Faculty Bibliography

BACKGROUND AND OBJECTIVES/OBJECTIVE:Bone metastases, affecting more than 4.8% of patients with cancer annually, and particularly spinal metastases require urgent intervention to prevent neurological complications. However, the current process of manually reviewing radiological reports leads to potential delays in specialist referrals. We hypothesized that natural language processing (NLP) review of routine radiology reports could automate the referral process for timely multidisciplinary care of spinal metastases. METHODS:We assessed 3 NLP models-a rule-based regular expression (RegEx) model, GPT-4, and a specialized Bidirectional Encoder Representations from Transformers (BERT) model (NYUTron)-for automated detection and referral of bone metastases. Study inclusion criteria targeted patients with active cancer diagnoses who underwent advanced imaging (computed tomography, MRI, or positron emission tomography) without previous specialist referral. We defined 2 separate tasks: task of identifying clinically significant bone metastatic terms (lexical detection), and identifying cases needing a specialist follow-up (clinical referral). Models were developed using 3754 hand-labeled advanced imaging studies in 2 phases: phase 1 focused on spine metastases, and phase 2 generalized to bone metastases. Standard McRae's line performance metrics were evaluated and compared across all stages and tasks. RESULTS:In the lexical detection, a simple RegEx achieved the highest performance (sensitivity 98.4%, specificity 97.6%, F1 = 0.965), followed by NYUTron (sensitivity 96.8%, specificity 89.9%, and F1 = 0.787). For the clinical referral task, RegEx also demonstrated superior performance (sensitivity 92.3%, specificity 87.5%, and F1 = 0.936), followed by a fine-tuned NYUTron model (sensitivity 90.0%, specificity 66.7%, and F1 = 0.750). CONCLUSION/CONCLUSIONS:An NLP-based automated referral system can accurately identify patients with bone metastases requiring specialist evaluation. A simple RegEx model excels in syntax-based identification and expert-informed rule generation for efficient referral patient recommendation in comparison with advanced NLP models. This system could significantly reduce missed follow-ups and enhance timely intervention for patients with bone metastases.

BACKGROUND AND PURPOSE/OBJECTIVE:Skeletal-related events (SRE) are a major source of morbidity and mortality across cancer types. Identification of risk factors for SRE and association with survival would facilitate more targeted preventive treatment. MATERIALS AND METHODS/METHODS:This retrospective cohort study included patients with bone metastases from solid tumors undergoing systemic imaging from February-March 2022 who had not received radiation within one year. Survival was analyzed using Cox models, and multi-state models assessed factors linked to SRE with death as a competing risk. Outcomes were SRE (including radiation for pain) and all-cause death. Variables included tumor type, metastasis site, and trial eligibility. RESULTS:Among 410 patients (median age 67 years; 48 % male), 162 (40 %) experienced SRE over a median follow-up of 26.8 months. Seventy-five (18.3 %) received radiation for pain alone. Experiencing any type of SRE (HR 1.98, 95 % CI 1.47-2.67, p < 0.001) or radiation for pain alone (HR 2.14, 95 % CI 1.57-2.92, p < 0.001) were both associated with increased mortality. Patients eligible for a trial of early radiation were more likely to develop SRE (HR 1.67, 95 % CI 1.18-2.37, p = 0.004). Prostate cancer histology (HR 1.70, p = 0.02) and metastases to the hip/acetabulum (HR 2.55, p = 0.02) were associated with SRE. CONCLUSION/CONCLUSIONS:Patients treated with radiation for pain alone demonstrated similar risk of death as those experiencing any type of SRE, supporting the inclusion of radiation in endpoint definitions. Prostate cancer type and hip/acetabulum metastasis location may help identify patients and lesions at elevated SRE risk, informing future preventive strategies.

IMPORTANCE/UNASSIGNED:Stereotactic body radiation therapy (SBRT) for spinal metastases improves symptomatic outcomes and local control compared to conventional radiotherapy. Treatment failure most often occurs within the epidural space, where dose is constrained by the risk of radiation myelitis (RM). Current constraints designed to prevent RM after spine SBRT are derived from limited data. OBJECTIVE/UNASSIGNED:To characterize the risk of RM after spine SBRT and to update the dosimetric constraints for preventing it. DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:This cohort study was conducted in a single tertiary cancer care center with patients treated for spinal metastases from 2014 to 2023. All included participants had undergone spine SBRT, had a minimum of 1-month follow-up with magnetic resonance imaging (MRI), a maximal cord dose to a voxel (Dmax) greater than 0 Gy, and no overlapping prior radiotherapy. In all, 2051 patients received SBRT to 2835 spinal metastases (levels C1-L2) during the study period. EXPOSURES/UNASSIGNED:Three-fraction spine SBRT to a prescription dose of 27 to 36 Gy. MAIN OUTCOMES AND MEASURES/UNASSIGNED:RM defined as radiographic evidence of spinal cord injury in the treatment field, classified as grade (G) 1 to G4 or G3 to G4 per the Common Terminology Criteria for Adverse Events, version 5.0. Multiple dosimetric parameters of the true spinal cord structure were assessed for an association with risk of RM to determine the important covariates associated with this toxicity. RESULTS/UNASSIGNED:The analysis included 1423 patients (mean [SD] age, 61.6 [12.9] years; 695 [48.8%] females and 728 [51.1%] males) who received SBRT for 1904 spinal metastases. Among them, 30 cases of RM were identified, 19 of which were classified as G3 to G4. Two years after SBRT, the rate of G1 to G4 RM was 1.8% (95% CI, 1.2%-2.5%) and the rate of G3 to G4 RM was 1.1% (95% CI, 0.7%-1.7%). The minimum dose to the 0.1 cm3 of spinal cord receiving the greatest dose (D0.1cc) was the most important covariate on univariable cause-specific hazards regression for RM (for G3 to G4: hazard ratio, 2.14; 95% CI, 1.68-2.72; P < .001). A true cord D0.1cc of 19.1 Gy and Dmax of 20.8 Gy estimated a 1.0% risk (95% CI, 0.3%-1.6% and 0.4%-1.6%, respectively) of G3 to G4 RM 2 years after SBRT. CONCLUSIONS AND RELEVANCE/UNASSIGNED:The findings of this cohort study indicate that a cord (myelogram or MRI-derived) D0.1cc constraint of 19.1 Gy and a Dmax constraint of 20.8 Gy correspond with a 1.0% risk of G3 to G4 RM at 2 years.

There is a much debate regarding optimal selection in patients with metastatic cancer who should undergo local treatment (surgery or radiation treatment) to the primary tumor and/or metastases. Additionally, the optimal treatment of newly diagnosed metastatic cancer is largely unclear. Current prognostication systems to best inform these clinical scenarios are limited, as all metastatic patients are grouped together as having Stage IV disease without further incorporation of patient and disease-specific covariates that significantly impact patient outcomes. Therefore, improving current prognostic scoring systems and incorporation of these covariates is essential to best individualize treatment for patients with metastatic cancer. In this narrative review article, we provide a detailed review of prognostication systems that can be used for both the site of metastasis and primary site to best tailor treatment in these patients. Additionally, we discuss the incorporation and ongoing developments in radiographic, genomic, and biostatistical techniques that can be used as prognostication tools.

Leptomeningeal metastases are increasingly becoming recognized as a treatable, yet generally incurable, complication of advanced cancer. As modern cancer therapeutics have prolonged the lives of patients with metastatic cancer, specifically in patients with parenchymal brain metastases, treatment options and clinical research protocols for patients with leptomeningeal metastases from solid tumors have similarly evolved to improve survival within specific populations. Recent expansion in clinical investigation, early diagnosis, and drug development have given rise to new unanswered questions. These include leptomeningeal metastasis biology and preferred animal modeling, epidemiology in the modern cancer population, ensuring validation and accessibility of newer leptomeningeal metastasis diagnostics, best clinical practices with multi-modality treatment options, clinical trial design and standardization of response assessments, and avenues worthy of further research. An international group of multi-disciplinary experts in the research and management of leptomeningeal metastases, supported by the Society for Neuro-Oncology and American Society of Clinical Oncology, were assembled to reach a consensus opinion on these pressing topics and provide a roadmap for future directions. Our hope is that these recommendations will accelerate collaboration and progress in the field of leptomeningeal metastases and serve as a platform for further discussion and patient advocacy.

IMPORTANCE/UNASSIGNED:For patients with nonspine bone metastases, short-course radiotherapy (RT) can reduce patient burden without sacrificing clinical benefit. However, there is great variation in uptake of short-course RT across practice settings. OBJECTIVE/UNASSIGNED:To evaluate whether a set of 3 implementation strategies facilitates increased adoption of a consensus recommendation to treat nonspine bone metastases with short-course RT (ie, ≤5 fractions). DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:This prospective, stepped-wedge, cluster randomized quality improvement study was conducted at 3 community-based cancer centers within an existing academic-community partnership. Rollout was initiated in 3-month increments between October 2021 and May 2022. Participants included treating physicians and patients receiving RT for nonspine bone metastases. Data analysis was performed from October 2022 to May 2023. EXPOSURES/UNASSIGNED:Three implementation strategies-(1) dissemination of published consensus guidelines, (2) personalized audit-and-feedback reports, and (3) an email-based electronic consultation platform (eConsult)-were rolled out to physicians. MAIN OUTCOMES AND MEASURES/UNASSIGNED:The primary outcome was adherence to the consensus recommendation of short-course RT for nonspine bone metastases. Mixed-effects logistic regression at the bone metastasis level was used to model associations between the exposure of physicians to the set of strategies (preimplementation vs postimplementation) and short-course RT, while accounting for patient and physician characteristics and calendar time, with a random effect for physician. Physician surveys were administered before implementation and after implementation to assess feasibility, acceptability, and appropriateness of each strategy. RESULTS/UNASSIGNED:Forty-five physicians treated 714 patients (median [IQR] age at treatment start, 67 [59-75] years; 343 women [48%]) with 838 unique nonspine bone metastases during the study period. Implementing the set of strategies was not associated with use of short-course RT (odds ratio, 0.78; 95% CI, 0.45-1.34; P = .40), with unadjusted adherence rates of 53% (444 lesions) preimplementation vs 56% (469 lesions) postimplementation; however, the adjusted odds of adherence increased with calendar time (odds ratio, 1.68; 95% CI, 1.20-2.36; P = .003). All 3 implementation strategies were perceived as being feasible, acceptable, and appropriate; only the perception of audit-and-feedback appropriateness changed before vs after implementation (19 of 29 physicians [66%] vs 27 of 30 physicians [90%]; P = .03, Fisher exact test), with 20 physicians (67%) preferring reports quarterly. CONCLUSIONS AND RELEVANCE/UNASSIGNED:In this quality improvement study, a multicomponent set of implementation strategies was not associated with increased use of short-course RT within an academic-community partnership. However, practice improved with time, perhaps owing to secular trends or physician awareness of the study. Audit-and-feedback was more appropriate than anticipated. Findings support the need to investigate optimal approaches for promoting evidence-based radiation practice across settings.

PURPOSE/OBJECTIVE:Treatment decisions for leptomeningeal disease (LMD) rely on patient risk stratification, since clinicians lack objective prognostic tools. The introduction of rare cell capture technology for identification of cerebrospinal fluid tumor cells (CSF-TCs), such as CNSide assay, improved the sensitivity of LMD diagnosis, but prognostic value is unknown. This study assesses the prognostic value of CSF-TC density in patients with LMD from solid tumors. METHODS:We conducted a retrospective cohort study of patients with newly diagnosed or previously treated LMD from a single institution who had CNSide assay testing for CSF-TCs from 2020 to 2023. Univariable and multivariable survival analyses were conducted with Cox proportional-hazards modeling. Maximally-selected rank statistics were used to determine an optimal cutpoint for CSF-TC density and survival. RESULTS:Of 31 patients, 29 had CSF-TCs detected on CNSide. Median (interquartile range [IQR]) CSF-TC density was 67.8 (4.7-639) TCs/mL. CSF cytology was positive in 16 of 29 patients with positive CNSide (CNSide diagnostic sensitivity = 93.5%, negative predictive value = 85.7%). Median (IQR) survival from time of CSF-TC detection was 176 (89-481) days. On univariable and multivariable analysis, CSF-TC density was significantly associated with survival. An optimal cutpoint for dichotomizing survival by CSF-TC density was 19.34 TCs/mL. The time-dependent sensitivity and specificity for survival using this stratification were 76% and 67% at 6 months and 65% and 67% at 1 year, respectively. CONCLUSIONS:CSF-TC density may carry prognostic value in patients with LMD from solid tumors. Integrating CSF-TC density into LMD patient risk-stratification may help guide treatment decisions.

This study explores the feasibility and potential dosimetric and time efficiency benefit of proton Pencil Beam Scanning (PBS) craniospinal irradiation with a single posterior-anterior (SPA) brain field. The SPA approach was compared to our current clinical protocol using Bilateral Posterior Oblique brain fields (BPO). Ten consecutive patients were simulated in the head-first supine position on a long BOS frame and scanned using 3 mm CT slice thickness. A customized thermoplastic mask immobilized the patient's head, neck, and shoulders. A vac-lock was used to secure the legs. PBS proton plans were robustly optimized with 3mm setup errors and 3.5% range uncertainties in the Eclipse V15.6 treatment planning system (n = 12 scenarios). In order to achieve a smooth gradient dose match at the junction area, at least 5 cm overlap region was maintained between the segments and 5 mm uncertainty along the cranial-cauda direction was applied to each segment independently as additional robust optimization scenarios. The brain doses were planned by SPA or BPO fields. All spine segments were planned with a single PA field. Dosimetric differences between the BPO and SPA approaches were compared, and the treatment efficiency was analyzed according to timestamps of beam delivery. Results: The maximum brain dose increases to 111.1 ± 2.1% for SPA vs. 109.0 ± 1.7% for BPO (p < 0.01). The dose homogeneity index (D5/D95) in brain CTV was comparable between techniques (1.037 ± 0.010 for SPA and 1.033 ± 0.008 for BPO). Lens received lower maximum doses by 2.88 ± 1.58 Gy (RBE) (left) and 2.23 ± 1.37 Gy (RBE) (right) in the SPA plans (p < 0.01). No significant cochlea dose change was observed. SPA reduced the treatment time by more than 4 minutes on average and ranged from 2 to 10 minutes, depending on the beam waiting and allocation time. SPA is dosimetrically comparable to BPO, with reduced lens doses at the cost of slightly higher dose inhomogeneity and hot spots. Implementation of SPA is feasible and can help to improve the treatment efficiency of PBS CSI treatment.