Bhandarkar, Archis

Abstract Background Biological sex and endocrine signaling influence cancer biology, immune response, and therapeutic outcomes. Recent evidence suggests that testosterone signaling may exert brain-context-dependent protective effects in glioblastoma through the hypothalamic-pituitary-adrenal axis, reduced glucocorticoid-mediated immune suppression, and altered tumor-immune interactions. We assessed whether testosterone replacement therapy (TRT) exposure was associated with survival in solid tumor central nervous system (CNS) metastases and glioblastoma (GBM, IDH-wildtype, WHO grade 4), settings in which post-diagnosis survival and TRT timing can be clinically defined. Methods We performed a retrospective Mayo Clinic cohort study of adult patients with molecularly confirmed glioblastoma and solid tumor CNS metastases confirmed from neuroimaging reports using large language model-assisted adjudication. TRT exposure was defined by testosterone-specific prescription evidence within prespecified peri-diagnostic windows. Overall survival was evaluated using propensity score-matched Cox models, 24-month administratively censored Cox models, time-dependent Cox sensitivity analyses, and 24-month restricted mean survival time. Results In the pooled solid tumor CNS metastasis cohort, TRT exposure was associated with improved overall survival after propensity score matching (HR 0.80, 95% CI 0.65–0.98, p=0.029) and a 3.22-month improvement in 24-month restricted mean survival time. In glioblastoma, TRT exposure was similarly associated with improved overall survival after propensity score matching (HR 0.56, 95% CI 0.38–0.82, p=0.003) and a 5.81-month improvement in 24-month restricted mean survival time. Conclusions TRT exposure was associated with improved survival in CNS metastases and glioblastoma. These hypothesis-generating findings support prospective studies incorporating TRT timing, hormone levels, corticosteroid exposure, immune correlates, and tumor-specific stratification. Key points TRT exposure is associated with improved survival in aggressive brain tumors. Survival associations persist after propensity score matching. Prospective studies should define TRT timing, hormone levels, and immune effects. Importance of Study This study evaluates testosterone replacement therapy (TRT) in aggressive intracranial tumors with clinically interpretable survival timelines. Building on recent evidence that androgen signaling may have protective brain-specific effects in glioblastoma, we extend this question to CNS metastases. In a multi-site Mayo Clinic cohort, TRT exposure was associated with improved survival in CNS metastases and glioblastoma, supporting androgen biology as an underrecognized factor in neuro-oncology outcomes. These findings provide rationale for prospective clinical trials incorporating TRT timing, hormone levels, corticosteroid exposure, immune profiling, and tumor-specific stratification.

Abstract Introduction Postoperative radiotherapy is an adjuvant therapy for brain metastases(BrMs) that improves local recurrence(LR). The radiosensitivity index(RSI) is a 10-gene index of pan-cancer molecular radiosensitivity previously validated for predicting radiotherapy response in clinical datasets. In this external validation study, we evaluate RSI as a predictor of local recurrence(LR) after radiotherapy in resected BrM. Methods Patients from University Health Network Toronto(UHN) and McGill University who had a resected BrM and subsequently underwent whole brain radiation(WBRT) or stereotactic radiosurgery(SRS). RSI was calculated using a ranked linear regression model from the weighted expression of 10 radiosensitivity genes, with higher values corresponding to radioresistance. Radiophenotype was determined using a previously validated RSI cutoff of < 0.375 to identify radiosensitive tumors. LR was defined as recurrence in the resection cavity and overall survival(OS) was measured from time of BrM diagnosis. Results 106 BrM patients were included, comprising 29 breast, 37 GI, 33 lung, and 7 melanoma primary sources. Mean age at surgery was 59.2 years, 31.1%(n = 33) were male, 67.9%(n = 72) had WBRT, 23.6%(n = 25) had a subtotal resection, and 30.2%(n = 32) had postoperative chemotherapy. Overall, 34.0%(n = 36) patients had LR. Patients with breast BrMs had median RSI 0.70(IQR:0.55-0.75, 10.3% radiosensitive) with 48.3% LR(n = 14); gastrointestinal BrMs had median RSI 0.43(IQR: 0.35-0.46, 29.7% radiosensitive) and 24.3% LR(n = 9); lung BrMs had median RSI of 0.67(IQR:0.66-0.71, 9% radiosensitive) and 33.3% LR(n = 11); melanoma BrMs had median RSI of 0.72(IQR:0.53-0.92, 0% radiosensitive) and 28.6% LR(n = 2). After adjusting for demographics, primary organ, radiotherapy, resection extent, among others, RSI did not predict LR(p = 0.56) or OS(p = 0.61). Discussion RSI did not predict either LR or OS in patients with surgically resected BrMs status post radiotherapy, further establishing the need for improved predictive scores in this population.

Background: Estimating operative time with low accuracy and precision leads to suboptimal scheduling of available operating rooms (ORs) and waste of healthcare resources. We aimed to develop a predictive machine learning (ML) model for operative time using our institutional data to improve the scheduling of neurosurgical ORs, minimize waste, and maximize resource utilization. Methods: We developed a predictive model using our institution’s multicenter registry, which included diverse neurosurgical cases across three tertiary hospitals. We applied multiple ML techniques, including linear regression, support vector regression, deep neural networks, and extreme gradient boosting regression (XGBR). The best model was combined with the first fit decreasing bin-packing algorithm to create a practical OR scheduling system based on prior data. Results: The XGBR model exhibited the best performance, with lower mean absolute errors (MAEs) (root mean squared error: 52.24, MAE: 37.25) and higher R2 (0.76) values than other models. Implementing this model in a simulated OR scheduling scenario using institutional data, we successfully scheduled 30 random procedures with minimal residual time, such that the model ran out of procedures for the week with 86 and 171 min remaining in the cranial and spinal ORs, respectively, demonstrating potential efficiency gains. The cranial procedures were scheduled over 4 days and spinal procedures over 6 days, optimizing available OR time. Conclusion: Integrating ML models into operative time prediction can significantly improve the accuracy and efficiency of surgical schedules. This approach minimizes wasted OR time and enhances resource allocation, potentially improving patient outcomes and satisfaction by reducing waiting times.

BACKGROUND AND OBJECTIVES: Cervical disk arthroplasty (CDA) offers the advantage of motion preservation in the treatment of focal cervical pathology. At present, implant sizing is performed using subjective tactile feedback and imaging of trial cages. This study aims to construct interpretable machine learning (IML) models to accurately predict postoperative range of motion (ROM) and identify the optimal implant sizes that maximize ROM in patients undergoing CDA. METHODS: Adult patients who underwent CDA for single-level disease from 2012 to 2020 were identified. Patient demographics, comorbidities, and outcomes were collected, including symptoms, examination findings, subsidence, and reoperation. Affected disk height, healthy rostral disk height, and implant height were collected at sequential time points. Linear regression and IML models, including bagged regression tree, bagged multivariate adaptive regression spline, and k-nearest neighbors, were used to predict ROM change. Model performance was assessed by calculating the root mean square error (RMSE) between predicted and actual changes in ROM in the validation cohort. Variable importance was assessed using RMSE loss. Area under the curve analyses were performed to identify the ideal implant size cutoffs in predicting improved ROM. RESULTS: Forty-seven patients were included. The average RMSE between predicted and actual ROM was 7.6° (range: 5.8-10.1) in the k-nearest neighbors model, 7.8° (range: 6.5-10.0) in the bagged regression tree model, 7.8° (range: 6.2-10.0) in the bagged multivariate adaptive regression spline model, and 15.8° (range: 14.3-17.5°) in a linear regression model. In the highest-performing IML model, graft size was the most important predictor with RMSE loss of 6.2, followed by age (RMSE loss = 5.9) and preoperative caudal disk height (RMSE loss = 5.8). Implant size at 110% of the normal adjacent disk height was the optimal cutoff associated with improved ROM. CONCLUSION: IML models can reliably predict change in ROM after CDA within an average of 7.6 degrees of error. Implants sized comparably with the healthy adjacent disk may maximize ROM.

Clinical prediction models for spine surgery applications are on the rise, with an increasing reliance on machine learning (ML) and deep learning (DL). Many of the predicted outcomes are uncommon; therefore, to ensure the models’ effectiveness in clinical practice it is crucial to properly evaluate them. This systematic review aims to identify and evaluate current research-based ML and DL models applied for spine surgery, specifically those predicting binary outcomes with a focus on their evaluation metrics. Overall, 60 papers were included, and the findings were reported according to the PRISMA guidelines. A total of 13 papers focused on lengths of stay (LOS), 12 on readmissions, 12 on non-home discharge, 6 on mortality, and 5 on reoperations. The target outcomes exhibited data imbalances ranging from 0.44% to 42.4%. A total of 59 papers reported the model’s area under the receiver operating characteristic (AUROC), 28 mentioned accuracies, 33 provided sensitivity, 29 discussed specificity, 28 addressed positive predictive value (PPV), 24 included the negative predictive value (NPV), 25 indicated the Brier score with 10 providing a null model Brier, and 8 detailed the F1 score. Additionally, data visualization varied among the included papers. This review discusses the use of appropriate evaluation schemes in ML and identifies several common errors and potential bias sources in the literature. Embracing these recommendations as the field advances may facilitate the integration of reliable and effective ML models in clinical settings.

OBJECTIVE High-grade gliomas (HGGs) are among the rarest yet most aggressive tumor types in neurosurgical practice. In the current literature, few studies have assessed the drivers of early outcomes following resection of these tumors and investigated their association with quality of care. The authors aimed to identify the clinical predictors for 30-day readmission and reoperation following HGG surgery using the American College of Surgeons (ACS) National Surgical Quality Improvement Project (NSQIP) database and sought to create web-based applications predicting each outcome. METHODS Using the ACS NSQIP database, the authors conducted a retrospective, multicenter cohort analysis of patients who underwent resection of supratentorial HGGs between January 1, 2016, and December 31, 2020. Demographics and comorbidities were extracted. The primary outcomes were 30-day unplanned readmission and reoperation. A stratified 80:20 split of the available data was carried out. Supervised machine learning algorithms were trained to predict 30-day outcomes. RESULTS A total of 9418 patients were included in our cohort. The observed rate of unplanned readmission within 30 days of surgery was 13.0% (n = 1221). In terms of predictors, weight, chronic steroid use, preoperative blood urea nitrogen level, and white blood cell count were associated with a higher risk of readmission. The observed rate of unplanned reoperation within 30 days of surgery was 5.2% (n = 489). In terms of predictors, increased weight, longer operative time, and more days between hospital admission and operation were associated with an increased risk of early reoperation. The random forest algorithm showed the highest predictive performance for early readmission (area under the curve [AUC] = 0.967), while the XGBoost algorithm showed the highest predictive performance for early reoperation (AUC = 0.985). Web-based tools for both outcomes were deployed ( https://glioma-readmission.herokuapp.com/ and https://glioma-reoperation.herokuapp.com/ ). CONCLUSIONS In this study, the authors provide the first nationwide analysis for short-term outcomes in patients undergoing resection of supratentorial HGGs. Multiple patient, hospital, and admission factors were associated with readmission and reoperation, confirmed by machine learning predicting patients’ prognosis, leading to better planning preoperatively and subsequently improved personalized patient care.