUC of 0.71 in the hold-out test set, a calibration slope of 0.76, and a calibration intercept of 0.03. The diagnosis-specific graded prognostic assessment had an AUC of 0.66. Patients were stratified into regular-risk, high-risk and very high-risk groups for death at 6 months; these strata strongly predicted both 6-month and longitudinal overall survival ( P < .0005). The model was implemented into a web application that can be accessed through http://brainmets.morethanml.com .

CONCLUSION

We developed and internally validated a prediction model that accurately predicts 6-month survival after neurosurgical resection for BM and allows for meaningful risk stratification. Future efforts should focus on external validation of our model.

Clinical utility of targeted next-generation sequencing assay in IDH-wildtype glioblastoma for therapy decision-making

BACKGROUND

Targeted gene NGS testing is available through many academic institutions and commercial entities and is increasingly incorporated in practice guidelines for glioblastoma (GBM). This single-center retrospective study aimed to evaluate the clinical utility of incorporating NGS results in the management of GBM patients at a clinical trials-focused academic center.

METHODS

We identified 1011 consecutive adult patients with pathologically confirmed GBM (IDHwt or IDHmut) who had somatic tumor sequencing (Oncopanel, ~500 cancer gene panel) at DFCI from 2013-2019. Clinical records of all IDHwt GBM patients were reviewed to capture clinical trial enrollment and off-label targeted therapy use based on NGS results.

RESULTS

Of the 557 IDHwt GBM patients with sequencing, 182 entered clinical trials at diagnosis (32.7%) and 213 (38.2%) entered after recurrence. Sequencing results for 130 patients (23.3%) were utilized for clinical trial enrollment for either targeted therapy indications (6.9 % upfront and 27.7% at recurrent clinical trials and 3.1% for off-label targeted therapy) or exploratory studies (55.4% upfront and 6.9% recurrent clinical trials). Median overall survival was 20.1 months with no survival difference seen between patients enrolled in clinical trials compared to those who were not, in a posthoc analysis.

CONCLUSIONS

While NGS testing has become essential for improved molecular diagnostics, our study illustrates that targeted gene panels remain underutilized for selecting therapy in GBM-IDHwt. Targeted therapy and clinical trial design remain to be improved to help leverage the potential of NGS in clinical care.

Intracranial complications of hypercoagulability and superinfection in the setting of COVID-19: illustrative cases

BACKGROUND

Hypercoagulability with thrombosis and associated inflammation has been well-documented in COVID-19, and catastrophic cerebral venous sinus thromboses (CVSTs) have been described. Another COVID-19–related complication is bacterial superinfection, including sinusitis. Here, the authors reported three cases of COVID-19–associated sinusitis, meningitis, and CVST and summarized the literature about septic intracranial thrombotic events as a cause of headache and fever in COVID-19.

OBSERVATIONS

The authors described three adolescent patients with no pertinent past medical history and no prior COVID-19 vaccinations who presented with subacute headaches, photosensitivity, nausea, and vomiting after testing positive for COVID-19. Imaging showed subdural collections, CVST, cerebral edema, and severe sinus disease. Two patients had decline in mental status and progression of neurological symptoms. In all three, emergency cranial and sinonasal washouts uncovered pus that grew polymicrobial cultures. After receiving broad-spectrum antimicrobials and various additional treatments, including two of three patients receiving anticoagulation, all patients eventually became neurologically intact with varying ongoing sequelae.

LESSONS

These cases demonstrated similar original presentations among previously healthy adolescents with COVID-19 infections, concurrent sinusitis precipitating CVST, and subdural empyemas. Better recognition and understanding of the multisystem results of severe acute respiratory syndrome coronavirus 2 and the complicated sequelae allows for proper treatment.

Tumor-Associated Macrophages/Microglia in Glioblastoma Oncolytic Virotherapy: A Double-Edged Sword

Oncolytic virotherapy is a rapidly progressing field that uses oncolytic viruses (OVs) to selectively infect malignant cells and cause an antitumor response through direct oncolysis and stimulation of the immune system. Despite demonstrated pre-clinical efficacy of OVs in many cancer types and some favorable clinical results in glioblastoma (GBM) trials, durable increases in overall survival have remained elusive. Recent evidence has emerged that tumor-associated macrophage/microglia (TAM) involvement is likely an important factor contributing to OV treatment failure. It is prudent to note that the relationship between TAMs and OV therapy failures is complex. Canonically activated TAMs (i.e., M1) drive an antitumor response while also inhibiting OV replication and spread. Meanwhile, M2 activated TAMs facilitate an immunosuppressive microenvironment thereby indirectly promoting tumor growth. In this focused review, we discuss the complicated interplay between TAMs and OV therapies in GBM. We review past studies that aimed to maximize effectiveness through immune system modulation-both immunostimulatory and immunosuppressant-and suggest future directions to maximize OV efficacy.

BIOM-30. BLOOD COUNTS THROUGH TREATMENT COURSE IN NEWLY DIAGNOSED GLIOBLASTOMA PATIENTS RECEIVING CHEMORADIATION

BACKGROUND

Glioblastoma (GBM) patients are treated with radiation therapy (RT), temozolomide, and corticosteroids which can affect hematologic and immunologic parameters. We examined lymphocytes, neutrophil-to-lymphocyte ratio and platelet measurements and their association with progression-free survival (PFS) overall survival (OS).

METHODS

We identified 759 newly diagnosed adult GBM patients treated at our institution in the temozolomide (TMZ) era with blood counts that could be automatically extracted from the electronic medical record during chemoradiation (CRT, defined as within 42 days of RT) and at first recurrence. Linear regression and Cox modeling were used to evaluate outcomes.

RESULTS

Median age was 60.3 years; 87% had KPS ≥ 70, 37.5% had gross total resection, and 90% received TMZ. Prior to RT, 56.4% (375/665) patients had a lymphocyte measurement &lt; 1.0 × 1000 cells [K]/μL. Within 42 days of CRT, 81.7% (536/656) had a lymphocyte measurement &lt; 1.0 K/μL, 37.8% (248/656) &lt; 0.5 K/μL. 10.7% (58/544) patients developed grade 2 or higher neutropenia, 9.1% (50/547) patients developed grade 2 or higher thrombocytopenia. On multivariable analysis (MVA), older age (AHR1.03, p&lt; 0.001), unmethylated MGMT status (AHR2.56,p&lt; 0.001), lower RT dose (&lt;54Gy, AHR 3.45, p&lt; 0.001), male sex (AHR1.45, p=0.02), non-gross total resection (AHR1.63, p&lt; 0.001), lymphopenia during CRT (AHR0.63, p=0.008) and higher NLR during CRT (AHR1.02, p=0.001) were significantly associated with worse OS.

Older age (AHR1.01, p=0.02), unmethylated MGMT status (AHR2.44, p&lt; 0.001), lower RT dose (AHR1.82, p=0.02), higher NLR during CRT (AHR1.03, p &lt; 0.001) were significantly associated with worse PFS on MVA. At first recurrence, median lymphocyte count was 0.7 K/μL with 74% (348/468) patients &lt; 1.0 K/μL and 27% &lt; 0.5 K/μL.

CONCLUSION

Lymphopenia and higher neutrophil-to-lymphocyte ratio are associated with inferior outcomes. Persistent lymphopenia at time of first recurrence may have implications for clinical trial eligibility and immunotherapy approaches in recurrent GBM.

CTNI-40. EVALUATING FEASIBILITY AND EFFICIENCY OF PHASE II ADAPTIVE PLATFORM TRIAL DESIGNS BASED ON THE INDIVIDUALIZED SCREENING TRIAL OF INNOVATIVE GLIOBLASTOMA THERAPY (INSIGhT) EXPERIENCE

BACKGROUND

The Individualized Screening Trial of Innovative Glioblastoma Therapy (INSIGhT) is a phase II platform trial with Bayesian adaptive randomization and deep genomic profiling to more efficiently test experimental agents in newly diagnosed glioblastoma and to prioritize therapies for late-stage testing.

METHODS

In the ongoing INSIGhT trial, patients with newly diagnosed MGMT-unmethylated glioblastoma are randomized to the control arm or one of three experimental therapy arms (CC-115, abemaciclib, and neratinib). The control arm therapy is radiotherapy with concomitant and adjuvant temozolomide, and primary endpoint is overall survival. Randomization has been adapted based on Bayesian estimation of biomarker-specific probability of treatment impact on progression-free survival (PFS). All tumors undergo detailed molecular sequencing, and this is facilitated with the companion ALLELE protocol. To evaluate feasibility of this approach, we assessed the status of this ongoing trial.

RESULTS

Since INSIGhT was activated 4.3 years ago, it has expanded to include 12 sites across the United States. A total of 247 patients have been enrolled. Randomization probabilities have been repeatedly adjusted over time based upon early PFS results to alter the randomization ratio from standard 1:1:1:1 randomization. All three arms have completed accrual and efficacy estimates are available based upon comparison to the common control arm in context of relevant biomarkers. There are 87 patients alive and in follow-up, and there are ongoing plans to add additional arms to evaluate further treatments in the future.

CONCLUSION

The INSIGhT trial demonstrates that a multi-center Bayesian adaptive platform trial is a feasible and effective approach to help prioritize therapies and biomarkers for newly diagnosed GBM. The trial has maintained robust accrual, and the simultaneous testing of multiple agents, sharing a common control arm and adaptive randomization serve as features to increase trial efficiency relative to traditional clinical trial designs.

NIMG-24. RANO CRITERIA DETECTS EARLY PROGRESSION SOONER THAN MODIFIED RANO CRITERIA IN PATIENTS WITH NEWLY DIAGNOSED GLIOBLASTOMA

BACKGROUND

Accurate response criteria are crucial for determining treatment efficacy. The response assessment in neuro-oncology (RANO) criteria was developed to standardize response assessment in neuro-oncology. Modified RANO (m-RANO) criteria were recently proposed to address some limitations of the initial criteria including the use of a post-radiation baseline and an additional scan to confirm progression. We sought to identify differences in the association of progression-free survival (PFS) and overall survival (OS) using RANO and m-RANO criteria.

METHODS

We conducted a retrospective review of newly diagnosed glioblastoma (GBM) patients treated at Dana-Farber Cancer Institute from January 2013 until December 2019. Patients with available clinical and imaging data obtained before initiation of treatment, after radiation completion and at intervals of 1 to 3 months were included. MRIs were evaluated by two independent readers, and PD dates determined using RANO and m-RANO criteria.

RESULTS

552 patients were included. 97.1% of the tumors were IDH wild-type. MGMT promoter was unmethylated in 51.4%, methylated in 35.1% and undetermined in 8.5%. Median OS among patients was 18.1 months. 72 patients (13%) did not have PD at the end of the study. 83 patients had treatment change while being clinically stable and without a confirmation scan and were excluded from the final analysis. PFS was 8.2 months with RANO and 8.4 months with mRANO. Difference in PD dates between RANO and m-RANO was detected in 76 patients (14%), where PFS was 3.5 months with RANO and 5.1 months with m-RANO. These patients had a worse median OS than those with identical RANO and m-RANO PD dates (15.2 vs. 22.4 months, p&lt; 0.0001).

CONCLUSION

RANO and m-RANO criteria resulted in identical PFS for most patients. 14% of patients had discordant PD dates and a worse prognosis. These patients progressed early, and their PD was identified sooner with RANO criteria.

CTNI-05. PRELIMINARY RESULTS OF THE NERATINIB ARM IN THE INDIVIDUALIZED SCREENING TRIAL OF INNOVATIVE GLIOBLASTOMA THERAPY (INSIGHT): A PHASE II PLATFORM TRIAL USING BAYESIAN ADAPTIVE RANDOMIZATION

BACKGROUND

EGFR is amplified in over 50% of glioblastoma and 20-30% have EGFRvIII mutations. Neratinib is a potent inhibitor of EGFR/HER2 approved for metastatic HER2+ breast cancer. To efficiently evaluate the potential impact of neratinib on overall survival (OS) in newly-diagnosed glioblastoma and to simultaneously develop information regarding potential genomic biomarker associations, neratinib was included as an arm on the Individualized Screening Trial of Innovative Glioblastoma Therapy (INSIGhT) trial. INSIGhT is a phase II platform trial using response adaptive randomization and deep genomic profiling to more efficiently test experimental agents in MGMT unmethylated glioblastoma and accelerate identification of novel therapies for phase III testing. Initial randomization was equal between neratinib, control, and two other experimental arms but subsequent randomization was adapted based on efficacy as determined by progression-free survival (PFS). We report preliminary results for the neratinib arm.

METHODS

Patients with newly diagnosed MGMT-unmethylated glioblastoma were randomized to receive either radiotherapy with concomitant and adjuvant temozolomide or standard radiochemotherapy followed by adjuvant neratinib (240 mg daily). Treatment continued until progression or development of unacceptable toxicities. The primary endpoint was OS. Association between neratinib efficacy and EGFR amplification was also investigated.

RESULTS

There were 144 patients (70 control; 74 neratinib). Neratinib was reasonably well-tolerated with no new toxicity signals identified. PFS was compared (HR 0.84; p=0.38, logrank test – not significant) between the neratinib (median 6.05 months) and control (median 5.82 months) arms. For patients EGFR pathway activation the PFS HR was 0.53 (p-value=0.03 – significant, median PFS: neratinib, 6.21 months, control, 5.26 months). However, there was no significant improvement in OS in EGFR amplified/mutated patients (HR 1.05; p-value 0.87) between neratinib (median 14.2) compared to the control arm (median 14.6).

CONCLUSION

Neratinib prolonged PFS in the EGFR positive subpopulation but there was no overall PFS benefit, or any OS improvement.

Deep Learning for Adjacent Segment Disease at Preoperative MRI for Cervical Radiculopathy

Background Patients who undergo surgery for cervical radiculopathy are at risk for developing adjacent segment disease (ASD). Identifying patients who will develop ASD remains challenging for clinicians. Purpose To develop and validate a deep learning algorithm capable of predicting ASD by using only preoperative cervical MRI in patients undergoing single-level anterior cervical diskectomy and fusion (ACDF). Materials and Methods In this Health Insurance Portability and Accountability Act-compliant study, retrospective chart review was performed for 1244 patients undergoing single-level ACDF in two tertiary care centers. After application of inclusion and exclusion criteria, 344 patients were included, of whom 60% (n = 208) were used for training and 40% for validation (n = 43) and testing (n = 93). A deep learning-based prediction model with 48 convolutional layers was designed and trained by using preoperative T2-sagittal cervical MRI. To validate model performance, a neuroradiologist and neurosurgeon independently provided ASD predictions for the test set. Validation metrics included accuracy, areas under the curve, and F1 scores. The difference in proportion of wrongful predictions between the model and clinician was statistically tested by using the McNemar test. Results A total of 344 patients (median age, 48 years; interquartile range, 41-58 years; 182 women) were evaluated. The model predicted ASD on the 93 test images with an accuracy of 88 of 93 (95%; 95% CI: 90, 99), sensitivity of 12 of 15 (80%; 95% CI: 60, 100), and specificity of 76 of 78 (97%; 95% CI: 94, 100). The neuroradiologist and neurosurgeon provided predictions with lower accuracy (54 of 93; 58%; 95% CI: 48, 68), sensitivity (nine of 15; 60%; 95% CI: 35, 85), and specificity (45 of 78; 58%; 95% CI: 56, 77) compared with the algorithm. The McNemar test on the contingency table demonstrated that the proportion of wrongful predictions was significantly lower by the model (test statistic, 2.000; P < .001). Conclusion A deep learning algorithm that used only preoperative cervical T2-weighted MRI outperformed clinical experts at predicting adjacent segment disease in patients undergoing surgery for cervical radiculopathy. © RSNA, 2021 An earlier incorrect version appeared online. This article was corrected on September 22, 2021.

Automating Clinical Chart Review: An Open-Source Natural Language Processing Pipeline Developed on Free-Text Radiology Reports From Patients With Glioblastoma

PURPOSE

The aim of this study was to develop an open-source natural language processing (NLP) pipeline for text mining of medical information from clinical reports. We also aimed to provide insight into why certain variables or reports are more suitable for clinical text mining than others.

MATERIALS AND METHODS

Various NLP models were developed to extract 15 radiologic characteristics from free-text radiology reports for patients with glioblastoma. Ten-fold cross-validation was used to optimize the hyperparameter settings and estimate model performance. We examined how model performance was associated with quantitative attributes of the radiologic characteristics and reports.

RESULTS

In total, 562 unique brain magnetic resonance imaging reports were retrieved. NLP extracted 15 radiologic characteristics with high to excellent discrimination (area under the curve, 0.82 to 0.98) and accuracy (78.6% to 96.6%). Model performance was correlated with the inter-rater agreement of the manually provided labels (ρ = 0.904; P < .001) but not with the frequency distribution of the variables of interest (ρ = 0.179; P = .52). All variables labeled with a near perfect inter-rater agreement were classified with excellent performance (area under the curve > 0.95). Excellent performance could be achieved for variables with only 50 to 100 observations in the minority group and class imbalances up to a 9:1 ratio. Report-level classification accuracy was not associated with the number of words or the vocabulary size in the distinct text documents.

CONCLUSION

This study provides an open-source NLP pipeline that allows for text mining of narratively written clinical reports. Small sample sizes and class imbalance should not be considered as absolute contraindications for text mining in clinical research. However, future studies should report measures of inter-rater agreement whenever ground truth is based on a consensus label and use this measure to identify clinical variables eligible for text mining.

Therapeutic cancer vaccines for pediatric malignancies: advances, challenges, and emerging technologies

Though outcomes for pediatric cancer patients have significantly improved over the past several decades, too many children still experience poor outcomes and survivors suffer lifelong, debilitating late effects after conventional chemotherapy, radiation, and surgical treatment. Consequently, there has been a renewed focus on developing novel targeted therapies to improve survival outcomes. Cancer vaccines are a promising type of immunotherapy that leverage the immune system to mediate targeted, tumor-specific killing through recognition of tumor antigens, thereby minimizing off-target toxicity. As such, cancer vaccines are orthogonal to conventional cancer treatments and can therefore be used alone or in combination with other therapeutic modalities to maximize efficacy. To date, cancer vaccination has remained largely understudied in the pediatric population. In this review, we discuss the different types of tumor antigens and vaccine technologies (dendritic cells, peptides, nucleic acids, and viral vectors) evaluated in clinical trials, with a focus on those used in children. We conclude with perspectives on how advances in combination therapies, tumor antigen (eg, neoantigen) selection, and vaccine platform optimization can be translated into clinical practice to improve outcomes for children with cancer.

SURG-02. SURVIVAL PREDICTION AFTER NEUROSURGICAL RESECTION OF BRAIN METASTASES: A MACHINE LEARNING APPROACH

INTRODUCTION

Survival prediction in brain metastases (BMs) remains challenging. Current prognostic models have been created and validated almost completely with data from patients receiving radiotherapy only, leaving uncertainty about surgical patients. Therefore, the aim of this study was to build and validate a model predicting 6-month survival after BM resection using different machine learning (ML) algorithms.

METHODS

An institutional database of 1062 patients who underwent resection for BM was split into a 80:20 training and testing set. Seven different ML algorithms were trained and assessed for performance. Moreover, an ensemble model was created incorporating random forest, adaptive boosting, gradient boosting, and logistic regression algorithms. Five-fold cross validation was used for hyperparameter tuning. Model performance was assessed using area under the receiver-operating curve (AUC) and calibration and was compared against the diagnosis-specific graded prognostic assessment (ds-GPA); the most established prognostic model in BMs.

RESULTS

The ensemble model showed superior performance with an AUC of 0.81 in the hold-out test set, a calibration slope of 1.14, and a calibration intercept of -0.08, outperforming the ds-GPA (AUC 0.68). Patients were stratified into high-, medium- and low-risk groups for death at 6 months; these strata strongly predicted both 6-months and longitudinal overall survival (p &lt; 0.001).

CONCLUSIONS

We developed and internally validated an ensemble ML model that accurately predicts 6-month survival after neurosurgical resection for BM, outperforms the most established model in the literature, and allows for meaningful risk stratification. Future efforts should focus on external validation of our model.

CTNI-11. CC-115 IN NEWLY DIAGNOSED MGMT UNMETHYLATED GLIOBLASTOMA IN THE INDIVIDUALIZED SCREENING TRIAL OF INNOVATIVE GLIOBLASTOMA THERAPY (INSIGHT): A PHASE II RANDOMIZED BAYESIAN ADAPTIVE PLATFORM TRIAL

BACKGROUND

CC-115 is an oral, CNS-penetrant, selective inhibitor of mammalian target of rapamycin kinase (mTOR) and deoxyribonucleic acid-dependent protein kinase (DNA-PK). Both targets are important in glioblastoma; PI3K/Akt/mTOR signaling is hyperactive in most glioblastomas, and DNA-PK is integral to repair of radiotherapy-mediated DNA damage. To investigate CC-115 in newly diagnosed glioblastoma and explore potential genomic biomarker associations, CC-115 was evaluated in the Individualized Screening Trial of Innovative Glioblastoma Therapy (INSIGhT) trial, an adaptive platform trial designed to efficiently test experimental agents.

METHODS

Adults with newly diagnosed MGMT-unmethylated glioblastoma, with genomic data available, are eligible for this ongoing trial. Patients are adaptively randomized to one of several experimental arms or the control arm: standard radiotherapy with concurrent and adjuvant temozolomide. The primary endpoint is overall survival (OS). Patients randomized to CC-115 (10mg po BID) received it concurrently with radiotherapy and as adjuvant monotherapy. As the first in-human use of CC-115 with radiation, a safety lead-in 3 + 3 design was used.

RESULTS

Twelve patients were randomized to CC-115; seven patients had possible treatment-related CTCAE grade &gt; 3 toxicity, including four pre-specified dose-limiting toxicities: liver function abnormality (n=1), hyperlipidemia (n=1), lipase elevation (n=1) and cerebral edema (n=1). There was no significant difference in progression-free survival (PFS, median 4.2 months [CC-115] vs. 5.2 months, p=0.9) or OS (median 10.1 months [CC-115] vs. 14.5 months, p=0.9) compared to the 50 patients randomized to the control arm. Based on early PFS results, randomization probability to CC-115 decreased from 25% to &lt; 10% at time of the trial arm closure.

CONCLUSION

Concurrent and adjuvant CC-115 was associated with toxicity and failed to improve PFS or OS. The INSIGhT trial design allowed for more efficient testing of CC-115, decreasing patients and resources allocated to a therapy that was discontinued due to concerns about toxicity and unfavorable risk-to-benefit ratio.

CTNI-12. PRELIMINARY RESULTS OF THE ABEMACICLIB ARM IN THE INDIVIDUALIZED SCREENING TRIAL OF INNOVATIVE GLIOBLASTOMA THERAPY (INSIGHT): A PHASE II PLATFORM TRIAL USING BAYESIAN ADAPTIVE RANDOMIZATION

BACKGROUND

The cyclin D-CDK4/6-Rb pathway is activated in most glioblastomas. Abemaciclib is a potent CDK4/6 inhibitor with good brain penetration approved for ER/PR/HER2- breast cancer. In order to efficiently evaluate the potential impact of abemaciclib on overall survival (OS) in newly diagnosed glioblastoma and to simultaneously develop information regarding potential genomic biomarker associations, abemaciclib was included as an arm on the Individualized Screening Trial of Innovative Glioblastoma Therapy (INSIGhT) trial. INSIGhT is a phase II platform trial using response adaptive randomization and deep genomic profiling to more efficiently test experimental agents in MGMT unmethylated glioblastoma and potentially accelerate identification of novel therapies for phase III testing. Initial randomization was equal between abemaciclib, control, and two other experimental arms but subsequent randomization was adapted based on efficacy as determined by progression-free survival (PFS). Ineffective arms were discontinued and new arms added by protocol amendment. We report preliminary results for the abemaciclib arm which has completed accrual.

METHODS

Patients with newly diagnosed MGMT-unmethylated glioblastoma were randomized to receive either radiotherapy with concomitant and adjuvant temozolomide at standard doses or standard radiochemotherapy followed by adjuvant abemaciclib (150–200 mg orally BID) without temozolomide. Treatment continued until progression or development of unacceptable toxicities. The primary endpoint was OS. Association between abemaciclib efficacy and cyclin D-CDK4/6-Rb pathway genomic alterations was also investigated.

RESULTS

There were 123 patients (50 control; 73 treated with abemaciclib). Abemaciclib was generally well-tolerated with no new toxicity signals identified. PFS was significantly longer (p=0.03, logrank test) with abemaciclib (median 6.31 months 95% CI [5.29, 8.18]) compared to the control arm (5.16 months 95% CI [4.37, 6.28]). 28/50 control and 36/73 abemaciclib patients remain alive.

CONCLUSION

Preliminary analysis suggests that abemaciclib increases PFS compared to control. Updated toxicity, PFS and survival data and potential genomic biomarker associations will be presented.

Survival prediction of glioblastoma patients-are we there yet? A systematic review of prognostic modeling for glioblastoma and its clinical potential

Glioblastoma is associated with a poor prognosis. Even though survival statistics are well-described at the population level, it remains challenging to predict the prognosis of an individual patient despite the increasing number of prognostic models. The aim of this study is to systematically review the literature on prognostic modeling in glioblastoma patients. A systematic literature search was performed to identify all relevant studies that developed a prognostic model for predicting overall survival in glioblastoma patients following the PRISMA guidelines. Participants, type of input, algorithm type, validation, and testing procedures were reviewed per prognostic model. Among 595 citations, 27 studies were included for qualitative review. The included studies developed and evaluated a total of 59 models, of which only seven were externally validated in a different patient cohort. The predictive performance among these studies varied widely according to the AUC (0.58-0.98), accuracy (0.69-0.98), and C-index (0.66-0.70). Three studies deployed their model as an online prediction tool, all of which were based on a statistical algorithm. The increasing performance of survival prediction models will aid personalized clinical decision-making in glioblastoma patients. The scientific realm is gravitating towards the use of machine learning models developed on high-dimensional data, often with promising results. However, none of these models has been implemented into clinical care. To facilitate the clinical implementation of high-performing survival prediction models, future efforts should focus on harmonizing data acquisition methods, improving model interpretability, and externally validating these models in multicentered, prospective fashion.

The Surgical Resection of Brainstem Glioma: Outcomes and Prognostic Factors

BACKGROUND

The management of brainstem glioma remains controversial, with increasing evidence supporting surgical resection as the primary treatment for a select subgroup of tumors. However, there remains no consensus on the specific benefits and risks, the selection of surgical candidates, and prognostic factors that may further refine surgical indications.

METHODS

A retrospective single-surgeon chart review was performed for all patients who underwent surgical treatment for radiographically suspected brainstem glioma between 2000 and 2017. Preoperative and postoperative radiographic evaluations on magnetic resonance imaging were conducted. Survival outcomes were collected, and machine-learning techniques were used for multivariate analysis.

RESULTS

Seventy-seven patients with surgical treatment of brainstem glioma were identified, with a median age of 9 years (range, 0-58 years). The cohort included 64% low-grade (I and II) and 36% high-grade (III and IV) tumors. For all patients, the 1-year and 5-year overall survival were 76.4% and 62.3%, respectively. Transient neurologic deficit was present in 34% of cases, and permanent deficit in a further 29%.

CONCLUSIONS

The radical surgical resection of brainstem gliomas can be performed with acceptable risk in well-selected cases and likely confers survival advantage for what is otherwise a rapidly and universally fatal disease. Various radiographic features are useful during patient selection and may guide treatment selection.

Contemporary assessment of extent of resection in molecularly defined categories of diffuse low-grade glioma: a volumetric analysis

OBJECTIVE

While the effect of increased extent of resection (EOR) on survival in diffuse infiltrating low-grade glioma (LGG) patients is well established, there is still uncertainty about the influence of the new WHO molecular subtypes. The authors designed a retrospective analysis to assess the interplay between EOR and molecular classes.

METHODS

The authors retrospectively reviewed the records of 326 patients treated surgically for hemispheric WHO grade II LGG at Brigham and Women's Hospital and Massachusetts General Hospital (2000-2017). EOR was calculated volumetrically and Cox proportional hazards models were built to assess for predictive factors of overall survival (OS), progression-free survival (PFS), and malignant progression-free survival (MPFS).

RESULTS

There were 43 deaths (13.2%; median follow-up 5.4 years) among 326 LGG patients. Median preoperative tumor volume was 31.2 cm3 (IQR 12.9-66.0), and median postoperative residual tumor volume was 5.8 cm3 (IQR 1.1-20.5). On multivariable Cox regression, increasing postoperative volume was associated with worse OS (HR 1.02 per cm3; 95% CI 1.00-1.03; p = 0.016), PFS (HR 1.01 per cm3; 95% CI 1.00-1.02; p = 0.001), and MPFS (HR 1.01 per cm3; 95% CI 1.00-1.02; p = 0.035). This result was more pronounced in the worse prognosis subtypes of IDH-mutant and IDH-wildtype astrocytoma, for which differences in survival manifested in cases with residual tumor volume of only 1 cm3. In oligodendroglioma patients, postoperative residuals impacted survival when exceeding 8 cm3. Other significant predictors of OS were age at diagnosis, IDH-mutant and IDH-wildtype astrocytoma classes, adjuvant radiotherapy, and increasing preoperative volume.

CONCLUSIONS

The results corroborate the role of EOR in survival and malignant transformation across all molecular subtypes of diffuse LGG. IDH-mutant and IDH-wildtype astrocytomas are affected even by minimal postoperative residuals and patients could potentially benefit from a more aggressive surgical approach.

Natural Language Processing for Automated Quantification of Brain Metastases Reported in Free-Text Radiology Reports

PURPOSE

Although the bulk of patient-generated health data are increasing exponentially, their use is impeded because most data come in unstructured format, namely as free-text clinical reports. A variety of natural language processing (NLP) methods have emerged to automate the processing of free text ranging from statistical to deep learning-based models; however, the optimal approach for medical text analysis remains to be determined. The aim of this study was to provide a head-to-head comparison of novel NLP techniques and inform future studies about their utility for automated medical text analysis.

PATIENTS AND METHODS

Magnetic resonance imaging reports of patients with brain metastases treated in two tertiary centers were retrieved and manually annotated using a binary classification (single metastasis v two or more metastases). Multiple bag-of-words and sequence-based NLP models were developed and compared after randomly splitting the annotated reports into training and test sets in an 80:20 ratio.

RESULTS

A total of 1,479 radiology reports of patients diagnosed with brain metastases were retrieved. The least absolute shrinkage and selection operator (LASSO) regression model demonstrated the best overall performance on the hold-out test set with an area under the receiver operating characteristic curve of 0.92 (95% CI, 0.89 to 0.94), accuracy of 83% (95% CI, 80% to 87%), calibration intercept of -0.06 (95% CI, -0.14 to 0.01), and calibration slope of 1.06 (95% CI, 0.95 to 1.17).

CONCLUSION

Among various NLP techniques, the bag-of-words approach combined with a LASSO regression model demonstrated the best overall performance in extracting binary outcomes from free-text clinical reports. This study provides a framework for the development of machine learning-based NLP models as well as a clinical vignette of patients diagnosed with brain metastases.

Severe Radiation Necrosis Refractory to Surgical Resection in Patients with Melanoma and Brain Metastases Managed with Ipilimumab/Nivolumab and Brain-Directed Stereotactic Radiation Therapy

BACKGROUND

The use of targeted therapies and immune checkpoint inhibitors has drastically changed the management of patients with melanoma and brain metastases. Specifically, combination therapy with ipilimumab, a cytotoxic T-lymphocyte antigen 4 inhibitor, and nivolumab, a programmed cell death protein 1 inhibitor, has become a preferred systemic therapy option for patients with melanoma and asymptomatic brain metastases. However, the efficacy and toxicity profile of these agents in combination with brain-directed radiation therapy is not well described.

CASE DESCRIPTION

In this case series, we highlight a series of patients with melanoma demonstrating severe radiation necrosis immediately refractory to surgical resection following brain-directed stereotactic radiation therapy with concurrent ipilimumab and nivolumab. Three patients described in this series each received stereotactic radiation therapy to a dose of 30 Gy in 5 fractions to a melanoma brain metastasis. These areas developed radiographic evidence of necrosis, which was managed surgically and progressed immediately and rapidly after resection. Re-resection, bevacizumab, steroids, and/or discontinuation of nivolumab was used to mitigate further necrosis with varying efficacy.

CONCLUSIONS

Patients with metastatic melanoma receiving brain-directed radiation therapy with concurrent ipilimumab and nivolumab are at risk for developing severe, surgically refractory radiation necrosis and should be closely followed clinically and with imaging. The exact mechanism for such severe necrosis is unknown, and future studies are needed to better understand this pathophysiology and identify optimal treatment strategies.

Fibrinolytics and Intraventricular Hemorrhage: A Systematic Review and Meta-analysis

Intraventricular hemorrhage (IVH) is an independent poor prognostic factor in subarachnoid and intra-parenchymal hemorrhage. The use of intraventricular fibrinolytics (IVF) has long been debated, and its exact effects on outcomes are unknown. A systematic review and meta-analysis were performed in accordance with the PRISMA guidelines to assess the impact of IVF after non-traumatic IVH on mortality, functional outcome, intracranial bleeding, ventriculitis, time until clearance of third and fourth ventricles, obstruction of external ventricular drains (EVD), and shunt dependency. Nineteen studies were included in the meta-analysis, totaling 1020 patients. IVF was associated with lower mortality (relative risk [RR] 0.58; 95% confidence interval [CI] 0.47-0.72), fewer EVD obstructions (RR 0.41; 95% CI 0.22-0.74), and a shorter time until clearance of the ventricles (median difference [MD] - 4.05 days; 95% CI - 5.52 to - 2.57). There was no difference in good functional outcome, RR 1.41 (95% CI 0.98-2.03), or shunt dependency, RR 0.93 (95% CI 0.70-1.22). Correction for publication bias predicted an increased risk of intracranial bleeding, RR 1.67 (95% CI 1.01-2.74) and a lower risk of ventriculitis, RR 0.68 (95% CI 0.45-1.03) in IVH patients treated with IVF. IVF was associated with improved survival, faster clearance of blood from the ventricles and fewer drain obstructions, but further research is warranted to elucidate the effects on ventriculitis, long-term functional outcomes, and re-hemorrhage.