Using Recurrent Neural Networks to Extract High-Quality Information From Lung Cancer Screening Computerized Tomography Reports for Inter-Radiologist Audit and Feedback Quality Improvement.

PURPOSE: Lung cancer screening programs generate a high volume of low-dose computed tomography (LDCT) reports that contain valuable information, typically in a free-text format. High-performance named-entity recognition (NER) models can extract relevant information from these reports automatically for inter-radiologist quality control. METHODS: Using LDCT report data from a longitudinal lung cancer screening program (8,305 reports; 3,124 participants; 2006-2019), we trained a rule-based model and two bidirectional long short-term memory (Bi-LSTM) NER neural network models to detect clinically relevant information from LDCT reports. Model performance was tested using F1 scores and compared with a published open-source radiology NER model (Stanza) in an independent evaluation set of 150 reports. The top performing model was applied to a data set of 6,948 reports for an inter-radiologist quality control assessment. RESULTS: The best performing model, a Bi-LSTM NER recurrent neural network model, had an overall F1 score of 0.950, which outperformed Stanza (F1 score = 0.872) and a rule-based NER model (F1 score = 0.809). Recall (sensitivity) for the best Bi-LSTM model ranged from 0.916 to 0.991 for different entity types; precision (positive predictive value) ranged from 0.892 to 0.997. Test performance remained stable across time periods. There was an average of a 2.86-fold difference in the number of identified entities between the most and the least detailed radiologists. CONCLUSION: We built an open-source Bi-LSTM NER model that outperformed other open-source or rule-based radiology NER models. This model can efficiently extract clinically relevant information from lung cancer screening computerized tomography reports with high accuracy, enabling efficient audit and feedback to improve quality of patient care.

Exploring social-emotional learning, school climate, and social network analysis.

There is currently limited research on student peer leadership in the social-emotional literature. This paper used exploratory methods of social network analysis to understand the structure of school peer relationships, peer leadership, and school climate. Self-report measures of perceptions of peer leadership and climate were given to students during the 2016-2017 school year. Data collected from a peer leadership survey were used to calculate closeness and indegree centrality values. The results showed that student Ambassadors have higher peer nominated leadership scores compared to non-Ambassador controls and the rest of the school. Additionally, Ambassadors did not demonstrate a change in centrality scores, non-Ambassador students increased in centrality scores, and school climate was not correlated with the leadership centrality score. Results suggest that influence spreads, and that good leadership may be emulated among students, leading to a diffusion effect. This supports the need for good leaders in schools. Additionally, climate may not be associated with leadership centrality scores due to the length of the intervention. Future studies should look toward behavioral data to unravel what comprises positive and negative influences in Social-Emotional and Character Development interventions.

Association of post-treatment longitudinal symptom severity clusters with subsequent survival in oropharyngeal cancer.

BACKGROUND: Patients with cancer often experience multiple symptoms concurrently. We identified patient clusters based on longitudinal symptom severity trajectories in oropharyngeal cancer (OPC) and evaluated the potential clinical utility of this approach. METHODS: A retrospective OPC patient cluster analysis using 6 months of symptom severity data from radiotherapy initiation. The clinico-demographic characteristics and overall survival of patients were compared between clusters. RESULTS: We identified four clusters of patients differing in longitudinal symptom severity. Cluster A (n = 168) included patients with the mildest longitudinal symptoms, cluster B (n = 59) and cluster C (n = 63) were intermediate, and cluster D (n = 30) included patients with the worst symptoms. The clusters differed in their HPV status, ECOG performance status, smoking history, drinking history, treatment modality, and 5-year survival. These clusters separated symptom severity trajectories more distinctly than individual clinico-demographic characteristics. CONCLUSIONS: Early symptom severity trajectory clustering revealed distinct patient clusters that were prognostic of overall survival.

Oncolytic viruses in melanoma.

Malignant melanoma recurrence remains heterogeneous in presentation, ranging from locoregional disease (i.e., local recurrence, satellites, in transit disease) to distant dermal and visceral metastases. This diverse spectrum of disease requires a personalized approach to management and has resulted in the development of both local (e.g., surgery, radiation, intralesional injection) and systemic (intravenous or oral) treatment strategies. Intralesional agents such as oncolytic viruses may also evoke local immune stimulation to induce and enhance the antitumor immune response. Further, it is hypothesized that these oncolytic viruses may convert immunologically "cold" tumors to more reactive "hot" tumor microenvironments and thereby overcome anti-PD-1 therapy resistance. Currently, talimogene laherparepvec (T-VEC), a modified herpes virus, is FDA-approved in this population, with many other oncolytic viruses under investigation in both preclinical and trial settings. Herein, we detail the scientific rationale, current landscape, and future directions of oncolytic viruses in melanoma.

Endoscopic resection with adjuvant treatment versus esophagectomy for early-stage esophageal cancer.

OBJECTIVE: To evaluate the outcome following the strategy of endoscopic R0 resection (ER) plus adjuvant treatment (AT) versus esophagectomy for esophageal squamous cell cancer in T1a invading muscularis mucosa (M3)-T1b stage. METHODS: We evaluated the outcomes of 46 esophageal squamous cell cancer (ESCC) patients with T1aM3-T1b stage who underwent ER + AT from the Esophageal Cancer Endoscopic Therapy Consortium (ECETC) and compared these outcomes to 92 patients who underwent esophagectomy. Propensity score matching (1:2) was used, with overall survival (OS) and relapse-free survival (RFS) being compared between the two groups. RESULTS: During a median follow-up of 32 months, there were no statistical differences (P = 0.226) in OS between the two groups. The 1-, 2-, and 3-year overall survival in the esophagectomy group was 95%, 91%, and 84%, respectively. There were no mortalities within three years in the ER + AT group. The RFS between the two groups was also not significantly different (P = 0.938). The 1-, 2-, and 3-year RFS of patients in the esophagectomy group was 90%, 90%, and 83%, respectively, while it was 97%, 94%, and 74% in the ER + AT group, respectively. The local recurrence rates between the two groups were not significantly different (P = 0.277). CONCLUSIONS: This first multicenter analysis showed similar outcomes were found regarding OS and RFS between the two groups in T1aM3-T1b stage patients. ER + AT may be considered in high-risk patients or for those who refuse esophagectomy.

Longitudinal Assessment of Health Utility Scores, Symptoms and Toxicities in Patients with Small Cell Lung Cancer Using Real World Data.

INTRODUCTION: Recent advances in small cell lung cancer (SCLC) treatments necessitate a better understanding of real-world health utility scores (HUS) in patients treated under current standards to facilitate robust pharmaco-economic assessments. METHODS: In this single institution cohort observational study, HUS were evaluated in patients with SCLC through EQ-5D questionnaires at outpatient visits (encounters). In addition, patients completed questionnaires relating to treatment toxicities and cancer symptoms. Clinical and pathological variables were abstracted from electronic medical records and disease status at each patient visit was documented. The impact of these variables on HUS were explored. RESULTS: There were 282 clinical encounters (12% newly diagnosed; 37% stable on treatment; 22% progressing on treatment; 29% stable off therapy/other) in 111 SCLC patients (58% male; 64% extensive stage (ES) SCLC). At the first encounter 29% of patients had an ECOG performance status (PS) ≥ 2. ES-SCLC, bone metastases, female sex, progressive disease and/or PS were each significantly associated with decreased HUS in multivariable analyses. Patients clinically stable on first line therapy had generally steady HUS longitudinally, with differences in HUS between limited disease (LD) and ES patients emerging as treatment progressed. Decreased HUS were associated with increased severity of the majority of measured symptoms (fatigue/tiredness, loss of appetite, pain, drowsiness, shortness of breath, anxiety, depression, and overall well-being; each p<0.001), supporting the value of EQ-5D-derived HUS in assessing health utility. CONCLUSION: Our HUS values in chemotherapy-treated SCLC are clinically relevant and are associated with specific clinico-demographic, symptom and toxicity factors.

Combined EZH2 Inhibition and IKAROS Degradation Leads to Enhanced Antitumor Activity in Diffuse Large B-cell Lymphoma.

PURPOSE: The efficacy of EZH2 inhibition has been modest in the initial clinical exploration of diffuse large B-cell lymphoma (DLBCL), yet EZH2 inhibitors are well tolerated. Herein, we aimed to uncover genetic and pharmacologic opportunities to enhance the clinical efficacy of EZH2 inhibitors in DLBCL. EXPERIMENTAL DESIGN: We conducted a genome-wide sensitizing CRISPR/Cas9 screen with tazemetostat, a catalytic inhibitor of EZH2. The sensitizing effect of IKZF1 loss of function was then validated and leveraged for combination treatment with lenalidomide. RNA sequencing (RNA-seq) and chromatin immunoprecipitation sequencing analyses were performed to elucidate transcriptomic and epigenetic changes underlying synergy. RESULTS: We identified IKZF1 knockout as the top candidate for sensitizing DLBCL cells to tazemetostat. Treating cells with tazemetostat and lenalidomide, an immunomodulatory drug that selectively degrades IKAROS and AIOLOS, phenocopied the effects of the CRISPR/Cas9 screen. The combined drug treatment triggered either cell-cycle arrest or apoptosis in a broad range of DLBCL cell lines, regardless of EZH2 mutational status. Cell-line-based xenografts also showed slower tumor growth and prolonged survival in the combination treatment group. RNA-seq analysis revealed strong upregulation of interferon signaling and antiviral immune response signatures. Gene expression of key immune response factors such as IRF7 and DDX58 were induced in cells treated with lenalidomide and tazemetostat, with a concomitant increase of H3K27 acetylation at their promoters. Furthermore, transcriptome analysis demonstrated derepression of endogenous retroviruses after combination treatment. CONCLUSIONS: Our data underscore the synergistic interplay between IKAROS degradation and EZH2 inhibition on modulating epigenetic changes and ultimately enhancing antitumor effects in DLBCL.

Costimulation Induces CD4 T Cell Antitumor Immunity via an Innate-like Mechanism.

Chronic exposure to tumor-associated antigens inactivates cognate T cells, restricting the repertoire of tumor-specific effector T cells. This problem was studied here by transferring TCR transgenic CD4 T cells into recipient mice that constitutively express a cognate self-antigen linked to MHC II on CD11c-bearing cells. Immunotherapeutic agonists to CD134 plus CD137, "dual costimulation," induces specific CD4 T cell expansion and expression of the receptor for the Th2-associated IL-1 family cytokine IL-33. Rather than producing IL-4, however, they express the tumoricidal Th1 cytokine IFNγ when stimulated with IL-33 or IL-36 (a related IL-1 family member) plus IL-12 or IL-2. IL-36, which is induced within B16-F10 melanomas by dual costimulation, reduces tumor growth when injected intratumorally as a monotherapy and boosts the efficacy of tumor-nonspecific dual costimulated CD4 T cells. Dual costimulation thus enables chronic antigen-exposed CD4 T cells, regardless of tumor specificity, to elaborate tumoricidal function in response to tumor-associated cytokines.

Direct CD137 costimulation of CD8 T cells promotes retention and innate-like function within nascent atherogenic foci.

Effector CD8 T cells infiltrate atherosclerotic lesions and are correlated with cardiovascular events, but the mechanisms regulating their recruitment and retention are not well understood. CD137 (4-1BB) is a costimulatory receptor induced on immune cells and expressed at sites of human atherosclerotic plaque. Genetic variants associated with decreased CD137 expression correlate with carotid-intimal thickness and its deficiency in animal models attenuates atherosclerosis. These effects have been attributed in part to endothelial responses to low and disturbed flow (LDF), but CD137 also generates robust effector CD8 T cells as a costimulatory signal. Thus, we asked whether CD8 T cell-specific CD137 stimulation contributes to their infiltration, retention, and IFNγ production in early atherogenesis. We tested this through adoptive transfer of CD8 T cells into recipient C57BL/6J mice that were then antigen primed and CD137 costimulated. We analyzed atherogenic LDF vessels in normolipidemic and PCSK9-mediated hyperlipidemic models and utilized a digestion protocol that allowed for lesional T-cell characterization via flow cytometry and in vitro stimulation. We found that CD137 activation, specifically of effector CD8 T cells, triggers their intimal infiltration into LDF vessels and promotes a persistent innate-like proinflammatory program. Residence of CD137+ effector CD8 T cells further promoted infiltration of endogenous CD8 T cells with IFNγ-producing potential, whereas CD137-deficient CD8 T cells exhibited impaired vessel infiltration, minimal IFNγ production, and reduced infiltration of endogenous CD8 T cells. Our studies thus provide novel insight into how CD137 costimulation of effector T cells, independent of plaque-antigen recognition, instigates their retention and promotes innate-like responses from immune infiltrates within atherogenic foci. NEW & NOTEWORTHY Our studies identify CD137 costimulation as a stimulus for effector CD8 T-cell infiltration and persistence within atherogenic foci, regardless of atherosclerotic-antigen recognition. These costimulated effector cells, which are generated in pathological states such as viral infection and autoimmunity, have innate-like proinflammatory programs in circulation and within the atherosclerotic microenvironment, providing mechanistic context for clinical correlations of cardiovascular morbidity with increased CD8 T-cell infiltration and markers of activation in the absence of established antigen specificity.

Activated T-effector seeds: cultivating atherosclerotic plaque through alternative activation.

Atherosclerosis is a chronic inflammatory pathology that precipitates substantial morbidity and mortality. Although initiated by physiological patterns of low and disturbed flow that differentially prime endothelial cells at sites of vessel branch points and curvature, the chronic, smoldering inflammation of atherosclerosis is accelerated by comorbidities involving inappropriate activation of the adaptive immune system, such as autoimmunity. The innate contributions to atherosclerosis, especially in the transition of monocyte to lipid-laden macrophage, are well established, but the mechanisms underpinning the infiltration, persistence, and effector dynamics of CD8 T cells in particular are not well understood. Adaptive immunity is centered on a classical cascade of antigen recognition and activation, costimulation, and effector cytokine secretion upon recall of antigen. However, chronic inflammation can generate alternative cues that supplant this behavior pattern and promote the retention and activation of peripherally activated T cells. Furthermore, the atherogenic foci that activated immune cell infiltrate are unique lipid-laden environments that offer a diverse array of stimuli, including those of survival, antigen hyporesponsiveness, and inflammatory cytokine expression. This review will focus on how known cardiovascular comorbidities may be influencing CD8 T-cell activation and how, once infiltrated within atherogenic foci, these T cells face a multitude of cues that skew the classical cascade of T-cell behavior, highlighting alternative modes of activation that may help contextualize associations of autoimmunity, viral infection, and immunotherapy with cardiovascular morbidity.

T cell-directed IL-17 production by lung granular γδ T cells is coordinated by a novel IL-2 and IL-1ß circuit.

Immune-mediated lung is considered the result of an exacerbated innate injury immune response, although a role for adaptive lymphocytes is emerging. αß T cells specific for S. aureus enterotoxin A orchestrate a Tγδ17 response during lung injury. However, the mechanism driving IL-17 production is unclear. Here, we show a role for IL-2 triggering IL-17 production by lung granular γδ T cells as IL-17 synthesis and neutrophil recruitment was reduced by IL-2 blocking mAbs in vitro and in vivo. Mass cytometry analysis revealed that lung γδ T cells responded directly to IL-2 as evident from STAT5 phosphorylation and RoRγt expression. IL-2 receptor blocking mAbs and JAK inhibition impaired STAT5 phosphorylation and IL-17 release. Moreover, inhalation of S. aureus enterotoxin A induced IL-2 secretion and caspase-1-dependent IL-1ß activation to drive IL-17 production. This T-cell-mediated inflammasome-dependent IL-17 response is maximum when lung Tγδ17 cells were sequentially stimulated first with IL-2 then IL-1ß. Interestingly, when IL-2 is given therapeutically to cancer patients it carries a known risk of lung injury that is largely indistinguishable from that seen in sepsis. Hence, this novel mechanism reveals therapeutic targets treating both acute lung injury and high-dose IL-2 toxicity in cancer.

Macrophage polarization and meta-inflammation.

Chronic overnutrition and obesity induces low-grade inflammation throughout the body. Termed "meta-inflammation," this chronic state of inflammation is mediated by macrophages located within the colon, liver, muscle, and adipose tissue. A sentinel orchestrator of immune activity and homeostasis, macrophages adopt variable states of activation as a function of time and environmental cues. Meta-inflammation phenotypically skews these polarization states and has been linked to numerous metabolic disorders. The past decade has revealed several key regulators of macrophage polarization, including the signal transducer and activator of transcription family, the peroxisome proliferator-activated receptor gamma, the CCAAT-enhancer-binding proteins (C/EBP) family, and the interferon regulatory factors. Recent studies have also suggested that microRNAs and long noncoding RNA influence macrophage polarization. The pathogenic alteration of macrophage polarization in meta-inflammation is regulated by both extracellular and intracellular cues, resulting in distinct secretome profiles. Meta-inflammation-altered macrophage polarization has been linked to insulin insensitivity, atherosclerosis, inflammatory bowel disease, cancer, and autoimmunity. Thus, further mechanistic exploration into the skewing of macrophage polarization promises to have profound impacts on improving global health.

Matrix elasticity of void-forming hydrogels controls transplanted-stem-cell-mediated bone formation.

The effectiveness of stem cell therapies has been hampered by cell death and limited control over fate. These problems can be partially circumvented by using macroporous biomaterials that improve the survival of transplanted stem cells and provide molecular cues to direct cell phenotype. Stem cell behaviour can also be controlled in vitro by manipulating the elasticity of both porous and non-porous materials, yet translation to therapeutic processes in vivo remains elusive. Here, by developing injectable, void-forming hydrogels that decouple pore formation from elasticity, we show that mesenchymal stem cell (MSC) osteogenesis in vitro, and cell deployment in vitro and in vivo, can be controlled by modifying, respectively, the hydrogel's elastic modulus or its chemistry. When the hydrogels were used to transplant MSCs, the hydrogel's elasticity regulated bone regeneration, with optimal bone formation at 60 kPa. Our findings show that biophysical cues can be harnessed to direct therapeutic stem cell behaviours in situ.

Patient preferences for testing for pulmonary embolism in the ED using a shared decision-making model.

INTRODUCTION: Shared decision making (SDM) is a process whereby patients and clinicians work together to make informed medical decisions that incorporate patient values. Recent data suggest that, for patients with low pretest probability of pulmonary embolism (PE), doubling the standard d-dimer cutoff may reduce the need for imaging with minimal increase in missed PE diagnoses. We used an SDM approach to determine patient preferences regarding this diagnostic approach. METHODS: We prospectively enrolled a consecutive sample of emergency department (ED) patients presenting with chest pain or dyspnea. We provided patients with a standardized description of the diagnostic workup for PE. We also provided image arrays describing the risks of computed tomography in low pretest probability patients and the risks of deferring imaging assuming a d-dimer was less than twice the value normally considered positive. We surveyed patients for their preference to undergo or defer imaging in this scenario. RESULTS: We enrolled 203 ED patients. Mean age was 55 ± 17 years, and 61% were male. Seventy-four patients (37%) elected to defer computed tomography of the pulmonary arteries testing. Patients with a previous PE diagnosis were less likely to defer computed tomography of the pulmonary arteries testing (P = .007). There was no association between the decision to defer testing and age, sex, family history of PE, or self-assessed risk-taking tendency. CONCLUSIONS: When presented with a hypothetical scenario, more than one-third of patients deferred imaging for PE based on low clinical probability and a d-dimer less than twice the normal threshold. An SDM approach is acceptable to patients and may decrease imaging for PE.