Citations

AbstractIntroductionWe sought to assess longitudinal electronic health records (EHRs) using machine learning (ML) methods to computationally derive probable Alzheimer's Disease (AD) and related dementia subphenotypes.MethodsA retrospective analysis of EHR data from a cohort of 7587 patients seen at a large, multi‐specialty urban academic medical center in New York was conducted. Subphenotypes were derived using hierarchical clustering from 792 probable AD patients (cases) who had received at least one diagnosis of AD using their clinical data. The other 6795 patients, labeled as controls, were matched on age and gender with the cases and randomly selected in the ratio of 9:1. Prediction models with multiple ML algorithms were trained on this cohort using 5‐fold cross‐validation. XGBoost was used to rank the variable importance.ResultsFour subphenotypes were computationally derived. Subphenotype A (n = 273; 28.2%) had more patients with cardiovascular diseases; subphenotype B (n = 221; 27.9%) had more patients with mental health illnesses, such as depression and anxiety; patients in subphenotype C (n = 183; 23.1%) were overall older (mean (SD) age, 79.5 (5.4) years) and had the most comorbidities including diabetes, cardiovascular diseases, and mental health disorders; and subphenotype D (n = 115; 14.5%) included patients who took anti‐dementia drugs and had sensory problems, such as deafness and hearing impairment.The 0‐year prediction model for AD risk achieved an area under the receiver operating curve (AUC) of 0.764 (SD: 0.02); the 6‐month model, 0.751 (SD: 0.02); the 1‐year model, 0.752 (SD: 0.02); the 2‐year model, 0.749 (SD: 0.03); and the 3‐year model, 0.735 (SD: 0.03), respectively. Based on variable importance, the top‐ranked comorbidities included depression, stroke/transient ischemic attack, hypertension, anxiety, mobility impairments, and atrial fibrillation. The top‐ranked medications included anti‐dementia drugs, antipsychotics, antiepileptics, and antidepressants.ConclusionsFour subphenotypes were computationally derived that correlated with cardiovascular diseases and mental health illnesses. ML algorithms based on patient demographics, diagnosis, and treatment demonstrated promising results in predicting the risk of developing AD at different time points across an individual's lifespan.

Abstract Background: Upamostat is an orally available small molecule serine protease inhibitor that is a highly potent inhibitor of trypsin 1, trypsin 2 and trypsin 3 (PRSS1/2/3) as well as urokinase-type plasminogen activator (uPA) which are expressed in many cancers and mediate cell migration, invasion and tissue remodeling. Opaganib (ABC294640), a novel, orally available small molecule is a specific inhibitor of sphingosine kinase 2 (SPHK2), which phosphorylates sphingosine to sphingosine-1-phosphate (S-1-P). While proliferation induced by S-1-P is regulated by both sphingosine kinase 1 (SPHK1) and SPHK2, SPHK2 appears to be more involved in cancer. We aimed to investigate the potential antitumor effect of upamostat and opaganib, individually and in combination, on cholangiocarcinoma (CCA) patient derived xenografts (PDX) in nude mice. Methods: PAX165, a PDX from a surgically resected CCA, expresses substantial levels of SPHK2, PRSS1, PRSS2 and PRSS3. 4 groups of 18 mice were treated with either drug or both. Mouse weights and tumor volumes were measured. In addition, experiments were conducted using the chorioallantoic membrane (CAM) of chicken embryos. Results: Table 1 shows the average tumor size for the control, upamostat, opaganib, and upamostat+opaganib groups at the study end point (Day 42). Tumor volumes in the upamostat, opaganib, and upamostat+opagnib groups were significantly decreased compared to the control group. The CAM experiments are ongoing and will be presented at the AACR Annual Meeting. Change in tumor volumes (mean) of CCA PDX after opaganib, upamostat or combination treatmentControlOpaganibUpamostatOpaganib+UpamostatPre-treatment129.9128.7118.8126.8Day 42198.6102.093.3186.09Percent change Day 0-42+53%-21%-21%-32%P value vs. control0.00020.00100.0008 Conclusion: This preclinical study demonstrated that upamostat and opaganib resulted in tumor regression in mice. Body weights of the mice showed no significant inter- or intra- group differences. The combination of upamostat and opaganib treatment showed greater regression compared to either upamostat or opaganib alone. Studies are underway to identify the molecular mechanisms of their interaction. Citation Format: Faizal Z. Asumda, Mohamed A. Hassan, Yo Han Kim, Nellie A. Campbell, Xin Luo, Daniel R. O'Brien, Sarah A. Buhrow, Joel M. Reid, Michael J. Moore, Vered Katz Ben-Yair, Reza Fathi, Mark L. Levitt, Fabrice Lucien-Matteoni, Jennifer L. Leiting, Mark J. Truty, Lewis R. Roberts. Effects of upamostat and opaganib on cholangiocarcinoma patient derived xenografts [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3078.

AbstractObjectiveTo identify depression subphenotypes from Electronic Health Records (EHRs) using machine learning methods, and analyze their characteristics with respect to patient demographics, comorbidities, and medications.Materials and MethodsUsing EHRs from the INSIGHT Clinical Research Network (CRN) database, multiple machine learning (ML) algorithms were applied to analyze 11 275 patients with depression to discern depression subphenotypes with distinct characteristics.ResultsUsing the computational approaches, we derived three depression subphenotypes: Phenotype_A (n = 2791; 31.35%) included patients who were the oldest (mean (SD) age, 72.55 (14.93) years), had the most comorbidities, and took the most medications. The most common comorbidities in this cluster of patients were hyperlipidemia, hypertension, and diabetes. Phenotype_B (mean (SD) age, 68.44 (19.09) years) was the largest cluster (n = 4687; 52.65%), and included patients suffering from moderate loss of body function. Asthma, fibromyalgia, and Chronic Pain and Fatigue (CPF) were common comorbidities in this subphenotype. Phenotype_C (n = 1452; 16.31%) included patients who were younger (mean (SD) age, 63.47 (18.81) years), had the fewest comorbidities, and took fewer medications. Anxiety and tobacco use were common comorbidities in this subphenotype.ConclusionComputationally deriving depression subtypes can provide meaningful insights and improve understanding of depression as a heterogeneous disorder. Further investigation is needed to assess the utility of these derived phenotypes to inform clinical trial design and interpretation in routine patient care.

Abstract Background: Deregulation of ceramide and sphingomyelin levels have been suggested to contribute to the pathogenesis of Alzheimer’s disease (AD). Ceramide transfer proteins (CERTs) are ceramide carriers, crucial for ceramide and sphingomyelin balance in cells. Extracellular forms of CERTs co-localize with amyloid-β (Aβ) plaques in AD brains. To date, the significance of these observations for the pathophysiology of AD remains uncertain.Methods: The plasmid expressing CERTL, the long isoform of CERTs, was used to study the interaction of CERTL with amyloid precursor protein (APP) by co-immunoprecipitation and immunofluorescence in HEK cells. The recombinant CERTL protein was employed to study interaction of CERTL with amyloid-β (Aβ), Aβ aggregation process in presence of CERTL, and the resulting changes in Aβ toxicity in neuroblastoma cells. CERTL was overexpressed in neurons by adeno associated virus (AAV) in a familial mouse model of familial AD (5xFAD). Ten weeks after transduction animal were challenged with behavior tests for memory, anxiety and locomotion. At week twelve brains were investigated for sphingolipid levels by mass spectrometry, plaques and neuroinflammation by immunohistochemistry, gene expression and/or immunoassay.Results: Here, we report that CERTL, binds to APP, modifies Aβ aggregation and reduces Aβ neurotoxicity in vitro. Furthermore, we show that intracortical injection of AAV, mediating the expression of CERTL, decreases levels of ceramide d18:1/16:0 and increases sphingomyelin levels in the brain of male transgenic mice, modelling familial AD (5xFAD). CERTL in vivo over-expression has a mild effect on animal locomotion and decreases Aβ formation and modulates microglia by decreasing their pro-inflammatory phenotype.Conclusion: Our results demonstrate a crucial role of CERTL in regulating ceramide levels in the brain, in amyloid plaque formation and neuroinflammation, thereby opening research avenues for therapeutic targets of AD and other neurodegenerative diseases.

AbstractIntroductionElectronic health record (EHR)‐driven phenotyping is a critical first step in generating biomedical knowledge from EHR data. Despite recent progress, current phenotyping approaches are manual, time‐consuming, error‐prone, and platform‐specific. This results in duplication of effort and highly variable results across systems and institutions, and is not scalable or portable. In this work, we investigate how the nascent Clinical Quality Language (CQL) can address these issues and enable high‐throughput, cross‐platform phenotyping.MethodsWe selected a clinically validated heart failure (HF) phenotype definition and translated it into CQL, then developed a CQL execution engine to integrate with the Observational Health Data Sciences and Informatics (OHDSI) platform. We executed the phenotype definition at two large academic medical centers, Northwestern Medicine and Weill Cornell Medicine, and conducted results verification (n = 100) to determine precision and recall. We additionally executed the same phenotype definition against two different data platforms, OHDSI and Fast Healthcare Interoperability Resources (FHIR), using the same underlying dataset and compared the results.ResultsCQL is expressive enough to represent the HF phenotype definition, including Boolean and aggregate operators, and temporal relationships between data elements. The language design also enabled the implementation of a custom execution engine with relative ease, and results verification at both sites revealed that precision and recall were both 100%. Cross‐platform execution resulted in identical patient cohorts generated by both data platforms.ConclusionsCQL supports the representation of arbitrarily complex phenotype definitions, and our execution engine implementation demonstrated cross‐platform execution against two widely used clinical data platforms. The language thus has the potential to help address current limitations with portability in EHR‐driven phenotyping and scale in learning health systems.

Abstract Background: Deregulation of ceramide and sphingomyelinlevels have been suggested tocontribute tothe pathogenesis of Alzheimer’s disease (AD).Ceramide transfer proteins (CERTs) are ceramide carriers, crucial for ceramide and sphingomyelin balance in cells.Extracellular forms of CERTs co-localize with amyloid-β (Aβ) plaques in AD brains. To date, the significance of these observations for the pathophysiology of AD remains uncertain.Methods: The plasmid expressing CERTL, the long isoform of CERTs, was used to study the interaction of CERTLwith amyloid precursor protein (APP) by co-immunoprecipitation and immunofluorescencein HEK cells.The recombinant CERTL protein wasemployed to study interaction of CERTLwith amyloid-β (Aβ), Aβ aggregation process in presence of CERTL, and the resulting changes inAβ toxicity in neuroblastoma cells. CERTLwas overexpressed in neurons by adeno associatedvirus (AAV) in a familial mouse model of familial AD (5xFAD). Ten weeks after transduction animal were challenged with behavior tests for memory, anxiety and locomotion. At week twelve brains were investigated for sphingolipid levels by mass spectrometry, plaques and neuroinflammation by immunohistochemistry, gene expression and/or immunoassay.Results:Here, we report that CERTL, binds to APP, modifies Aβ aggregation and reduces Aβ neurotoxicity in vitro. Furthermore, we show that intracortical injection of AAV, mediating the expression of CERTL, decreases levels of ceramide d18:1/16:0 and increases sphingomyelin levels in the brain of male transgenic mice, modelling familial AD (5xFAD). CERTLin vivo over-expression hasa mild effect on animal locomotion and decreases Aβ formation and modulates microglia by decreasing their pro-inflammatory phenotype.Conclusion: Our results demonstratea crucial role of CERTL in regulatingceramidelevels in the brain, in amyloid plaque formation and neuroinflammation,thereby opening research avenuesfor therapeutic targets of AD and other neurodegenerative diseases.

e21631 Background: The efficacy of osimertinib in previously EGFR-TKI treated NSCLC patients without T790M mutation remains unclear in real-world practice. We investigated whether osimertinib can provide survival benefit in EGFR-mutant patients without T790M mutation after 1st/2nd generation TKI treatment. Methods: Between January 1, 2009, and March 31, 2019, 417 patients had stage III-IV NSCLC harboring EGFR mutation and 154 out of 417 patients receiving osimertinib as≥2nd-line EGFR-TKI treatment were identified at Mayo Clinic. The time to treatment failure of osimertinib was analyzed by the Kaplan-Meier (KM) estimates. The risk of death post diagnosis was analyzed by Cox proportional hazard models. Results: Among 417 EGFR-mutant patients, higher risk of death was found in patients with age above 65 years, non-adenocarcinoma, no surgery treatment, no radiation treatment, non-exon 19 deletion/exon 21 L858R mutation, higher ECOG PS (2-4), PD-L1 expression of 50% or more, bone metastasis, live metastasis, and adrenal metastasis (all p < 0.05). Moreover, osimertinib as ≥2nd-line TKI treatment in patients with or without T790M revealed lower risk of death compared to 1st/2nd generation TKI treatment without subsequent osimertinib (HR = 0.33; 0.46, and p = 0.0002; 0.0232, respectively). However, among patients receiving osimertinib as ≥2nd-line TKI treatment, patients with T790M did not have superior survival than those without (p = 0.2803). Among 154 patients receiving osimertinib, a higher risk of treatment failure for osimertinib was found in male (HR = 1.72; p = 0.0327), patients with 1st-line TKI duration ≤12 months (HR = 2.16; p = 0.0019), BMI drop > 10% (HR = 1.85; p = 0.0207), PD-L1 levels of 50% or more (HR = 4.28; p = 0.0008), and 1st-line TKI with afatinib (HR = 2.19; p = 0.0136). Nonetheless, osimertinib as ≥2nd-line TKI in patients without 790M mutation did not have higher risk of treatment failure than those with T790M (p = 0.1236). Conclusions: This is the first study to demonstrate that osimertinib can provide similar survival benefit in previously EGFR-TKI treated NSCLC patients without T790M mutation as those with T790M in real-world practice. Additionally, EGFR-mutant patients with PD-L1 expression ≥50% had a higher risk of treatment failure for osimertinib and worse overall survival than those with PD-L1 expression < 50% and may potentially gain benefit from optimizing treatment strategies including immunotherapy.

3552 Background: Delta-like 3 (DLL3) is highly and specifically expressed in solid tumors, such as neuroendocrine carcinomas (NECs), malignant melanoma (MM), and medullary thyroid carcinoma (MTC). Rovalpituzumab tesirine (Rova-T) is a DLL3-targeting antibody-drug conjugate. Methods: This Phase 1/2 study (NCT02709889) enrolled patients with relapsed/refractory DLL3+ (>1% by IHC) advanced solid tumors and ECOG performance status of 0-1. Rova-T was given IV at 0.2, 0.3, or 0.4 mg/kg on d 1 of each 6-wk cycle (q6wk) for dose escalation (3+3 design) in disease-specific cohorts in Phase I. The recommended Phase 2 dose (RP2D) was tested in Phase II. Safety and dose-limiting toxicities (DLTs) were primary endpoints; efficacy outcomes were secondary endpoints. Results: The study enrolled 200 patients; 101 had NECs (large cell NEC [n=13], neuroendocrine prostate cancer [n=21], high-grade gastroenteropancreatic NEC [n=36], other [n=31]) and 99 had other solid tumors (MM [n=20], MTC [n=13], glioblastoma [GBM; n=23], other [n=43]). The median age was 61 y (range, 28-84); 63% were male. The RP2D was 0.3 mg/kg q6wk for 2 cycles in all cohorts. There were 7 DLTs in 5 patients: 2 with 0.2 mg/kg (Grade [Gr] 3 photosensitivity reaction, Gr 3 dyspnea), 2 with 0.3 mg/kg (1 with Gr 2 effusion, Gr 3 tumor lysis syndrome, and Gr 3 rhabdomyolysis; 1 with Gr 4 kidney injury), and 1 with 0.4 mg/kg (Gr 4 thrombocytopenia). Despite only 1 DLT identified with 0.4 mg/kg, the totality of the safety data suggested that this dose is not well tolerated. Common adverse events (AEs) in patients given 0.3 mg/kg (n=145) are shown (Table). Serious AEs occurred in 77/145 patients (53%), most commonly (≥3%) malignant neoplasm progression (n=18; 12%), pleural effusion (n=7; 5%), pericardial effusion (n=6; 4%), and dyspnea (n=5; 3%). The objective response rate (ORR) was 11% (21/200): 14 had NEC, 2 had MM, 2 had MTC, 2 had small cell carcinoma (SCC) not of lung origin (all partial responses), and 1 had GBM (complete response). In patients with NECs given 0.3 mg/kg, ORR, clinical benefit rate, and progression-free survival trended in favor of those with high DLL3-expressing tumors (≥50% by IHC) which represented 51% of NECs. Conclusions: Rova-T was tolerable in patients with advanced solid tumors at 0.3 mg/kg q6wk for 2 cycles. Antitumor activity was observed in patients with NEC, MM, MTC, SCC, and GBM. Clinical trial information: NCT02709889 . [Table: see text]

Abstract Objective The n2c2/UMass Lowell spin-off shared task focused on medical concept normalization (MCN) in clinical records. This task aimed to assess state-of-the-art methods for matching salient medical concepts from clinical records to a controlled vocabulary. We describe the task and the dataset used, compare the participating systems, and identify the strengths and limitations of the current approaches and directions for future research. Materials and Methods Participating teams were asked to link preselected text spans in discharge summaries (henceforth referred to as concept mentions) to the corresponding concepts in the SNOMED CT (Systematized Nomenclature of Medicine Clinical Terms) and RxNorm vocabularies from the Unified Medical Language System. The shared task used the MCN corpus created by the organizers, which maps all mentions of problems, treatments, and tests in the 2010 i2b2/VA challenge data to the Unified Medical Language System concepts. Submitted systems represented 4 broad categories of approaches: cascading dictionary matching, cosine distance, deep learning, and retrieve-and-rank systems. Disambiguation modules were common across all approaches. Results A total of 33 teams participated in the shared task. The best-performing team achieved an accuracy of 0.8526. The median and mean performances among all teams were 0.7733 and 0.7426, respectively. Conclusions Overall performance among the top 10 teams was high. However, particularly challenging for all teams were mentions requiring disambiguation of misspelled words, acronyms, abbreviations, and mentions with more than 1 possible semantic type. Complex mentions of long, multiword terms were also challenging and, in the future, will require better methods for learning contextualized representations of concept mentions and better use of domain knowledge.