Extraction of Unstructured Electronic Health Records to Evaluate Glioblastoma Treatment Patterns.

PURPOSE: Data on lines of therapy (LOTs) for cancer treatment are important for clinical oncology research, but LOTs are not explicitly recorded in electronic health records (EHRs). We present an efficient approach for clinical data abstraction and a flexible algorithm to derive LOTs from EHR-based medication data on patients with glioblastoma multiforme (GBM). METHODS: Nonclinicians were trained to abstract the diagnosis of GBM from EHRs, and their accuracy was compared with abstraction performed by clinicians. The resulting data were used to build a cohort of patients with confirmed GBM diagnosis. An algorithm was developed to derive LOTs using structured medication data, accounting for the addition and discontinuation of therapies and drug class. Descriptive statistics were calculated and time-to-next-treatment (TTNT) analysis was performed using the Kaplan-Meier method. RESULTS: Treating clinicians as the gold standard, nonclinicians abstracted GBM diagnosis with a sensitivity of 0.98, specificity 1.00, positive predictive value 1.00, and negative predictive value 0.90, suggesting that nonclinician abstraction of GBM diagnosis was comparable with clinician abstraction. Of 693 patients with a confirmed diagnosis of GBM, 246 patients contained structured information about the types of medications received. Of them, 165 (67.1%) received a first-line therapy (1L) of temozolomide, and the median TTNT from the start of 1L was 179 days. CONCLUSION: We described a workflow for extracting diagnosis of GBM and LOT from EHR data that combines nonclinician abstraction with algorithmic processing, demonstrating comparable accuracy with clinician abstraction and highlighting the potential for scalable and efficient EHR-based oncology research.

Ultrasonic cerebrospinal fluid clearance improves outcomes in hemorrhagic brain injury models.

Impaired clearance of the byproducts of aging and neurologic disease from the brain exacerbates disease progression and severity. We have developed a noninvasive, low intensity transcranial focused ultrasound protocol that facilitates the removal of pathogenic substances from the cerebrospinal fluid (CSF) and the brain interstitium. This protocol clears neurofilament light chain (NfL) - an aging byproduct - in aged mice and clears red blood cells (RBCs) from the central nervous system in two mouse models of hemorrhagic brain injury. Cleared RBCs accumulate in the cervical lymph nodes from both the CSF and interstitial compartments, indicating clearance through meningeal lymphatics. Treating these hemorrhagic brain injury models with this ultrasound protocol reduced neuroinflammatory and neurocytotoxic profiles, improved behavioral outcomes, decreased morbidity and, importantly, increased survival. RBC clearance efficacy was blocked by mechanosensitive channel antagonism and was effective when applied in anesthetized subjects, indicating a mechanosensitive channel mediated mechanism that does not depend on sensory stimulation or a specific neural activity pattern. Notably, this protocol qualifies for an FDA non-significant risk designation given its low intensity, making it readily clinically translatable. Overall, our results demonstrate that this low-intensity transcranial focused ultrasound protocol clears hemorrhage and other harmful substances from the brain via the meningeal lymphatic system, potentially offering a novel therapeutic tool for varied neurologic disorders.

Selective prediction for extracting unstructured clinical data.

OBJECTIVE: While there are currently approaches to handle unstructured clinical data, such as manual abstraction and structured proxy variables, these methods may be time-consuming, not scalable, and imprecise. This article aims to determine whether selective prediction, which gives a model the option to abstain from generating a prediction, can improve the accuracy and efficiency of unstructured clinical data abstraction. MATERIALS AND METHODS: We trained selective classifiers (logistic regression, random forest, support vector machine) to extract 5 variables from clinical notes: depression (n = 1563), glioblastoma (GBM, n = 659), rectal adenocarcinoma (DRA, n = 601), and abdominoperineal resection (APR, n = 601) and low anterior resection (LAR, n = 601) of adenocarcinoma. We varied the cost of false positives (FP), false negatives (FN), and abstained notes and measured total misclassification cost. RESULTS: The depression selective classifiers abstained on anywhere from 0% to 97% of notes, and the change in total misclassification cost ranged from -58% to 9%. Selective classifiers abstained on 5%-43% of notes across the GBM and colorectal cancer models. The GBM selective classifier abstained on 43% of notes, which led to improvements in sensitivity (0.94 to 0.96), specificity (0.79 to 0.96), PPV (0.89 to 0.98), and NPV (0.88 to 0.91) when compared to a non-selective classifier and when compared to structured proxy variables. DISCUSSION: We showed that selective classifiers outperformed both non-selective classifiers and structured proxy variables for extracting data from unstructured clinical notes. CONCLUSION: Selective prediction should be considered when abstaining is preferable to making an incorrect prediction.

Noninvasive ultrasonic induction of cerebrospinal fluid flow enhances intrathecal drug delivery.

Intrathecal drug delivery is routinely used in the treatment and prophylaxis of varied central nervous system conditions, as doing so allows drugs to directly bypass the blood-brain barrier. However, the utility of this route of administration is limited by poor brain and spinal cord parenchymal drug uptake from the cerebrospinal fluid. We demonstrate that a simple noninvasive transcranial ultrasound protocol can significantly increase influx of cerebrospinal fluid into the perivascular spaces of the brain, to enhance the uptake of intrathecally administered drugs. Specifically, we administered small (~1 kDa) and large (~155 kDa) molecule agents into the cisterna magna of rats and then applied low, diagnostic-intensity focused ultrasound in a scanning protocol throughout the brain. Using real-time magnetic resonance imaging and ex vivo histologic analyses, we observed significantly increased uptake of small molecule agents into the brain parenchyma, and of both small and large molecule agents into the perivascular space from the cerebrospinal fluid. Notably, there was no evidence of brain parenchymal damage following this intervention. The low intensity and noninvasive approach of transcranial ultrasound in this protocol underscores the ready path to clinical translation of this technique. In this manner, this protocol can be used to directly bypass the blood-brain barrier for whole-brain delivery of a variety of agents. Additionally, this technique can potentially be used as a means to probe the causal role of the glymphatic system in the variety of disease and physiologic processes to which it has been correlated.

Clinical landscape of oncolytic virus research in 2020.

Oncolytic viruses (OVs) are a new class of cancer therapeutics. This review was undertaken to provide insight into the current landscape of OV clinical trials. A PubMed search identified 119 papers from 2000 to 2020 with 97 studies reporting data on 3233 patients. The viruses used, presence of genetic modifications and/or transgene expression, cancer types targeted, inclusion of combination strategies and safety profile were reported. In addition, information on viral bioshedding across the studies, including which tissues or body fluids were evaluated and how virus was detected (eg, PCR, plaque assay or both), is also reported. Finally, the number of studies evaluating antiviral and antitumor humoral and cellular immune responses were noted. We found that adenovirus (n=30) is the most common OV in clinical trials with approximately two-thirds (n=63) using modified or recombinant viral backbones and granulocyte-macrophage colony-stimulating factor (n=24) was the most common transgene. The most common tumors targeted were melanoma (n=1000) and gastrointestinal (GI; n=577) cancers with most using monotherapy OVs given by intratumoral (n=1482) or intravenous (n=1347) delivery. The most common combination included chemotherapy (n=36). Overall, OV treatment-related adverse events were low-grade constitutional and local injection site reactions. Viral shedding was frequently measured although many studies restricted this to blood and tumor tissue and used PCR only. While most studies did report antiviral antibody titers (n=63), only a minority of studies reported viral-specific T cell responses (n=10). Tumor immunity was reported in 48 studies and largely relied on general measures of immune activation (eg, tumor biopsy immunohistochemistry (n=25) and serum cytokine measurement (n=19)) with few evaluating tumor-specific immune responses (n=7). Objective responses were reported in 292 (9%) patients and disease control was achieved in 681 (21.1%) patients, although standard reporting criteria were only used in 53% of the trials. Completed clinical trials not reported in the peer-reviewed literature were not included in this review potentially underestimating the impact of OV treatment. These data provide insight into the current profile of OV clinical trials reporting and identifies potential gaps where further studies are needed to better define the role of OVs, alone and in combination, for patients with cancer.