Deep learning segmentation of organs-at-risk with integration into clinical workflow for pediatric brain radiotherapy.

PURPOSE: Radiation therapy (RT) of pediatric brain cancer is known to be associated with long-term neurocognitive deficits. Although target and organs-at-risk (OARs) are contoured as part of treatment planning, other structures linked to cognitive functions are often not included. This paper introduces a novel automatic segmentation tool specifically designed for the unique challenges posed by pediatric patients undergoing brain RT, as well as its seamless integration into the existing clinical workflow. METHODS AND MATERIALS: Images of 47 pediatric brain cancer patients aged 1 to 20 years old and 33 two-year-old healthy infants were used to train a vision transformer, UNesT, for the segmentation of five brain OARs. The trained model was then incorporated to clinical workflow via DICOM connections between a treatment planning system (TPS) and a server hosting the trained model such that scans are sent from TPS to the server, automatically segmented, and sent back to TPS for treatment planning. RESULTS: The proposed automatic segmentation framework achieved a median dice similarity coefficient of 0.928 (frontal white matter), 0.908 (corpus callosum), 0.933 (hippocampi), 0.819 (temporal lobes), and 0.960 (brainstem) with a mean ± SD run time of 1.8 ± 0.67 s over 20 test cases. CONCLUSIONS: The pediatric brain segmentation tool showed promising performance on five OARs linked to neurocognitive functions and can easily be extended for additional structures. The proposed integration to the clinic enables easy access to the tool from clinical platforms and minimizes disruption to existing workflow while maximizing its benefits.

Impact of Consolidative Radiation on Overall and Progression-Free Survival in Pediatric, Adolescent, and Young Adult Metastatic Bone and Soft Tissue Sarcoma.

PURPOSE: To determine the association between consolidative radiation (RT) and survival in children, adolescents, and young adults with metastatic sarcoma. METHODS AND MATERIALS: Eligibility criteria included patients aged ≤39 years with newly diagnosed metastatic bone or soft tissue sarcoma who completed local control of the primary tumor without disease progression. Consolidative RT was defined as RT to all known sites of metastatic disease. The Kaplan-Meier method was used to estimate overall survival (OS) and progression-free survival (PFS). The least absolute shrinkage and selection operator Cox provided adjusted estimates. To account for immortal time bias, consolidative RT was used as a time-varying covariate in a time dependent Cox model. Distant failure was estimated using the Fine-Gray model. RESULTS: Patients (n = 85) had a median age at diagnosis of 14.8 years. Most common histology was Ewing Sarcoma (45.9%) followed by rhabdomyosarcoma (40.0%). Receipt of consolidative RT was associated with Ewing Sarcoma (P < .001) and local control modality as those who underwent local control with surgery and RT compared with surgery alone were more likely to be treated with consolidative RT (P = .034). Consolidative RT was independently associated with improved OS (hazard ratio [HR], 0.41; 95% CI, 0.17-0.98; P = .045) and improved PFS (HR, 0.37; 95% CI, 0.16-0.88; P = .024) after adjusting for confounding variables and immortal time bias. Patients treated with consolidative RT also experienced a lower risk of distant failure (HR, 0.33; 95% CI, 0.17-0.64; P = .001). In an independent data set of patients with metachronous progression (n = 36), consolidative RT remained independently associated with improved OS. CONCLUSIONS: Consolidative RT was independently associated with improved OS and PFS and decreased risk of distant failure in child, adolescent, and young adult patients with metastatic sarcoma. Future work should evaluate biomarkers to optimize patient selection, timing, and dose for consolidative RT.

Impact of socioeconomic status and chemotherapy on neurocognitive performance in children with brain tumors.

Background: Although the relationship between radiation and neurocognition has been extensively studied in the pediatric brain tumor population, it is increasingly recognized that neurocognitive impairment is multifactorial. Therefore, we quantified the effect of socioeconomic status (SES) and chemotherapy on neurocognitive impairment and decline post-treatment. Methods: Eligible patients included those diagnosed with a brain tumor at < 22 years of age with ≥1 neurocognitive assessment. Neurocognitive impairment was defined as performance 1.5 standard deviations below the normative mean using age-standardized measures of intellectual function. Neurocognitive decline was defined as a negative slope. Neurocognitive outcomes included Wechsler indices of Full-Scale Intelligence Quotient (IQ). Logistic regression identified variables associated with neurocognitive impairment. Longitudinal data was analyzed using linear mixed models. Results: Eligible patients (n = 152, median age at diagnosis = 9.6 years) had a mean neurocognitive follow-up of 50.2 months. After accounting for age and receipt of craniospinal irradiation, patients with public insurance had 8-fold increased odds of impaired IQ compared to private insurance (odds ratio [OR]: 7.59, P < .001). After accounting for age, change in IQ was associated with chemotherapy use (slope: -0.45 points/year with chemotherapy vs. 0.71 points/year without chemotherapy, P = .012). Conclusions: Public insurance, an indicator of low SES, was associated with post-treatment impairment in IQ, highlighting the need to incorporate SES measures into prospective studies. Chemotherapy was associated with change in IQ. Further work is needed to determine whether impairment associated with low SES is secondary to baseline differences in IQ prior to brain tumor diagnosis, brain tumor/therapy itself, or some combination thereof.

Clinical Characterization of a Table Mounted Range Shifter Board for Synchrotron-Based Intensity Modulated Proton Therapy for Pediatric Craniospinal Irradiation.

Purpose: To report our design, manufacturing, commissioning and initial clinical experience with a table-mounted range shifter board (RSB) intended to replace the machine-mounted range shifter (MRS) in a synchrotron-based pencil beam scanning (PBS) system to reduce penumbra and normal tissue dose for image-guided pediatric craniospinal irradiation (CSI). Methods: A custom RSB was designed and manufactured from a 3.5 cm thick slab of polymethyl methacrylate (PMMA) to be placed directly under patients, on top of our existing couch top. The relative linear stopping power (RLSP) of the RSB was measured using a multi-layer ionization chamber, and output constancy was measured using an ion chamber. End-to-end tests were performed using the MRS and RSB approaches using an anthropomorphic phantom and radiochromic film measurements. Cone beam CT (CBCT) and 2D planar kV X-ray image quality were compared with and without the RSB present using image quality phantoms. CSI plans were produced using MRS and RSB approaches for two retrospective pediatric patients, and the resultant normal tissue doses were compared. Results: The RLSP of the RSB was found to be 1.163 and provided computed penumbra of 6.9 mm in the phantom compared to 11.8 mm using the MRS. Phantom measurements using the RSB demonstrated errors in output constancy, range, and penumbra of 0.3%, -0.8%, and 0.6 mm, respectively. The RSB reduced mean kidney and lung dose compared to the MRS by 57.7% and 46.3%, respectively. The RSB decreased mean CBCT image intensities by 86.8 HU but did not significantly impact CBCT or kV spatial resolution providing acceptable image quality for patient setup. Conclusions: A custom RSB for pediatric proton CSI was designed, manufactured, modeled in our TPS, and found to significantly reduce lateral proton beam penumbra compared to a standard MRS while maintaining CBCT and kV image-quality and is in routine use at our center.

The clinical and molecular characteristics of progressive hypothalamic/optic pathway pilocytic astrocytoma.

BACKGROUND: Unresectable hypothalamic/optic pathway pilocytic astrocytoma (PA) often progresses despite multiple therapies. Identifying clinical and molecular characteristics of progressive tumors may aid in prognostication and treatment. METHODS: We collected 72 unresectable, non-neurofibromatosis type 1-associated hypothalamic/optic pathway PA to identify clinical and biologic factors associated with tumor progression. Tumors that progressed after therapy, metastasized, or resulted in death were categorized into Cohort B; those that did not meet these criteria were categorized into Cohort A. DNA methylation and transcriptome analyses were performed on treatment-naïve tumors, and the findings were validated by immunohistochemistry (IHC). RESULTS: The median follow-up of the entire cohort was 12.3 years. Cohort B was associated with male sex (M:F = 2.6:1), younger age at diagnosis (median 3.2 years vs 6.7 years, P = .005), and high incidence of KIAA1549-BRAF fusion (81.5% vs 38.5%, P = .0032). Cohort B demonstrated decreased CpG methylation and increased RNA expression in mitochondrial genes and genes downstream of E2F and NKX2.3. Transcriptome analysis identified transcription factor TBX3 and protein kinase PIM1 as common downstream targets of E2F and NKX2.3. IHC confirmed increased expression of TBX3 and PIM1 in Cohort B tumors. Gene enrichment analysis identified enrichment of MYC targets and MAPK, PI3K/AKT/mTOR, and p53 pathways, as well as pathways related to mitochondrial function. CONCLUSIONS: We identified risk factors associated with progressive PA. Our results support the model in which the p53-PIM1-MYC axis and TBX3 act alongside MAPK and PI3K/AKT/mTOR pathways to promote tumor progression, highlighting potential new targets for combination therapy and refining disease prognostication.

Dosimetric evaluation of cone-beam CT-based synthetic CTs in pediatric patients undergoing intensity-modulated proton therapy.

PURPOSE: To evaluate dosimetric changes detected using synthetic computed tomography (sCT) derived from online cone-beam CTs (CBCT) in pediatric patients treated using intensity-modulated proton therapy (IMPT). METHODS: Ten pediatric patients undergoing IMPT and aligned daily using proton gantry-mounted CBCT were identified for retrospective analysis with treated anatomical sites fully encompassed in the CBCT field of view. Dates were identified when the patient received both a CBCT and a quality assurance CT (qCT) for routine dosimetric evaluation. sCTs were generated based on a deformable registration between the initial plan CT (pCT) and CBCT. The clinical IMPT plans were re-computed on the same day qCT and sCT, and dosimetric changes due to tissue change or response from the initial plan were computed using each image. Linear regression analysis was performed to determine the correlation between dosimetric changes detected using the qCT and the sCT. Gamma analysis was also used to compare the dose distributions computed on the qCT and sCT. RESULTS: The correlation coefficients (p-values) between qCTs and sCTs for changes detected in target coverage, overall maximum dose, and organ at risk dose were 0.97 (< .001), 0.84 (.002) and 0.91 (< .001), respectively. Mean ± SD gamma pass rates of the sCT-based dose compared to the qCT-based dose at 3%/3 mm, 3%/2 mm, and 2%/2 mm criteria were 96.5%±4.5%, 93.2%±6.3%, and 91.3%±7.8%, respectively. Pass rates tended to be lower for targets near lung. CONCLUSION: While insufficient for re-planning, sCTs provide approximate dosimetry without administering additional imaging dose in pediatric patients undergoing IMPT. Dosimetric changes detected using sCTs are correlated with changes detected using clinically-standard qCTs; however, residual differences in dosimetry remain a limitation. Further improvements in sCT image quality may both improve online dosimetric evaluation and reduce imaging dose for pediatric patients by reducing the need for routine qCTs.

Facilitating MR-Guided Adaptive Proton Therapy in Children Using Deep Learning-Based Synthetic CT.

PURPOSE: To determine whether self-attention cycle-generative adversarial networks (cycle-GANs), a novel deep-learning method, can generate accurate synthetic computed tomography (sCT) to facilitate adaptive proton therapy in children with brain tumors. MATERIALS AND METHODS: Both CT and T1-weighted magnetic resonance imaging (MRI) of 125 children (ages 1-20 years) with brain tumors were included in the training dataset. A model introducing a self-attention mechanism into the conventional cycle-GAN was created to enhance tissue interfaces and reduce noise. The test dataset consisted of 7 patients (ages 2-14 years) who underwent adaptive planning because of changes in anatomy discovered on MRI during proton therapy. The MRI during proton therapy-based sCT was compared with replanning CT (ground truth). RESULTS: The Hounsfield unit-mean absolute error was significantly reduced with self-attention cycle-GAN, as compared with conventional cycle-GAN (65.3 ± 13.9 versus 88.9 ± 19.3, P < .01). The average 3-dimensional gamma passing rates (2%/2 mm criteria) for the original plan on the anatomy of the day and for the adapted plan were high (97.6% ± 1.2% and 98.9 ± 0.9%, respectively) when using sCT generated by self-attention cycle-GAN. The mean absolute differences in clinical target volume (CTV) receiving 95% of the prescription dose and 80% distal falloff along the beam axis were 1.1% ± 0.8% and 1.1 ± 0.9 mm, respectively. Areas of greatest dose difference were distal to the CTV and corresponded to shifts in distal falloff. Plan adaptation was appropriately triggered in all test patients when using sCT. CONCLUSION: The novel cycle-GAN model with self-attention outperforms conventional cycle-GAN for children with brain tumors. Encouraging dosimetric results suggest that sCT generation can be used to identify patients who would benefit from adaptive replanning.

Toward MR-only proton therapy planning for pediatric brain tumors: Synthesis of relative proton stopping power images with multiple sequence MRI and development of an online quality assurance tool.

PURPOSE: To generate synthetic relative proton stopping power (sRPSP) images from magnetic resonance imaging (MRI) sequence(s) and develop an online quality assurance (QA) tool for sRPSP to facilitate safe integration of magnetic resonance (MR)-only proton planning into clinical practice. MATERIALS AND METHODS: Planning computed tomography (CT) and MR images of 195 pediatric brain tumor patients were utilized (training: 150, testing: 45). Seventeen consistent-cycle generative adversarial network (ccGAN) models were trained separately using paired CT-converted RPSP and MRI datasets to transform a subject's MRI into sRPSP. T1-weighted (T1W), T2-weighted (T2W), and FLAIR MRI were permutated to form 17 combinations, with or without preprocessing, for determining the optimal training sequence(s). For evaluation, sRPSP images were converted to synthetic CT (sCT) and compared to the real CT in terms of mean absolute error (MAE) in Hounsfield units (HU). For QA, sCT was deformed and compared to a reference template built from training dataset to produce a flag map, highlighting pixels that deviate by >100 HU and fall outside the mean ± standard deviation reference intensity. The gamma intensity analysis (10%/3 mm) of the deformed sCT against the QA template on the intensity difference was investigated as a surrogate of sCT accuracy. RESULTS: The sRPSP images generated from a single T1W or T2W sequence outperformed that generated from multi-MRI sequences in terms of MAE (all p < 0.05). Preprocessing with N4 bias and histogram matching reduced MAE of T2W MRI-based sCT (54 ± 21 HU vs. 42 ± 13 HU, p = 0.002). The gamma intensity analysis of sCT against the QA template was highly correlated with the MAE of sCT against the real CT in the testing cohort (r = -0.89 for T1W sCT; r = -0.93 for T2W sCT). CONCLUSION: Accurate sRPSP images can be generated from T1W/T2W MRI for proton planning. A QA tool highlights regions of inaccuracy, flagging problematic cases unsuitable for clinical use.

Rethinking the Definition of High Risk in Pediatric Salivary Gland Carcinoma.

OBJECTIVE: Indications for adjuvant radiation in pediatric salivary gland carcinoma rely on high-risk criteria extrapolated from adult data. We sought to determine whether adult-derived high-risk criteria were prognostic in children aged ≤21 years or young adults aged 22 to 39 years. STUDY DESIGN: Cross-sectional analysis of a hospital-based national registry. SETTING: Patients were identified from the National Cancer Database between 2004 and 2015. METHODS: High-risk criteria were defined as adenoid cystic histology, intermediate/high grade, T3/T4, positive margins, and/or lymph node involvement. Exact matching was used to adjust for differences in baseline characteristics between pediatric and young adult patients. RESULTS: We identified 215 pediatric patients aged ≤21 years, 317 patients aged 22 to 30 years, and 466 patients aged 31 to 39 years. Within the pediatric cohort, there was no significant difference in overall survival (OS) between low- and high-risk groups (5-year OS, 100% vs 98.5%; P = .29). In contrast, within the young adult cohorts, there was a significant difference in OS between low- and high-risk groups in patients aged 22 to 30 years (5-year OS, 100% vs 96.1%; P = .01) and 31 to 39 years (5-year OS, 100% vs 88.5%; P < .001). When high-risk patients were matched 1:1 on high-risk criteria and race, pediatric patients were associated with better OS than those aged 22 to 30 years (P = .044) and those aged 31 to 39 years (P = .005). CONCLUSION: Children have excellent OS, irrespective of adult-derived high-risk status. These findings underscore the need to understand how age modifies clinicopathologic risk factors.

Association Between Brain Substructure Dose and Cognitive Outcomes in Children With Medulloblastoma Treated on SJMB03: A Step Toward Substructure-Informed Planning.

PURPOSE: To characterize the association between neurocognitive outcomes (memory and processing speed) and radiation (RT) dose to the hippocampus, corpus callosum (CC), and frontal white matter (WM) in children with medulloblastoma treated on a prospective study, SJMB03. PATIENTS AND METHODS: Patients age 3-21 years with medulloblastoma were treated at a single institution on a phase III study. The craniospinal RT dose was 23.4 Gy for average-risk patients and 36-39.6 Gy for high-risk patients. The boost dose was 55.8 Gy to the tumor bed. Patients underwent cognitive testing at baseline and once yearly for 5 years. Performance on tests of memory (associative memory and working memory) and processing speed (composite processing speed and perceptual speed) was analyzed. Mixed-effects models were used to estimate longitudinal trends in neurocognitive outcomes. Reliable change index and logistic regression were used to define clinically meaningful neurocognitive decline and identify variables associated with decline. RESULTS: One hundred and twenty-four patients were eligible for inclusion, with a median neurocognitive follow-up of 5 years. Mean right and left hippocampal doses were significantly associated with decline in associative memory in patients without posterior fossa syndrome (all P < .05). Mean CC and frontal WM doses were significantly associated with decline in both measures of processing speed (all P < .05). Median brain substructure dose-volume histograms were shifted to the right for patients with a decline in associative memory or processing speed. The odds of decline in associative memory and composite processing speed increased by 23%-26% and by 10%-15% for every 1-Gy increase in mean hippocampal dose and mean CC or frontal WM dose, respectively. CONCLUSION: Increasing RT dose to the CC or frontal WM and hippocampus is associated with worse performance on tests of processing speed and associative memory, respectively. Brain substructure-informed RT planning may mitigate neurocognitive impairment.

Training a deep neural network coping with diversities in abdominal and pelvic images of children and young adults for CBCT-based adaptive proton therapy.

PURPOSE: To train a deep neural network for correcting abdominal and pelvic cone-beam computed tomography (CBCT) of children and young adults in the presence of diverse patient size, anatomic extent, and scan parameters. MATERIALS AND METHODS: Pretreatment CBCT and planning/repeat CT image pairs from 64 children and young adults treated with proton therapy (aged 1-23 years) were analyzed. To evaluate the impact of anatomic extent in CBCT and data size in the training data, we compared the performance of three cycle-consistent generative adversarial network models that were separately trained by three datasets comprising abdominal (n = 21), pelvic (n = 29), and combined abdominal-pelvic image pairs (n = 50), respectively. The maximum body width of each patient was normalized to a fixed width before training and model application to reduce the impact of variations in body size. The corrected CBCT images by the three models were comparatively evaluated against the repeat CT closest in time to the CBCT (median gap, 0 days; range, 0-6 days) in HU accuracy, estimated dose distribution, and proton range. RESULTS: The network model trained by the combined dataset significantly outperformed the abdomen and pelvis models in mean absolute HU error of the corrected CBCT from 14 testing patients (47 ± 7 HU versus 51 ± 8 HU; paired Wilcoxon signed-rank test, P < 0.01). The larger error (60 ± 7 HU) without the body-size normalization confirmed the efficacy of the preprocessing. The model trained with the combined dataset resulted in gamma passing rates of 98.5 ± 1.9% (2%/2 mm criterion) and the range (80% distal fall-off) differences from the reference within ±3 mm for 91.2 ± 11.5% beamlets. CONCLUSION: Combining data from adjacent anatomic sites and normalizing age-dependent body sizes in children and young adults were beneficial in training a neural network to accurately estimate proton dose from CBCT despite limited training data size and anatomic diversities.

Cardiac-Sparing and Breast-Sparing Whole Lung Irradiation Using Intensity-Modulated Proton Therapy.

PURPOSE: Whole lung irradiation (WLI) is indicated for certain pediatric patients with lung metastases. This study investigated whether WLI delivered as intensity-modulated proton therapy (IMPT) could significantly spare the heart and breasts when compared with conventional WLI delivered with anteroposterior/posteroanterior photon fields and with intensity-modulated photon therapy (IMRT) WLI. MATERIALS AND METHODS: Conventional, IMRT, and IMPT plans were generated for 5 patients (aged 5-22 years). The prescription dose was 16.5 GyRBE in 1.5-GyRBE fractions. Conventional plans used 6-MV photons prescribed to the midline and a field-in-field technique to cover the planning target volume (the internal target volume [ITV] + 1 cm). IMRT plans used 6-MV photons with a 7-beam arrangement with dose prescribed to the planning target volume. IMPT plans used scenario-based optimization with 5% range uncertainty and 5-mm positional uncertainty to cover the ITV robustly. Monte Carlo dose calculation was used for all IMPT plans. Doses were compared with paired Student t test. RESULTS: The ITV Dmean was similar for the IMPT, conventional, and IMRT plans, but the IMPT plans had a lower Dmin and a higher Dmax at tissue interfaces than conventional plans (Dmean ratio: 0.96, P > .05; Dmin ratio: 0.9, P < .001; Dmax ratio: 1.1, P = .014). Dmeans for breast and heart substructures were lower with IMPT plans than with conventional/IMRT plans (heart ratios, 0.63:0.73; left ventricle ratios, 0.61:0.72; right ventricle ratios, 0.45:0.57; left atrium ratios, 0.79:0.85; right atrium ratios, 0.81:0.86; left breast ratios, 0.40:0.51; right breast ratio, 0.46:0.52; all P < .05). CONCLUSIONS: IMPT resulted in comparable ITV coverage and lower mean doses to the heart and breasts when compared with other techniques. Whole lung irradiation delivered as IMPT warrants prospective evaluation in pediatric patients.

A rare manifestation of choriocarcinoma syndrome in a child with primary intracranial germ cell tumor and extracranial metastases: A case report and review of the literature.

Choriocarcinoma syndrome is an uncommon, potentially fatal complication of germ cell tumors (GCTs) in adults, but it is not well documented in children. Pediatric central nervous system (CNS) GCTs comprise a rare group of malignancies not usually associated with extra-CNS metastatic disease. Here, we report the case of a pediatric patient with a suprasellar mixed GCT and pulmonary metastases who presented with intratumoral hemorrhage and stroke. Choriocarcinoma syndrome developed soon after initiating chemotherapy. The primary tumor and pulmonary metastases were successfully treated using a multidisciplinary approach, including neurovascular intervention, chemotherapy, and craniospinal irradiation.

Adaptive Proton Therapy for Pediatric Patients: Improving the Quality of the Delivered Plan With On-Treatment MRI.

PURPOSE: Pencil-beam scanning proton therapy is particularly sensitive to anatomic changes, which may affect the delivered dose distribution. This study examined whether offline adaptation using on-treatment magnetic resonance imaging (MRI) scan during proton therapy could improve plan quality for pediatric patients. METHODS AND MATERIALS: Pediatric patients with at least 1 MRI scan in the treatment position (MRItx) during proton therapy between January 2017 and July 2019 were retrospectively reviewed. Patients underwent MRI and computed tomography simulation. Cases were planned with scenario-based optimization with 3 mm/3% positional/range uncertainty. Patients demonstrating anatomic change on MRItx were recontoured. The original plans were applied to the anatomy-of-the-day for dose recalculation (delivered plans). Plans were subsequently reoptimized offline, using original beam angles and dose-volume constraints (adapted plans). Delivered plans were compared with adapted plans to detect significant changes in plan quality, defined as a ≥5% decrease in the clinical target volume (CTV) receiving 95% of the prescription dose (V95) or a ≥5% increase in the dose-volume parameter used as an organ-at-risk constraint. RESULTS: Seventy-three pediatric patients were eligible, with 303 MRI scans (73 simulation and 230 MRItx scans) available for analysis. The median MRItx scans per patient was 3 (range, 1-7). Twenty patients (27%) showed anatomic change, with 11 (55%) demonstrating a significant change in delivered plan quality. Significant changes were noted on MRItx from week 2 (n = 3) or week 3 (n = 8). Seven of these 11 patients (64%) had a significantly reduced CTV V95 (median decrease, 7.6%; range, 5%-16%). Four (36%) had a significantly increased dose to the brain stem, hippocampus, and/or optic apparatus. Eight had a suprasellar low-grade glioma or head and neck rhabdomyosarcoma. CONCLUSION: On-treatment MRI was useful in detecting anatomic changes during proton therapy. MRI-based offline adaptation improved plan quality for most patients with anatomic changes. Further studies should determine the clinical value of MRI-based adaptive therapy for pediatric patients.

Internal dose escalation associated with increased local control for melanoma brain metastases treated with stereotactic radiosurgery.

OBJECTIVE: The internal high-dose volume varies widely for a given prescribed dose during stereotactic radiosurgery (SRS) to treat brain metastases (BMs). This may be altered during treatment planning, and the authors have previously shown that this improves local control (LC) for non-small cell lung cancer BMs without increasing toxicity. Here, they seek to identify potentially actionable dosimetric predictors of LC after SRS for melanoma BM. METHODS: The records of patients with unresected melanoma BM treated with single-fraction Gamma Knife RS between 2006 and 2017 were reviewed. LC was assessed on a per-lesion basis, defined as stability or a decrease in lesion size. Outcome-oriented approaches were utilized to determine optimal dichotomization for dosimetric variables relative to LC. Univariable and multivariable Cox regression analysis was implemented to evaluate the impact of collected parameters on LC. RESULTS: Two hundred eighty-seven melanoma BMs in 79 patients were identified. The median age was 56 years (range 31-86 years). The median follow-up was 7.6 months (range 0.5-81.6 months), and the median survival was 9.3 months (range 1.3-81.6 months). Lesions were optimally stratified by volume receiving at least 30 Gy (V30) greater than or equal to versus less than 25%. V30 was ≥ and < 25% in 147 and 140 lesions, respectively. For all patients, 1-year LC was 83% versus 66% for V30 ≥ and < 25%, respectively (p = 0.001). Stratifying by volume, lesions 2 cm or less (n = 215) had 1-year LC of 82% versus 70% (p = 0.013) for V30 ≥ and < 25%, respectively. Lesions > 2 to 3 cm (n = 32) had 1-year LC of 100% versus 43% (p = 0.214) for V30 ≥ and < 25%, respectively. V30 was still predictive of LC even after controlling for the use of immunotherapy and targeted therapy. Radionecrosis occurred in 2.8% of lesions and was not significantly associated with V30. CONCLUSIONS: For a given prescription dose, an increased internal high-dose volume, as indicated by measures such as V30 ≥ 25%, is associated with improved LC but not increased toxicity in single-fraction SRS for melanoma BM. Internal dose escalation is an independent predictor of improved LC even in patients receiving immunotherapy and/or targeted therapy. This represents a dosimetric parameter that is actionable at the time of treatment planning and warrants further evaluation.

Risk stratification in pediatric low-grade glioma and glioneuronal tumor treated with radiation therapy: an integrated clinicopathologic and molecular analysis.

BACKGROUND: Management of unresectable pediatric low-grade glioma and glioneuronal tumor (LGG/LGGNT) is controversial. There are no validated prognostic features to guide use of radiation therapy (RT). Our study aimed to identify negative prognostic features in patients treated with RT using clinicopathologic and molecular data and validate these findings in an external dataset. METHODS: Children with non-metastatic, biopsy-proven unresectable LGG/LGGNT treated with RT at a single institution between 1997 and 2017 were identified. Recursive partitioning analysis (RPA) was used to stratify patients into low- and high-risk prognostic groups based on overall survival (OS). CNS9702 data were used for validation. RESULTS: One hundred and fifty patients met inclusion criteria. Median follow-up was 11.4 years. RPA yielded low- and high-risk groups with 10-year OS of 95.6% versus 76.4% (95% CI: 88.7%-98.4% vs 59.3%-87.1%, P = 0.003), respectively. These risk groups were validated using CNS9702 dataset (n = 48) (4-year OS: low-risk vs high-risk: 100% vs 64%, P < 0.001). High-risk tumors included diffuse astrocytoma or location within thalamus/midbrain. Low-risk tumors included pilocytic astrocytoma/ganglioglioma located outside of the thalamus/midbrain. In the subgroup with known BRAF status (n = 49), risk stratification remained prognostic independently of BRAF alteration (V600E or fusion). Within the high-risk group, delayed RT, defined as RT after at least one line of chemotherapy, was associated with a further decrement in overall survival (P = 0.021). CONCLUSION: A high-risk subgroup of patients, defined by diffuse astrocytoma histology or midbrain/thalamus tumor location, have suboptimal long-term survival and might benefit from timely use of RT. These results require validation.

Evaluating pediatric spinal low-grade gliomas: a 30-year retrospective analysis.

PURPOSE: Most pediatric spinal tumors are low-grade gliomas (LGGs). Characterization of these tumors has been difficult given their heterogeneity and rare incidence. The objective was to characterize such tumors diagnosed at our institution. METHODS: Spinal tumors diagnosed in our pediatric patients between 1984 and 2014 were reviewed retrospectively. Demographics, presentation, pathology, imaging, management, and sequelae were examined. RESULTS: Forty patients had spinal LGG tumors, 24 (62%) of which were pilocytic astrocytomas. The most common initial presentations were pain (n = 15), partial extremity paralysis (n = 13), and ataxia (n = 11), with the diagnosis frequently delayed by months (median = 5.9 months, range 4 days-6.2 years). Twenty-nine patients had some tumor resection, and 8 required adjuvant therapy with chemotherapy (n = 4) or radiation (n = 4) post-resection. Ten other patients received only biopsy for histologic diagnosis, who were treated with chemotherapy (n = 4) or radiation (n = 5) post biopsy. Tumor progression was noted in 16 patients (2 after gross-total resection; 10, partial resection; and 4, biopsy). During the evaluation period, 3 patients died secondary to tumor progression. BRAF status could have shortened progression-free survival: patients with BRAFV600E mutations (n = 3) all experienced progression within 10 months. Long-term sequelae of the disease/treatment were mostly residual neurologic deficits (paresthesia, paralysis), chemotherapy-induced hearing loss, and scoliosis. CONCLUSIONS: Spinal LGG is a rare entity with significant long-term effects. Although surgery is the most common initial treatment option, more in-depth analysis of molecular biomarkers may improve stratification and prognostication.