Stepwise Transfer Learning for Expert-Level Pediatric Brain Tumor MRI Segmentation in a Limited Data Scenario.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop, externally test, and evaluate clinical acceptability of a deep learning (DL) pediatric brain tumor segmentation model using stepwise transfer learning. Materials and Methods In this retrospective study, the authors leveraged two T2-weighted MRI datasets (May 2001-December 2015) from a national brain tumor consortium (n = 184; median age, 7 years (range: 1-23 years); 94 male) and a pediatric cancer center (n = 100; median age, 8 years (range: 1-19 years); 47 male) to develop and evaluate DL neural networks for pediatric low-grade glioma segmentation using a novel stepwise transfer learning approach to maximize performance in a limited data scenario. The best model was externally-tested on an independent test set and subjected to randomized, blinded evaluation by three clinicians, wherein they assessed clinical acceptability of expert- and artificial intelligence (AI)-generated segmentations via 10-point Likert scales and Turing tests. Results The best AI model used in-domain, stepwise transfer learning (median DSC: 0.88 [IQR 0.72-0.91] versus 0.812 [0.56-0.89] for baseline model; P = .049). On external testing, AI model yielded excellent accuracy using reference standards from three clinical experts (Expert-1: 0.83 [0.75-0.90]; Expert-2: 0.81 [0.70-0.89]; Expert-3: 0.81 [0.68-0.88]; mean accuracy: 0.82)). On clinical benchmarking (n = 100 scans), experts rated AI-based segmentations higher on average compared with other experts (median Likert score: median 9 [IQR 7-9]) versus 7 [IQR 7-9]) and rated more AI segmentations as clinically acceptable (80.2% versus 65.4%). Experts correctly predicted the origin of AI segmentations in an average of 26.0% of cases. Conclusion Stepwise transfer learning enabled expert-level, automated pediatric brain tumor auto-segmentation and volumetric measurement with a high level of clinical acceptability. ©RSNA, 2024.

Patterns of recurrence after radiotherapy for high-risk neuroblastoma: Implications for radiation dose and field.

BACKGROUND: Prognosis for patients with high-risk neuroblastoma (HR-NBL) is guarded despite aggressive therapy, and few studies have characterized outcomes after radiotherapy in relation to radiation treatment fields. METHODS: Multi-institutional retrospective cohort of 293 patients with HR-NBL who received autologous stem cell transplant (ASCT) and EBRT between 1997-2021. LRR was defined as recurrence at the primary site or within one nodal echelon beyond disease present at diagnosis. Follow-up was defined from the end of EBRT. Event-free survival (EFS) and OS were analyzed by Kaplan-Meier method. Cumulative incidence of locoregional progression (CILP) was analyzed using competing risks of distant-only relapse and death with Gray's test. RESULTS: Median follow-up was 7.0 years (range: 0.01-22.4). Five-year CILP, EFS, and OS were 11.9 %, 65.2 %, and 77.5 %, respectively. Of the 31 patients with LRR and imaging review, 15 (48.4 %) had in-field recurrences (>12 Gy), 6 (19.4 %) had marginal failures (≤12 Gy), and 10 (32.3 %) had both in-field and marginal recurrences. No patients receiving total body irradiation (12 Gy) experienced marginal-only failures (p = 0.069). On multivariable analyses, MYCN amplification had higher risk of LRR (HR: 2.42, 95 % CI: 1.06-5.50, p = 0.035) and post-consolidation isotretinoin and anti-GD2 antibody therapy (HR: 0.42, 95 % CI: 0.19-0.94, p = 0.035) had lower risk of LRR. CONCLUSIONS: Despite EBRT, LRR remains a contributor to treatment failure in HR-NBL with approximately half of LRRs including a component of marginal failure. Future prospective studies are needed to explore whether radiation fields and doses should be defined based on molecular features such as MYCN amplification, and/or response to chemotherapy.

Plasma Kidney Injury Molecule-1 for Preoperative Prediction of Renal Cell Carcinoma Versus Benign Renal Masses, and Association With Clinical Outcomes.

PURPOSE: Both clear cell and papillary renal cell carcinomas (RCCs) overexpress kidney injury molecule-1 (KIM-1). We investigated whether plasma KIM-1 (pKIM-1) may be a useful risk stratification tool among patients with suspicious renal masses. METHODS: Prenephrectomy pKIM-1 was measured in two independent cohorts of patients with renal masses. Cohort 1, from the prospective K2 trial, included 162 patients found to have clear cell RCC (cases) and 162 patients with benign renal masses (controls). Cohort 2 included 247 patients with small (cT1a) renal masses from an academic biorepository, of whom 184 had RCC. We assessed the relationship between pKIM-1, surgical pathology, and clinical outcomes. RESULTS: In Cohort 1, pKIM-1 distinguished RCC versus benign masses with area under the receiver operating curve (AUC-ROC, 0.81 [95% CI, 0.76 to 0.86]). In Cohort 2 (cT1a only), pKIM-1 distinguished RCC versus benign masses (AUC-ROC, 0.74 [95% CI, 0.67 to 0.80]) and the addition of pKIM-1 to an established nomogram for predicting malignancy improved the model AUC-ROC (0.65 [95% CI, 0.57 to 0.74] v 0.78 [95% CI, 0.72 to 0.85]). A pKIM-1 cutpoint identified using Cohort 2 demonstrated sensitivity of 92.5% and specificity of 60% for identifying RCC in Cohort 1. In long-term follow-up of RCC cases (Cohort 1), higher prenephrectomy pKIM-1 was associated with worse metastasis-free survival (multivariable MFS hazard ratio [HR] 1.29 per unit increase in log pKIM-1, 95% CI, 1.10 to 1.53) and overall survival (multivariable OS HR 1.31 per unit increase in log pKIM-1, 95% CI, 1.10 to 1.54). In long-term follow-up of Cohort 2, no metastatic events occurred, consistent with the favorable prognosis of resected cT1a RCC. CONCLUSION: Among patients with renal masses, pKIM-1 is associated with malignant pathology, worse MFS, and risk of death. pKIM-1 may be useful for selecting patients with renal masses for intervention versus surveillance.

Treatment-free survival outcomes from the phase II study of nivolumab and salvage nivolumab/ipilimumab in advanced clear cell renal cell carcinoma (HCRN GU16-260-Cohort A).

BACKGROUND: As part of a partitioned survival analysis, treatment-free survival (TFS) can characterize the overall survival time patients spend between the cessation of immunotherapy and the start of subsequent therapy; both with and without toxicity. Significant TFS was reported for the nivolumab/ipilimumab arms of the CheckMate 067 and 214 trials for patients with advanced melanoma or renal cell carcinoma (aRCC), respectively, where immunotherapy was often halted for toxicity rather than a predefined treatment endpoint. We therefore sought to assess TFS in the HCRN GU16-260 trial, which was designed to reduce toxicity and cap immunotherapy duration. METHODS: Data were analyzed from 128 patients with clear-cell aRCC treated with first-line nivolumab monotherapy for up to 2 years. Salvage nivolumab/ipilimumab for up to 1 year was provided to eligible patients with disease progression at any point or stable disease at 48 weeks (29% of patients). TFS was defined as the area between Kaplan-Meier curves for a time from registration to protocol therapy cessation and for a time from registration to subsequent systemic therapy initiation or death, estimated from 36-month mean times. The time on or off protocol treatment with grade 3+treatment-related adverse events (TRAEs) was also captured. RESULTS: At 36 months from enrollment, 68.3% of patients were alive: 96.8% of International Metastatic RCC Database Consortium (IMDC) favorable-risk patients and 56.6% of those with intermediate/poor-risk, respectively. The 36-month mean time on protocol therapy was 11.5 months including 0.6 months with grade 3+TRAEs (16.0 months for favorable-risk patients and 9.6 months for intermediated/poor-risk patients). The 36-month mean TFS for the whole population was 9.4 months (12.9 months including 1.5 months with grade 3+TRAEs for favorable-risk and 8.0 months including 1.0 months with grade 3+TRAEs for intermediate/poor-risk). At 36 months, 65.6% of favorable-risk patients and 27.1% of intermediate/poor-risk patients were alive and subsequent systemic treatment-free. CONCLUSIONS: Nivolumab monotherapy with salvage nivolumab/ipilimumab in non-responders is an active treatment approach in treatment-naïve patients with aRCC and, similar to nivolumab/ipilimumab in CheckMate 214, results in substantial TFS and toxicity-free TFS. TFS was greatest in patients with favorable-risk disease, supporting the use of an immunotherapy-only regimen in this population.

Large language models to identify social determinants of health in electronic health records.

Social determinants of health (SDoH) play a critical role in patient outcomes, yet their documentation is often missing or incomplete in the structured data of electronic health records (EHRs). Large language models (LLMs) could enable high-throughput extraction of SDoH from the EHR to support research and clinical care. However, class imbalance and data limitations present challenges for this sparsely documented yet critical information. Here, we investigated the optimal methods for using LLMs to extract six SDoH categories from narrative text in the EHR: employment, housing, transportation, parental status, relationship, and social support. The best-performing models were fine-tuned Flan-T5 XL for any SDoH mentions (macro-F1 0.71), and Flan-T5 XXL for adverse SDoH mentions (macro-F1 0.70). Adding LLM-generated synthetic data to training varied across models and architecture, but improved the performance of smaller Flan-T5 models (delta F1 + 0.12 to +0.23). Our best-fine-tuned models outperformed zero- and few-shot performance of ChatGPT-family models in the zero- and few-shot setting, except GPT4 with 10-shot prompting for adverse SDoH. Fine-tuned models were less likely than ChatGPT to change their prediction when race/ethnicity and gender descriptors were added to the text, suggesting less algorithmic bias (p < 0.05). Our models identified 93.8% of patients with adverse SDoH, while ICD-10 codes captured 2.0%. These results demonstrate the potential of LLMs in improving real-world evidence on SDoH and assisting in identifying patients who could benefit from resource support.

PD-1 Expression on Intratumoral Regulatory T Cells Is Associated with Lack of Benefit from Anti-PD-1 Therapy in Metastatic Clear-Cell Renal Cell Carcinoma Patients.

PURPOSE: Programmed cell death protein 1 (PD-1) expression on CD8+TIM-3-LAG-3- tumor-infiltrating cells predicts positive response to PD-1 blockade in metastatic clear-cell renal cell carcinoma (mccRCC). Because inhibition of PD-1 signaling in regulatory T cells (Treg) augments their immunosuppressive function, we hypothesized that PD-1 expression on tumor-infiltrating Tregs would predict resistance to PD-1 inhibitors. EXPERIMENTAL DESIGN: PD-1+ Tregs were phenotyped using multiparametric immunofluorescence in ccRCC tissues from the CheckMate-025 trial (nivolumab: n = 91; everolimus: n = 90). Expression of CD8, PD-1, TIM-3, and LAG-3 was previously determined (Ficial and colleagues, 2021). Clinical endpoints included progression-free survival (PFS), overall survival (OS), and objective response rate (ORR). RESULTS: In the nivolumab (but not everolimus) arm, high percentage of PD-1+ Tregs was associated with shorter PFS (3.19 vs. 5.78 months; P = 0.021), shorter OS (18.1 vs. 27.7 months; P = 0.013) and marginally lower ORR (12.5% vs. 31.3%; P = 0.059). An integrated biomarker (PD-1 Treg/CD8 ratio) was developed by calculating the ratio between percentage of PD-1+Tregs (marker of resistance) and percentage of CD8+PD-1+TIM-3-LAG-3- cells (marker of response). In the nivolumab (but not everolimus) arm, patients with high PD-1 Treg/CD8 ratio experienced shorter PFS (3.48 vs. 9.23 months; P < 0.001), shorter OS (18.14 vs. 38.21 months; P < 0.001), and lower ORR (15.69% vs. 40.00%; P = 0.009). Compared with the individual biomarkers, the PD-1 Treg/CD8 ratio showed improved ability to predict outcomes to nivolumab versus everolimus. CONCLUSIONS: PD-1 expression on Tregs is associated with resistance to PD-1 blockade in mccRCC, suggesting that targeting Tregs may synergize with PD-1 inhibition. A model that integrates PD-1 expression on Tregs and CD8+TIM-3-LAG-3- cells has higher predictive value.

Automated temporalis muscle quantification and growth charts for children through adulthood.

Lean muscle mass (LMM) is an important aspect of human health. Temporalis muscle thickness is a promising LMM marker but has had limited utility due to its unknown normal growth trajectory and reference ranges and lack of standardized measurement. Here, we develop an automated deep learning pipeline to accurately measure temporalis muscle thickness (iTMT) from routine brain magnetic resonance imaging (MRI). We apply iTMT to 23,876 MRIs of healthy subjects, ages 4 through 35, and generate sex-specific iTMT normal growth charts with percentiles. We find that iTMT was associated with specific physiologic traits, including caloric intake, physical activity, sex hormone levels, and presence of malignancy. We validate iTMT across multiple demographic groups and in children with brain tumors and demonstrate feasibility for individualized longitudinal monitoring. The iTMT pipeline provides unprecedented insights into temporalis muscle growth during human development and enables the use of LMM tracking to inform clinical decision-making.

Development and Validation of an Automated Image-Based Deep Learning Platform for Sarcopenia Assessment in Head and Neck Cancer.

Importance: Sarcopenia is an established prognostic factor in patients with head and neck squamous cell carcinoma (HNSCC); the quantification of sarcopenia assessed by imaging is typically achieved through the skeletal muscle index (SMI), which can be derived from cervical skeletal muscle segmentation and cross-sectional area. However, manual muscle segmentation is labor intensive, prone to interobserver variability, and impractical for large-scale clinical use. Objective: To develop and externally validate a fully automated image-based deep learning platform for cervical vertebral muscle segmentation and SMI calculation and evaluate associations with survival and treatment toxicity outcomes. Design, Setting, and Participants: For this prognostic study, a model development data set was curated from publicly available and deidentified data from patients with HNSCC treated at MD Anderson Cancer Center between January 1, 2003, and December 31, 2013. A total of 899 patients undergoing primary radiation for HNSCC with abdominal computed tomography scans and complete clinical information were selected. An external validation data set was retrospectively collected from patients undergoing primary radiation therapy between January 1, 1996, and December 31, 2013, at Brigham and Women's Hospital. The data analysis was performed between May 1, 2022, and March 31, 2023. Exposure: C3 vertebral skeletal muscle segmentation during radiation therapy for HNSCC. Main Outcomes and Measures: Overall survival and treatment toxicity outcomes of HNSCC. Results: The total patient cohort comprised 899 patients with HNSCC (median [range] age, 58 [24-90] years; 140 female [15.6%] and 755 male [84.0%]). Dice similarity coefficients for the validation set (n = 96) and internal test set (n = 48) were 0.90 (95% CI, 0.90-0.91) and 0.90 (95% CI, 0.89-0.91), respectively, with a mean 96.2% acceptable rate between 2 reviewers on external clinical testing (n = 377). Estimated cross-sectional area and SMI values were associated with manually annotated values (Pearson r = 0.99; P < .001) across data sets. On multivariable Cox proportional hazards regression, SMI-derived sarcopenia was associated with worse overall survival (hazard ratio, 2.05; 95% CI, 1.04-4.04; P = .04) and longer feeding tube duration (median [range], 162 [6-1477] vs 134 [15-1255] days; hazard ratio, 0.66; 95% CI, 0.48-0.89; P = .006) than no sarcopenia. Conclusions and Relevance: This prognostic study's findings show external validation of a fully automated deep learning pipeline to accurately measure sarcopenia in HNSCC and an association with important disease outcomes. The pipeline could enable the integration of sarcopenia assessment into clinical decision making for individuals with HNSCC.

In situ psycho-cognitive assessments support self-determined urban green exercise time.

Prescribed nature walks frequently yield improvements to mood and cognition as observed in experimental studies. Research that uses real life settings such as self-determined time exercising outdoors for restorative health benefits may more accurately elicit effects than time-specified study protocols. This study examined in situ psycho-cognitive outcomes of routine walks in urban greenspace to test the concept that self-set exposure duration and not context alone is related to magnitude of psycho-cognitive benefit. Pre-post measurements taken on a diverse participant pool of individuals walking in urban parks and recruited on random days over a two-week period found significant associations between outdoor activity duration and cognitive and mood improvements. Greater outdoor walking duration linearly predicted stronger processing speeds but non-linearly in tests of other cognitive domains. Results of fixed effects model for mean mood change following green exercise show outdoor walking influenced mood change at highest levels of significance, even after accounting for individual level variability in duration. Mood improved for all durations of outdoor walking under a random effects model with high significance. Untethering fixed intervals of outdoor exercise from formal study design revealed briefer but more frequent nature engagement aligned with nature affinity. The influence of unmeasured factors, e.g., nature affinity or restorative conditioning, for prescriptive durations of urban green exercise merits further investigation toward designing wellbeing interventions directed at specific urban populations.

Multi-institutional Characterization of Outcomes for Pediatric and Young Adult Patients With High-Risk Myxopapillary Ependymoma After Radiation Therapy.

PURPOSE: Myxopapillary ependymoma (MPE) is a rare, typically slow-growing subtype of spinal ependymomas. There are no standard guidelines for radiotherapy and long-term outcomes after radiation, particularly patterns of relapse, for pediatric and young adult (YA) patients with MPE remain under-characterized. METHODS AND MATERIALS: This is an Institutional Review Board-approved multi-institutional retrospective cohort study of 60 pediatric and YA patients diagnosed with MPE and received radiotherapy between 2000-2020. Clinical and treatment characteristics, and long-term outcomes were recorded. Site(s) of progression was compared to radiation fields. Survival outcomes were analyzed using Kaplan-Meier method. Cumulative incidence of local in-field progression (CILP) after initial radiotherapy was analyzed using Gray's method with out-of-field-only progression as a competing risk. Univariate analyses were performed using Cox proportional hazard's model. RESULTS: The median age at radiation was 14.8 years (range: 7.1-26.5). At time of radiotherapy, 45 (75.0%) and 35 (58.3%) patients had gross residual and multifocal disease, respectively. Forty-eight (80.0%), seven (11.7%) and five (8.3%) patients received involved field radiotherapy, craniospinal irradiation, and whole spine radiation, respectively. Median follow-up from end of radiotherapy was 6.2 years (range: 0.6-21.0). Five-year overall survival, progression-free survival, and CILP were 100%, 60.8% and 4.1%, respectively. Both local recurrences were at sites of gross residual disease. Of the eighteen out-of-field first recurrences after radiotherapy, all were superior to the initial treatment field and nine had intracranial relapse. On univariate analyses, distant-only recurrence before radiation (HR: 4.00, 95% CI: 1.54-10.43, p = 0.005) was significantly associated with shorter time to progression. CONCLUSIONS: While the risk of recurrence within the radiation field is low, pediatric and YA patients with high-risk MPE remain at risk for recurrences in the spine above the radiation field and intracranially after radiotherapy. Future prospective studies are needed to investigate the appropriate radiation field and dose based on the extent of metastases.

Bevacizumab in Combination With Oxaliplatin, Fluorouracil, and Leucovorin (FOLFOX4) for Previously Treated Metastatic Colorectal Cancer: Results From the Eastern Cooperative Oncology Group Study E3200.

PURPOSE: Colorectal cancer is the second leading cause of cancer mortality in the United States. Antiangiogenic therapy with bevacizumab combined with chemotherapy improves survival in previously untreated metastatic colorectal cancer. This study was conducted to determine the effect of bevacizumab (at 10 mg/kg) on survival duration for oxaliplatin-based chemotherapy in patients with previously treated metastatic colorectal cancer. PATIENTS AND METHODS: Eight hundred twenty-nine metastatic colorectal cancer patients previously treated with a fluoropyrimidine and irinotecan were randomly assigned to one of three treatment groups: oxaliplatin, fluorouracil, and leucovorin (FOLFOX4) with bevacizumab; FOLFOX4 without bevacizumab; or bevacizumab alone. The primary end point was overall survival, with additional determinations of progression-free survival, response, and toxicity. RESULTS: The median duration of survival for the group treated with FOLFOX4 and bevacizumab was 12.9 months compared with 10.8 months for the group treated with FOLFOX4 alone (corresponding hazard ratio for death = 0.75; P = .0011), and 10.2 months for those treated with bevacizumab alone. The median progression-free survival for the group treated with FOLFOX4 in combination with bevacizumab was 7.3 months, compared with 4.7 months for the group treated with FOLFOX4 alone (corresponding hazard ratio for progression = 0.61; P < .0001), and 2.7 months for those treated with bevacizumab alone. The corresponding overall response rates were 22.7%, 8.6%, and 3.3%, respectively (P < .0001 for FOLFOX4 with bevacizumab v FOLFOX4 comparison). Bevacizumab was associated with hypertension, bleeding, and vomiting. CONCLUSION: The addition of bevacizumab to oxaliplatin, fluorouracil, and leucovorin improves survival duration for patients with previously treated metastatic colorectal cancer.

A genomic score to predict local control among patients with brain metastases managed with radiation.

BACKGROUND: Clinical predictors of local recurrence following radiation among patients with brain metastases (BrM) provide limited explanatory power. We developed a DNA-based signature of radiotherapeutic efficacy among patients with BrM to better characterize recurrence risk. METHODS: We identified 570 patients with 1487 BrM managed with whole-brain (WBRT) or stereotactic radiation therapy at Brigham and Women's Hospital/Dana-Farber Cancer Institute (2013-2020) for whom next-generation sequencing panel data (OncoPanel) were available. Fine/Gray's competing risks regression was utilized to compare local recurrence on a per-metastasis level among patients with versus without somatic alterations of likely biological significance across 84 genes. Genes with a q-value ≤ 0.10 were utilized to develop a "Brain-Radiation Prediction Score" ("Brain-RPS"). RESULTS: Genomic alterations in 11 (ATM, MYCL, PALB2, FAS, PRDM1, PAX5, CDKN1B, EZH2, NBN, DIS3, and MDM4) and 2 genes (FBXW7 and AURKA) were associated with decreased or increased risk of local recurrence, respectively (q-value ≤ 0.10). Weighted scores corresponding to the strength of association with local failure for each gene were summed to calculate a patient-level RPS. On multivariable Fine/Gray's competing risks regression, RPS [1.66 (1.44-1.91, P < .001)], metastasis-associated edema [1.60 (1.16-2.21), P = .004], baseline size [1.02 (1.01-1.03), P < .001] and receipt of WBRT without local therapy [4.04 (2.49-6.58), P < .001] were independent predictors of local failure. CONCLUSIONS: We developed a genomic score to quantify local recurrence risk following brain-directed radiation. To the best of our knowledge, this represents the first study to systematically correlate DNA-based alterations with radiotherapeutic outcomes in BrM. If validated, Brain-RPS has potential to facilitate clinical trials aimed at genome-based personalization of radiation in BrM.

Phase II study of nivolumab and salvage nivolumab/ipilimumab in treatment-naïve patients with advanced non-clear cell renal cell carcinoma (HCRN GU16-260-Cohort B).

BACKGROUND: To determine the efficacy and toxicity of nivolumab monotherapy in treatment-naïve patients with non-clear cell renal cell carcinoma (nccRCC) and the efficacy of nivolumab/ipilimumab salvage therapy in patients with tumors unresponsive to initial nivolumab monotherapy. METHODS: Eligible patients with treatment-naïve nccRCC received nivolumab until progressive disease (PD), toxicity, or completion of 96 weeks of treatment (Part A). Patients with PD prior to, or stable disease (SD) at 48 weeks (prolonged SD) were potentially eligible to receive salvage nivolumab/ipilimumab (Part B). Patients were required to submit tissue from a metastatic lesion obtained within 12 months prior to study entry and prior to Part B for correlative studies. RESULTS: 35 patients with nccRCC were enrolled: 19 (54%) had papillary, 6 (17%) had chromophobe and 10 (29%) had unclassified histology. At median follow-up of 22.9 months, RECIST-defined objective response rate (ORR) was 5 of 35 (14.3% 95% CI 4.8% to 30.3%) (complete response (CR) 2 (5.7%) and partial response (PR) 3 (8.6%)). ORR by histology was: papillary-1/19 (5%); chromophobe-1/6 (17%); and unclassified-3/10 (30%). Nine patients (26%) had tumors with sarcomatoid features with 3 (33%) (2 unclassified and 1 papillary) responding. ORR was 0/18, 3/11 (27%) and 2/6 (33%) for patients with tumor progammed death ligand 1 (PD-L1) expression of <5%, ≥5% or not measured, respectively. Median progression-free survival was 4.0 (2.7-4.3) months. Two of five responders have progressed. Thirty-two patients had PD or prolonged SD and therefore, were potentially eligible for salvage nivolumab/ipilimumab (Part B), but 15 patients did not enroll due to grade 2-3 toxicity (6) on nivolumab, symptomatic disease progression (5), or other reasons including no biopsy tissue (4). In the 17 Part B patients, there was one PR (6%) (unclassified/non-sarcomatoid). Grade >3 treatment-related adverse events were seen in 7/35 (20%) on nivolumab and 7/17 (41%) on salvage nivolumab/ipilimumab with one patient experiencing sudden death. CONCLUSIONS: Nivolumab monotherapy has limited activity in treatment-naïve nccRCC with most responses (4 of 5) seen in patients with sarcomatoid and/or unclassified tumors. Toxicity is consistent with prior nivolumab studies. Salvage treatment with nivolumab/ipilimumab was provided in half of these patients with minimal activity. TRIAL REGISTRATION NUMBER: NCT03117309.

Randomized Study of Temozolomide or Temozolomide and Capecitabine in Patients With Advanced Pancreatic Neuroendocrine Tumors (ECOG-ACRIN E2211).

PURPOSE: Patients with advanced pancreatic neuroendocrine tumors (NETs) have few treatment options that yield objective responses. Retrospective and small prospective studies suggest that capecitabine and temozolomide are associated with high response rates (RRs) and long progression-free survival (PFS). PATIENTS AND METHODS: E2211 was a multicenter, randomized, phase II trial comparing temozolomide versus capecitabine/temozolomide in patients with advanced low-grade or intermediate-grade pancreatic NETs. Key eligibility criteria included progression within the preceding 12 months and no prior temozolomide, dimethyl-triazeno-imidazole-carboxamide or dacarbazine, capecitabine or fluorouracil. The primary end point was PFS; secondary endpoints were overall survival, RR, safety, and methylguanine methyltransferase (MGMT) by immunohistochemistry and promoter methylation. RESULTS: A total of 144 patients were enrolled between April 2013 and March 2016 to temozolomide (n = 72) or capecitabine and temozolomide (n = 72); the primary analysis population included 133 eligible patients. At the scheduled interim analysis in January 2018, the median PFS was 14.4 months for temozolomide versus 22.7 months for capecitabine/temozolomide (hazard ratio = 0.58), which was sufficient to reject the null hypothesis for the primary end point (stratified log-rank P = .022). In the final analysis (May 2021), the median overall survival was 53.8 months for temozolomide and 58.7 months for capecitabine/temozolomide (hazard ratio = 0.82, P = .42). MGMT deficiency was associated with response. CONCLUSION: The combination of capecitabine/temozolomide was associated with a significant improvement in PFS compared with temozolomide alone in patients with advanced pancreatic NETs. The median PFS and RR observed with capecitabine/temozolomide are the highest reported in a randomized study for pancreatic NETs. MGMT deficiency was associated with response, and although routine MGMT testing is not recommended, it can be considered for select patients in need of objective response (ClinicalTrials.gov identifier: NCT01824875).

Incidence proportion and prognosis of leptomeningeal disease among patients with breast vs. non-breast primaries.

BACKGROUND: Leptomeningeal disease (LMD) is a relatively uncommon manifestation of advanced cancer. Patients with LMD carry a poor prognosis and often decline rapidly, complicating inclusion in clinical trials. Identification of LMD subsets of greater incidence and more favorable prognosis might facilitate dedicated clinical trials in the future. We hypothesized that patients with breast cancer may represent such a population and sought to assess the relative incidence and prognosis of LMD secondary to breast vs. non-breast primaries. METHODS: We identified 2411 patients with intracranial metastases secondary to breast (N = 501) and non-breast (N = 1910) primaries at Brigham and Women's Hospital/Dana-Farber Cancer Institute between 1996 and 2020, of whom 112 presented with and an additional 161 subsequently developed LMD. A log-rank test and Cox modeling were used to compare outcomes in patients with breast vs. non-breast primaries. RESULTS: Among patients with newly diagnosed intracranial disease, the incidence proportion of concurrent LMD was 11.4% vs. 2.9% among patients with breast vs. non-breast primaries (P < .001). Development of LMD among initially LMD-naïve patients was also more common among patients with breast vs. non-breast primaries (HR = 1.49 [1.05-2.11], P = .03). Patients with LMD secondary to breast vs. non-breast primaries displayed lower all-cause mortality (HR 0.70 [0.52-0.93], P = .01; median survival: 5.2 vs. 2.4 months, respectively), with a greater numerical difference observed in patients with LMD at intracranial involvement (7.4 vs. 2.6 months, respectively). CONCLUSIONS: Patients with breast cancer and LMD may represent an ideal population for clinical trials given the higher incidence and potentially more favorable prognosis seen in this population.

Clinical validation of deep learning algorithms for radiotherapy targeting of non-small-cell lung cancer: an observational study.

BACKGROUND: Artificial intelligence (AI) and deep learning have shown great potential in streamlining clinical tasks. However, most studies remain confined to in silico validation in small internal cohorts, without external validation or data on real-world clinical utility. We developed a strategy for the clinical validation of deep learning models for segmenting primary non-small-cell lung cancer (NSCLC) tumours and involved lymph nodes in CT images, which is a time-intensive step in radiation treatment planning, with large variability among experts. METHODS: In this observational study, CT images and segmentations were collected from eight internal and external sources from the USA, the Netherlands, Canada, and China, with patients from the Maastro and Harvard-RT1 datasets used for model discovery (segmented by a single expert). Validation consisted of interobserver and intraobserver benchmarking, primary validation, functional validation, and end-user testing on the following datasets: multi-delineation, Harvard-RT1, Harvard-RT2, RTOG-0617, NSCLC-radiogenomics, Lung-PET-CT-Dx, RIDER, and thorax phantom. Primary validation consisted of stepwise testing on increasingly external datasets using measures of overlap including volumetric dice (VD) and surface dice (SD). Functional validation explored dosimetric effect, model failure modes, test-retest stability, and accuracy. End-user testing with eight experts assessed automated segmentations in a simulated clinical setting. FINDINGS: We included 2208 patients imaged between 2001 and 2015, with 787 patients used for model discovery and 1421 for model validation, including 28 patients for end-user testing. Models showed an improvement over the interobserver benchmark (multi-delineation dataset; VD 0·91 [IQR 0·83-0·92], p=0·0062; SD 0·86 [0·71-0·91], p=0·0005), and were within the intraobserver benchmark. For primary validation, AI performance on internal Harvard-RT1 data (segmented by the same expert who segmented the discovery data) was VD 0·83 (IQR 0·76-0·88) and SD 0·79 (0·68-0·88), within the interobserver benchmark. Performance on internal Harvard-RT2 data segmented by other experts was VD 0·70 (0·56-0·80) and SD 0·50 (0·34-0·71). Performance on RTOG-0617 clinical trial data was VD 0·71 (0·60-0·81) and SD 0·47 (0·35-0·59), with similar results on diagnostic radiology datasets NSCLC-radiogenomics and Lung-PET-CT-Dx. Despite these geometric overlap results, models yielded target volumes with equivalent radiation dose coverage to those of experts. We also found non-significant differences between de novo expert and AI-assisted segmentations. AI assistance led to a 65% reduction in segmentation time (5·4 min; p<0·0001) and a 32% reduction in interobserver variability (SD; p=0·013). INTERPRETATION: We present a clinical validation strategy for AI models. We found that in silico geometric segmentation metrics might not correlate with clinical utility of the models. Experts' segmentation style and preference might affect model performance. FUNDING: US National Institutes of Health and EU European Research Council.

Phase II Study of Nivolumab and Salvage Nivolumab/Ipilimumab in Treatment-Naive Patients With Advanced Clear Cell Renal Cell Carcinoma (HCRN GU16-260-Cohort A).

PURPOSE: To determine the value of tumor cell programmed death-ligand 1 (PD-L1) expression as a predictive biomarker of nivolumab monotherapy efficacy in treatment-naive patients with clear cell renal cell carcinoma (ccRCC) and the efficacy of salvage nivolumab/ipilimumab in patients with tumors unresponsive to nivolumab monotherapy. METHODS: Eligible patients with treatment-naive ccRCC received nivolumab until progressive disease (PD), toxicity, or completing 96 treatment weeks (part A). Patients with PD before or stable disease at 48 weeks could receive salvage nivolumab/ipilimumab (part B). The primary end point was improvement in 1-year progression-free survival in patients with tumor PD-L1 expression > 20% versus 0%. RESULTS: One hundred twenty-three patients were enrolled. The objective response rate (ORR) was 34.1% (95% CI, 25.8 to 43.2). ORR by International Metastatic RCC Database Consortium category was favorable-risk 57.1%, intermediate-risk/poor-risk 25.0%, and by sarcomatoid features 36.4%. The ORR was 26.9%, 50.0%, and 75.0% for patients with the tumor PD-L1 expression of 0, 1-20, or > 20%, respectively (trend test P value = .002). The median duration of response was 27.6 (19.3 to not reached) months, with 26 of 42 responders including 17 of 20 with favorable-risk disease remaining progression-free. The 1-year progression-free survival was 34.6% and 75.0% in the PD-L1 = 0% and > 20% categories, respectively (P = .050). Ninety-seven patients with PD or prolonged stable disease were potentially eligible for part B, and 35 were enrolled. The ORR for part B was 11.4%. Grade ≥ 3 treatment-related adverse events occurred in 35% of patients on nivolumab and 43% of those on salvage nivolumab/ipilimumab. CONCLUSION: Nivolumab monotherapy is active in treatment-naive ccRCC. Although efficacy appears to be less than that of nivolumab/ipilimumab in patients with intermediate-risk/poor-risk disease, favorable-risk patients had notable benefit. Efficacy correlated with tumor PD-L1 status. Salvage nivolumab/ipilimumab was frequently not feasible and of limited benefit.

Frequency, etiologies, risk factors, and sequelae of falls among patients with brain metastases: A population- and institutional-level analysis.

Background: Falls in patients with cancer harbor potential for serious sequelae. Patients with brain metastases (BrM) may be especially susceptible to falls but supporting investigations are lacking. We assessed the frequency, etiologies, risk factors, and sequelae of falls in patients with BrM using 2 data sources. Methods: We identified 42 648 and 111 patients with BrM utilizing Surveillance, Epidemiology, and End Results (SEER)-Medicare data (2008-2016) and Brigham and Women's Hospital/Dana-Farber Cancer Institute (BWH/DFCI) institutional data (2015), respectively, and characterized falls in these populations. Results: Among SEER-Medicare patients, 10 267 (24.1%) experienced a fall that prompted medical evaluation, with cumulative incidences at 3, 6, and 12 months of 18.0%, 24.3%, and 34.1%, respectively. On multivariable Fine/Gray's regression, older age (≥81 or 76-80 vs 66-70 years, hazard ratio [HR] 1.18 [95% CI, 1.11-1.25], P < .001 and HR 1.10 [95% CI, 1.04-1.17], P < .001, respectively), Charlson comorbidity score of >2 vs 0-2 (HR 1.08 [95% CI, 1.03-1.13], P = .002) and urban residence (HR 1.08 [95% CI, 1.01-1.16], P = .03) were associated with falls. Married status (HR 0.94 [95% CI, 0.90-0.98], P = .004) and Asian vs white race (HR 0.90 [95% CI, 0.81-0.99], P = .03) were associated with reduced fall risk. Identified falls were more common among BWH/DFCI patients (N = 56, 50.4% of cohort), resulting in emergency department visits, hospitalizations, fractures, and intracranial hemorrhage in 33%, 23%, 11%, and 4% of patients, respectively. Conclusions: Falls are common among patients with BrM, especially older/sicker patients, and can have deleterious consequences. Risk-reduction measures, such as home safety checks, physical therapy, and medication optimization, should be considered in this population.

Incidence and Predictors of Neurologic Death in Patients with Brain Metastases.

OBJECTIVE: Neurologic death is the most serious consequence of intracranial disease among patients with brain metastases. Identifying patients with brain metastases at increased risk of neurologic death can improve care and guide further research. We sought to delineate factors predictive of neurologic death among patients with brain metastases. METHODS: We identified 1218 patients with newly diagnosed brain metastases managed at Brigham and Women's Hospital/Dana-Farber Cancer Institute from 2008-2015. Factors predictive of neurologic death were assessed via univariable and multivariable Fine and Gray competing risks regression. RESULTS: On multivariable analysis, neurologic death was associated with number of brain metastases (hazard ratio [HR] 1.01 per 1 metastasis increase, 95% confidence interval [CI] 1.01-1.02, P < 0.001) and 3 primary tumor sites (reference=non-small cell lung cancer): melanoma (HR 4.67, 95% CI 3.27-6.68, P < 0.001), small cell lung cancer (HR 2.33, 95% CI 1.47-3.68, P < 0.001), and gastrointestinal cancer (HR 2.21, 95% CI 1.28-3.82, P = 0.005). Conversely, a reduction in neurologic death was found in patients with good Karnofsky performance status (90-100 vs. 30-80, HR 0.67, 95% CI 0.48-0.95, P = 0.03) and progressive extracranial metastases at diagnosis of intracranial disease (HR 0.50, 95% CI 0.38-0.67, P = 0.001). Among patients with breast primaries, HER2+ patients displayed increased neurologic death relative to the reference of HR+/HER2- (univariable analysis only: HR 2.41, 95% CI 1.00-5.84, P = 0.05). CONCLUSIONS: Patients with melanoma, small cell lung cancer, gastrointestinal cancer, and HER2+ breast cancer primaries, as well as greater intracranial versus extracranial disease burden, harbor significant risk of neurologic death. Future research investigating novel intracranial approaches should focus on these populations.