Stratified Cohorts Research Articles

Amyloid and SCD jointly predict cognitive decline across Chinese and German cohorts.

Subjective cognitive decline (SCD) in amyloid-positive (Aβ+) individuals was proposed as a clinical indicator of Stage 2 in the Alzheimer's disease (AD) continuum, but this requires further validation across cultures, measures, and recruitment strategies. Eight hundred twenty-one participants from SILCODE and DELCODE cohorts, including normal controls (NC) and individuals with SCD recruited from the community or from memory clinics, underwent neuropsychological assessments over up to 6 years. Amyloid positivity was derived from positron emission tomography or plasma biomarkers. Global cognitive change was analyzed using linear mixed-effects models. In the combined and stratified cohorts, Aβ+ participants with SCD showed steeper cognitive decline or diminished practice effects compared with NC or Aβ- participants with SCD. These findings were confirmed using different operationalizations of SCD and amyloid positivity, and across different SCD recruitment settings. Aβ+ individuals with SCD in German and Chinese populations showed greater global cognitive decline and could be targeted for interventional trials. SCD in amyloid-positive (Aβ+) participants predicts a steeper cognitive decline. This finding does not rely on specific SCD or amyloid operationalization. This finding is not specific to SCD patients recruited from memory clinics. This finding is valid in both German and Chinese populations. Aβ+ older adults with SCD could be a target population for interventional trials.

The Five-Year Incidence of Progression to Osteoarthritis and Total Joint Arthroplasty in Patients Prescribed Glucagon-Like Peptide 1 Receptor Agonists

BackgroundResearch has suggested that glucagon-like peptide-1 receptor agonists (GLP-1-RAs) may have therapeutic effects on osteoarthritis of the hip and knee, in addition to managing diabetes and obesity. However, there is a lack of understanding regarding the association between GLP-1-RA use and the diagnosis of osteoarthritis (OA) of the hip and knee. MethodsA collaborative network analytics platform was queried for obese diabetic (n = 1,094,198), obese nondiabetic (n = 916,235), and nonobese diabetic (n = 157,305) patients who had an index visit between 2015 and 2017. Patients who had pre-existing hip and/or knee OA were excluded. A 1:1 propensity score matching was used to balance GLP-1-RA use in stratified cohorts for age, sex, race, body mass index, and hemoglobin A1c. The primary outcomes were rates of progression to hip OA, knee OA, major joint injections, total hip arthroplasty, and total knee arthroplasty. Cox proportional hazards models determined hazard ratios (HRs) between cohorts prescribed and not prescribed GLP-1-RAs. ResultsAll patients had a five-year follow-up. Rates of progression to hip and knee OA were higher among the GLP-1-RA users in both obese diabetic (hip HR: 1.63, 95% confidence interval [CI]: 1.46 to 1.82; knee HR: 1.52, CI: 1.41 to 1.64) and nonobese diabetic (hip HR: 1.78, CI: 1.50 to 2.10; knee HR: 1.58, CI: 1.39 to 1.80) cohorts. These diabetic cohorts received higher rates of major joint injections, though there was no difference in rates of total hip arthroplasty or total knee arthroplasty. No differences in five-year outcomes were seen when comparing obese, nondiabetic patients who were prescribed GLP-1-RAs with obese, nondiabetic patients not exposed to GLP-1-RAs. ConclusionsThis five-year analysis found a greater risk of progression to hip and knee OA among obese and non-obese diabetic GLP-1-RA users. Further studies should explore GLP-1-RA effects upon glucose management, weight loss, and lower extremity arthritis development. Level of EvidenceIII, retrospective cohort study.

Comparison of tumor immune microenvironments (TIMEs) between primary and metastatic sites (Mets) in triple-negative breast cancer (TNBC).

1097 Background: The TIME is critical in determining response to immune checkpoint inhibitors (ICIs), but TIME differences across primary and mets in TNBC are not well understood. Since ICIs are standard of care for patients with TNBC, we studied the TIMEs of primary TNBC and mets to investigate how immune composition may affect ICI efficacy. Secondary analysis stratified cohorts by race and disease sites to compare TIMEs in Black or African American (B/AA) with White patients, since TNBC has a high prevalence of B/AA women who are underrepresented in studies and have a worse prognosis. Methods: We retrospectively analyzed de-identified next-generation sequencing data from TNBC patients (n=1,044) in the Tempus Database. Tumors from primary breast (PB, n=553), liver (n=153), lymph node (LN, n=174), lung (n=111), and bone (n=53) sites were sequenced with the Tempus xT DNA (648-gene panel) and xR RNA assays. Histologies included invasive ductal carcinoma (82%), lobular carcinoma (2.2%), mixed ductal and lobular carcinoma (1.7%), and other (14%). Demographics, TMB, MSI, PD-L1, and proportions of B, T (CD4+, CD8+), NK cells, and macrophages were compared across sites. LN were used as a positive control. Chi-squared/Fisher’s exact or Kruskal-Wallis tests were used to assess statistical significance (two-sided, evaluated at the 0.05 alpha level). Results: The cohort (median age=57 years, IQR 46-66) comprised a diverse population (65% White, 23% B/AA, 2.8% Asian, and 9.3% other). Liver exhibited a lower percentage of B cells and higher percentage of macrophages compared to PB (p<0.0001 for both), lung (p<0.0001 for both), and bone (B cells p<0.0001, macrophages p<0.05). Liver also exhibited lower percentages of CD8+ T cells compared to PB (p<0.05) and CD4+ and NK cells compared to lung (CD4+ cells p<0.01, NK cells p<0.0001). Bone had lower percentages of CD8+ cells compared to PB (p<0.0001), lung (p<0.001), and liver (p<0.01). Compared to lung, bone had a lower percentage of CD4+ cells (p<0.0001) and higher percentage of macrophages (p<0.01). PD-L1 positivity, TMB, and MSI exhibited no differences across sites. Secondary analysis compared the TIME between B/AA and White cohorts in PB (B/AA n=85 vs White n=204: %B cells, 15 vs 11, p=0.017; %CD8 cells, 4.9 vs 7.3, p=0.025), liver (B/AA n=15 vs White n=48: %B cells, 8 vs 4, p=0.2; %CD4, 23 vs 14, p=0.2; %CD8, 7.2 vs 4.6, p=0.07), and lung (B/AA n=9 vs White n=33: %B cells, 22 vs 11, p=0.021; %CD8, 7.9 vs 5.2, p=0.5). Conclusions: Compared to PB and lung, liver and bone had immune cell infiltrates indicating a less immunogenic TIME in the overall TNBC population. Although limited by sample size, this is one of the first studies to assess the TIME across race and sites of disease. These findings are hypothesis generating and provide further rationale to better understand how the TIME across disease sites and race may alter the efficacy of ICIs in TNBC.

Sex‐specific effects of genetically predicted BMI and education on risk of AD

AbstractBackgroundExamining sex differences on the impact of modifiable risk factors in Alzheimer’s disease (AD) risk can help us better understand the mechanisms underlying sex differences in the prevalence and incidence of AD. Here, we used polygenic risk scores (PRS) and Mendelian randomization (MR) to investigate sex‐specific effects of sleep duration, insomnia, blood pressure, diabetes, alcohol intake, smoking, body mass index, high cholesterol, and education on AD risk.MethodWe obtained combined and sex‐stratified genome‐wide association study (GWAS) summary statistics for each risk factor from the UK Biobank and used them as the base dataset for constructing PRS and exposure datasets in the MR analysis. Linkage disequilibrium clumping was performed to identify independent genome‐wide significant single nucleotide polymorphisms (SNPs) across the sex‐combined, male‐specific, and female‐specific GWAS. The combined lead SNPs were then weighted by their strata‐specific effect sizes in the stratified PRS and MR analyses. PRS were constructed for each risk factor in participants from the Alzheimer’s Disease Genetics Consortium. Linear regression was used to investigate the association of each PRS with AD risk, adjusting for age, principal components, and cohort. MR was used to estimate sex‐stratified causal effects of each risk factor on AD. Sex differences in PRS associations and MR causal estimates were determined using Fisher’s Z score method.ResultsAssociation testing of the PRS with AD risk in sex stratified cohorts revealed that the BMI PRS was associated with differential effects in men and women (OR [95%CI]: males: 1.05 [1.00, 1.10] vs females: 0.96 [0.93, 0.99], p = 0.003). Furthermore, the university completion PRS was non‐significant in men but was associated with reduced risk in women (OR [95%CI]: males: 0.96 [0.92, 1.01] vs females: 0.92 [0.89, 0.96], p = 0.12). In the follow‐up MR analysis, university completion was also causally associated with reduced risk in women only (OR [95%CI]: males: 1.09 [0.77, 1.53] vs females: 0.55 [0.42, 0.72], p = 0.002) (Figure 1). No other risk factors showed evidence of sex‐differences.ConclusionOur study found sex‐specific effects of genetically predicted BMI and educational attainment on AD risk. These findings suggest the need for sex‐specific approaches to AD prevention and management.

Implementing an end-to-end pathway for detection, diagnosis, and management of atrial fibrillation in the risk-stratified patients: results from the atrial fibrillation stroke prevention hub program

Abstract Background Atrial fibrillation (AF) is a prevalent cardiac arrhythmia linked with a five-fold increased risk of ischemic stroke. Despite the need for pulse rhythm checks of reasonable duration in selectively identified high-risk patients, the NHS is under pressure to perform these tasks during daily practice in primary care to prevent stroke. Purpose This program aimed to establish an end-to-end pathway to identify, detect, diagnose, and manage high-risk patients with no prior AF diagnosis. Methods The AF Stroke Prevention Hub program was aimed at patients aged 65 and above with a history of heart failure or stroke/transient ischemic attack. Data from electronic patient records identified these patients, while exclusion criteria consisted of known AF, implanted cardiac devices, end-stage renal disease and end-of-life care. The program used a medically certified smartphone application to monitor heart rate and rhythm and track symptoms using photoplethysmography (PPG). Patients were facilitated to perform a minimum of two measurements per day, for seven days. Those who were digitally excluded were offered an assessment in a face-to-face clinic appointment. Based on the PPG recordings, patients with a positive finding received a confirmatory ECG examination and anticoagulation therapy, once the diagnosis was established, within 48 hours. Results Between February 2022 and February 2023, after applying inclusion and exclusion criteria, six hundred and sixty-nine patients were found to be eligible from 4 primary care practices. Two hundred and sixty-seven patients were issued PPG applications after obtaining consent. In total, two hundred and ten patients completed the PPG-based, 7-day monitoring period. The technology adoption rate was 78.65% in this group of patients. Among the high-risk group of cardiac failure, four patients were detected with possible AF based on the PPG recordings. All four were confirmed via a 12-lead ECG or a Holter monitor, with an AF detection rate of 9.09%. Combining all stratified risk patient cohorts, ten (4.76%) were detected with possible AF based on the PPG recordings and six (2.8%) were verified based on a confirmatory 12-lead ECG or a 7-day Holter. All patients with newly diagnosed AF received anticoagulation therapy and were managed accordingly, while the remaining patients received advice regarding self-management, lifestyle, and yearly health checks. Conclusion Compared to the current NHS opportunistic pulse check, where the detection rate is <1%, the AF Stroke Prevention Hub program successfully identified patients with a significantly higher detection rate. The hub delivered an end-to-end pathway allowing real-time reporting and triaging of patients, early detection, appropriate confirmation, and rapid treatment with favorable real-life technology adoption. Expanding the data-driven program to a wider difficult-to-reach population could reduce the burden on NHS and improve patient outcomes.

Looking Back, Looking Forward: A Study Protocol for a Mixed-Methods Multiple-Case Study to Examine Improvement Sustainability of Large-Scale Initiatives in Tertiary Hospitals.

Background Hospitals invest extensive resources in large-scale initiatives to improve patient safety and quality at an organizational level. However, initial success, if any, does not guarantee longer-term improvement. Empirical and theoretical knowledge that informs hospitals on how to attain sustained improvement from large-scale change is lacking. Aim The proposed study aims to examine improvement sustainability of two large-scale initiatives in an Australian tertiary hospital and translate the lessons into strategies for achieving sustained improvement from large-scale change in hospital settings. Design and Methods The study employs a single-site, multiple-case study design to evaluate the initiatives separately and comparatively using mixed methods. Semi-structured staff interviews will be conducted in stratified cohorts across the organizational hierarchy to capture different perspectives from various staff roles involved in the initiatives. The output and impact of the initiatives will be examined through organizational documents and relevant routinely collected organizational indicators. The obtained data will be analyzed thematically and statistically before being integrated for a synergic interpretation. Implications Capturing a comprehensive organizational view of large-scale change, the findings will have the potential to guide the practice and contribute to the theoretical understandings for achieving meaningful and longer-term organizational improvement in patient safety and quality.

Reduction of Sarcopenia Using HIFEM and RF in Two Age Stratified Cohorts: A Prospective Clinical Trial

Reduction of Sarcopenia Using HIFEM and RF in Two Age Stratified Cohorts: A Prospective Clinical Trial

Predictive Models for Length of Stay and Discharge Disposition in Elective Spine Surgery: Development, Validation, and Comparison to the ACS NSQIP Risk Calculator.

A retrospective study at a single academic institution. The purpose of this study is to utilize machine learning to predict hospital length of stay (LOS) and discharge disposition following adult elective spine surgery, and to compare performance metrics of machine learning models to the American College of Surgeon's National Surgical Quality Improvement Program's (ACS NSQIP) prediction calculator. A total of 3678 adult patients undergoing elective spine surgery between 2014 and 2019, acquired from the electronic health record. Patients were divided into three stratified cohorts: cervical degenerative, lumbar degenerative, and adult spinal deformity groups. Predictive variables included demographics, body mass index, surgical region, surgical invasiveness, surgical approach, and comorbidities. Regression, classification trees, and least absolute shrinkage and selection operator (LASSO) were used to build predictive models. Validation of the models was conducted on 16% of patients (N=587), using area under the receiver operator curve (AUROC), sensitivity, specificity, and correlation. Patient data were manually entered into the ACS NSQIP online risk calculator to compare performance. Outcome variables were discharge disposition (home vs. rehabilitation) and LOS (days). Of 3678 patients analyzed, 51.4% were male (n=1890) and 48.6% were female (n=1788). The average LOS was 3.66 days. In all, 78% were discharged home and 22% discharged to rehabilitation. Compared with NSQIP (Pearson R2 =0.16), the predictions of poisson regression ( R2 =0.29) and LASSO ( R2 =0.29) models were significantly more correlated with observed LOS ( P =0.025 and 0.004, respectively). Of the models generated to predict discharge location, logistic regression yielded an AUROC of 0.79, which was statistically equivalent to the AUROC of 0.75 for NSQIP ( P =0.135). The predictive models developed in this study can enable accurate preoperative estimation of LOS and risk of rehabilitation discharge for adult patients undergoing elective spine surgery. The demonstrated models exhibited better performance than NSQIP for prediction of LOS and equivalent performance to NSQIP for prediction of discharge location.

Who to Enroll in Parkinson Disease Prevention Trials? The Case for Genetically At-Risk Cohorts.

Therapies that prevent the occurrence of Parkinson disease (PD) (primary prevention) or mitigate the progression of symptoms in those with early disease (secondary prevention) are a critical unmet need in disease management. Despite great promise, PD prevention trials have not yet demonstrated success. Initiation of treatment too late in the disease course and the heterogeneity of disease are obstacles that may have contributed to the failure. Genetically stratified groups offer many advantages to primary and secondary prevention trials. In addition to their ease of identification, they decrease disease heterogeneity on several levels. Particularly, they comprise a phenotypically and pathologically enriched group with defined clinical features, pathogenic mechanisms and associated proteins that may serve as specific trial endpoints, therapeutic targets and biomarkers for disease state, and pharmacodynamic and pharmacokinetic status. However, challenges arise from genetic variant heterogeneity, from reduced penetrance whereby many carriers will not develop PD, and in recruiting a population that will meet the desired outcome in the proposed study duration. In this review, we discussed the opportunities afforded by the enrollment of genetically stratified cohorts (i.e., leucine-rich repeat kinase 2 and glucocerebrosidase 1) into prevention trials with a primary focus on primary prevention trials. We also outlined challenges surrounding the enrollment of these cohorts and offered suggestions to leverage their many advantages.

Antibiotic use in end-of-life care patients: A nationwide Veterans’ Health Administration cohort study

Background: Antibiotic use during end-of-life (EOL) care is an increasingly important target for antimicrobial stewardship given the high prevalence of antibiotic use in this setting with limited evidence on safety and effectiveness to guide antibiotic decision making. We estimated antibiotic use during the last 6 months of life for patients under hospice or palliative care, and we identified potential targets (ie time points) during the EOL period when antimicrobial stewardship interventions could be targeted for maximal benefit. Methods: We conducted a retrospective cohort study of nationwide Veterans’ Affairs (VA) patients, 18 years and older who died between January 1, 2014, and December 31, 2019, and who had been hospitalized within 6 months prior to death. Data from the VA’s integrated electronic medical record (EMR) were collected including demographics, comorbid conditions, and duration of inpatient antibiotics administered, along with outpatient antibiotics dispensed. A propensity-score matched-cohort analysis was conducted to compare antibiotic use between patients placed into palliative care or hospice matched to patients not receiving palliative care or hospice care. Repeated measures ANOVA and repeated measures linear regression methods were used to analyze difference in difference (D-I-D) of days of therapy (DOT) between the 2 cohorts. Results: There were 251,822 patients in the cohort, including 23,746 in hospice care, 89,768 in palliative care, and 138,308 without palliative or hospice care. The median days from last discharge to death was 9 days. The most common comorbidities were chronic obstructive pulmonary diseases (50%), malignancy (46%), and diabetes mellitus (43%). Overall, 18,296 (77%) of 23,746 hospice patients, and 71,812 (80%) of 89,768 palliative care patients received at least 1 antibiotic, whereas 95,167 (69%) of 138,308 who were not placed in hospice or did not receive palliative care received antibiotics. In the primary matched cohort analysis that compared patients placed into hospice or palliative care to propensity-score matched controls, entry into palliative care was associated with a 11% absolute increase in antibiotic prescribing, and entry into hospice was associated with a 4% absolute increase during the 7–14 days after entry versus the 7–14 days before entry (Fig. 1). The stratified cohorts had very similar balanced covariates as the overall cohort. Conclusions: In our large cohort study, we observed that patients receiving EOL care had high levels of antibiotic exposure across VA population, particularly on entry to hospice or during admissions when they received palliative care consultation. Future studies are needed to identify the optimal EOL strategies for collaboration between antimicrobial stewardship and palliative care.Funding: NoneDisclosures: None

Evaluation of multiple myeloma clinical decision support tool value for assessing treatment selection variation when populated with community health system data.

e18781 Background: Clinical decision support (CDS) technology has the potential to improve health outcomes by offering physicians an informational resource to support review and application of best practices. The Multiple Myeloma Research Foundation (MMRF) and Intermountain Healthcare (IMH) conducted a study to assess the suitability of a single health system’s data for a myeloma-specific CDS tool that visualizes treatment pathways, and to assess the effort needed to support a CDS program. This research is part of a longer-term effort to explore how CDS technology can help: Increase awareness of and apply treatment guidelines by visualizing pathways for specific MM patient cohorts, Improve understanding of treatment variation for quality improvement within healthcare systems, Improve outcomes research by visualizing relationships between treatments and outcomes. This abstract focuses on the second use case, showing suitability of community health system data to assess treatment variability within the health system. Methods: IA12 data from the CoMMpass study was used to create a CDS tool prototype. These data were aggregated into state and transition maps to identify nodes and pathways with corresponding outcomes, including response, progression-free survival (PFS), and overall survival (OS). We also tested if EMR data from a community health system (i.e., IMH) could support such visualization. Inclusion criteria included patients with active MM between January 2016–June 2018; adult aged 18 years to 89 years at diagnosis of active or smoldering MM. An IMH-specific data dictionary was assessed for variable importance, quantity, and ease of acquisition. Results: Ninety-six of an initial 146 patients meeting eligibility criteria had sufficient data usable for the study, reflecting 44 unique drug combinations across 9 lines of therapy. The tool was able to associate and visualize all patients and their clinical states and transitions to their outcomes. Clinical data was typically complete (99% of the time), including key clinician-derived data, such as ECOG scores (78%) and treatment response (99%). Conclusions: The IMH portion of the study supports the hypothesis that a community health system can provide sufficiently high-quality information to power a CDS tool with priority features including display of treatment selection variability. Only 65% (96/146) of the initial study group had usable data because some patients had received partial care outside of the IMH integrated delivery network (IDN) leaving associated data inaccessible. Initial biostatistical analysis suggests that roughly 750-1000 complete patient records would be required for statistically significant outcomes research with granularly stratified cohorts. The MMRF plans to recruit additional large IDNs to obtain the patients to power more generalizable functionality.

878 Comparing Treatment Outcomes for Unruptured Brain Arteriovenous Malformations: A Retrospective Cohort Study

Abstract Background Unruptured brain arteriovenous malformations (bAVMs) carry a lifetime risk of haemorrhage. Treatment strategies include conservative management, microsurgical excision, endovascular treatment (EVT) and radiosurgery (SRS). Optimal treatment selection remains unclear. Method A single-centre retrospective cohort study of adult unruptured bAVMs (2007-2019). Patients who underwent intervention were propensity matched using baseline features (age, sex, size, deep drainage, eloquence, and Spetzler-Martin grade) with patients conservatively managed. Rates of neurological disability and mortality due to intervention or bleed were compared. Results 137 patients (mean age 48 years [SD = 16], males 64) were included; 34 (25%) EVT, 20 (15%) surgery, 31 (22%) SRS and 51 (37%) conservative. After a median follow-up of 49 months (IQR 23-75), rates of disability were as follows: surgery 35%, EVT 21%, SRS 13% and conservative 8%. Matched cohorts (intervention/conservative) were: surgery-19/18, SRS-30/22 and EVT-33/34. Comparison of disability rates across matched cohorts revealed no statistically significant differences (surgery p = 0.07, SRS p = 0.65 and EVT p = 0.11). Three conservatively managed patients died. Conclusions Unruptured bAVMs carry a significant risk of neurological morbidity, regardless of intervention choice. Treatment choice may have an impact on patient outcomes but requires investigation of stratified cohorts. Findings are consistent with the nuances of AVM treatment selection.

Integrative analysis identifies an immune-relevant epigenetic signature for prognostication of non-G-CIMP glioblastomas

ABSTRACT The clinical and molecular implications of DNA methylation alterations remain unclear among the majority of glioblastomas (GBMs) without glioma-CpGs island methylator phenotype (G-CIMP); integrative multi-level molecular profiling may provide useful information. Independent cohorts of non-G-CIMP GBMs or IDH wild type (wt) lower-grade gliomas (LGGs) from local and public databases with DNA methylation and gene expression microarray data were included for discovery and validation of a multimarker signature, combined using a RISK score model. Bioinformatic and in vitro functional analyses were employed for biological validation. Using a strict multistep selection approach, we identified eight CpGs, each of which was significantly correlated with overall survival (OS) of non-G-CIMP GBMs, independent of age, the O-6-methylguanine-DNA methyltransferase (MGMT) methylation status, treatments and other identified CpGs. An epigenetic RISK signature of the 8 CpGs was developed and validated to robustly and independently prognosticate prognosis in different cohorts of not only non-G-GIMP GBMs, but also IDHwt LGGs. It also showed good discriminating value in stratified cohorts by current clinical and molecular factors. Bioinformatic analysis revealed consistent correlation of the epigenetic signature to distinct immune-relevant transcriptional profiles of GBM bulks. Functional experiments showed that S100A2 appeared to be epigenetically regulated by one identified CpG and was associated with GBM cell proliferation, apoptosis, invasion, migration and immunosuppression. The prognostic 8-CpGs RISK score signature may be of promising value for refining current glioma risk classification, and its potential links to distinct immune phenotypes make it a promising biomarker candidate for predicting response to anti-glioma immunotherapy.

Evaluation of Multiple Myeloma Clinical Decision Support Tool Value When Populated with Community Health System Data

Introduction Clinical decision support (CDS) technology has the potential to improve health outcomes by offering physicians an informational resource to support review and application of best pratices.1 The Multiple Myeloma Research Foundation (MMRF) and Intermountain Healthcare (IMH) conducted a study to assess the suitability of a single health system's data for a myeloma-specific CDS tool that visualizes treatment pathways, and to assess the effort needed to support a CDS program.2 This research is part of a longer-term effort to explore how CDS technology can help: - increase awareness of and apply treatment guidelines by visualizing pathways for specific MM patient cohorts - improve understanding of treatment variation within health systems - improve outcomes research by showing relationships between treatments and outcomes Methods IA12 data from the CoMMpass study3 was used to create a CDS tool prototype. These data were aggregated into state and transition maps to identify nodes and pathways with corresponding outcomes, including response, progression-free survival (PFS), and overall survival (OS). Intervening patient states were displayed using Sankey diagrams [Fig. 1]. We also tested if EMR data from a community health system (i.e., IMH) could support such visualization. The team designed a study protocol and obtained IRB approval. Inclusion criteria included patients with active MM between January 2016-June 2018; adult aged 18 years to 89 years at diagnosis of active or smoldering MM. An IMH-specific data dictionary was assessed for variable importance, quantity, and ease of acquisition. [Table 1]. IMH then manually abstracted prioritized structured (eg: labs) and non-structured (eg: notes) data for use in the tool. Results Ninety-six of an initial 146 patients meeting eligibility criteria had sufficient data usable for the study, reflecting 44 unique drug combinations across 9 lines of therapy. The tool was able to associate and visualize all patients and their clinical states and transitions to their outcomes. Clinical data was typically complete (99% of the time), including key clinician-derived data, such as ECOG scores (78%) and treatment response (99%). 569 person-hours were required to conduct the abstraction activity on 96 cases, averaging 5.9 hours/patient Discussion The IMH portion of the study supports the hypothesis that a community health system can provide sufficient high-quality information to power a CDS tool with priority features. Only 65% (96/146) of the initial study group had usable data because some patients had received partial care outside of the IMH integrated delivery network (IDN) leaving associated data inaccessible. Initial biostatistical analysis suggests that roughly 750-1000 complete patient records would be required for statistically significant outcomes research with granularly stratified cohorts. The MMRF is currently recruiting 5-7 additional large IDNs to obtain the patients to power more generalizable functionality. References 1 McKie PM, Kor DJ, Cook DA, Kessler ME, Carter RE, Wilson PM, et al. Computerized advisory decision support for cardiovascular diseases in primary care: a cluster randomized trial. Am J Med [Internet]. 2019 Dec 18 [cited 2020 Mar 5]. Available from: https://doi.org/10.1016/j.amjmed.2019.10.039 2 Garcelon N, Burgun A, Salomon R, Neuraz A. Electronic health records for the diagnosis of rare diseases. Kidney Int [Internet]. 2020 Jan 14 [cited 2020 Mar 5]. Available from: https://doi.org/10.1016/ j.kint.2019.11.037 3 Christofferson A, Nasser S, Aldrich J, Penaherrera D, Legendre C, Benard B, et al. Integrative analysis of the genomic landscape underlying multiple myeloma at diagnosis: an Mmrf Commpass analysis. Blood. 2017 Dec 7; 130 (Supplement 1): 326. Disclosures No relevant conflicts of interest to declare.

PCN8 Real World Treatment Patterns, Healthcare Resource Utilization (HCRU), Economic Outcomes and Survival Associated with Esophageal Cancer in JAPAN

Describe the demographics, clinical status, treatment patterns, HCRU, economic outcomes and survival of patients with esophageal cancer (EC) in Japan. This retrospective observational study included patients with a primary ICD-10 code of EC in ≥2 hospital claims from Jan 1st, 2012 through Dec 31st, 2018 in the Medical Data Vision Co., Ltd (Tokyo, Japan) database of records from >300 hospitals across Japan. The first date of an EC admission was defined as the index date. Patients were followed for a minimum ± 30 days from the index date and were stratified by EC stage and clinical presentation (ie, resectable vs. non-resectable advanced). Key characteristics such as demographics, clinical parameters, medication utilization, HCRU, costs incurred, and survival were tracked for the overall population and stratified cohorts. The included cohort (N=6645) were predominantly male (n=5607, 84.4%) former or current smokers (n=3948, 59.4%) of mean (SD) age 69.2 (9.2) years. At the index date, 20.5% (n=1364) of patients had resectable, 14.8% (n=983) non-resectable advanced and 64.7% (n=4298) unknown clinical presentation; 36.0% had EC of stage 0–2, 35.5% stage 3–4 and 28.5% stage unknown. After the index date, nearly half the patients (n=3241, 48.8%) received first line (1L) chemotherapy with the most common being cisplatin + fluorouracil combination (n=1517, 46.8% of 1L patients) and fluorouracil monotherapy (n=566, 17.5% of 1L patients). The annualized EC-related HCRU costs for inpatient, outpatient and pharmacy was ¥ 731,000 in 2019¥. The average patient follow-up was 19.3 months (95% CI: 2.0 – 58.4) with the majority of patients (n=5178, 77.9%) alive at the end of the observation period. In this cohort of patients with EC, the most common 1L chemotherapies were fluoropyrimidine and platinum-containing combinations. EC-related HCRU costs appear to be a substantial financial burden in Japan.

Identification of a novel prognostic DNA methylation signature for lung adenocarcinoma based on consensus clustering method.

Abnormal DNA methylation persists throughout carcinogenesis and cancer development. Hence, gene promoter methylation may act as a prognostic tool and provide new potential therapeutic targets for patients with lung adenocarcinoma (LUAD). In this study, to explore prognostic methylation signature, data regarding DNA methylation and RNA‐seq, and clinical data of patients with LUAD from the Cancer Genome Atlas database (TCGA) were downloaded. After data preprocessing, the methylation data were divided into training (N = 405) and test sets (N = 62). Then, patients in the training set were assigned to five subgroups based on their different methylation levels using the consensus clustering method. We comprehensively analyzed the survival information, methylation levels, and clinical variables, including American Joint Committee on Cancer (AJCC) stage, tumor‐node‐metastasis (TNM) staging, age, smoking history, and gender of these five groups. Subsequently, we identified a 16‐CpG prognostic signature and constructed a prognostic model, which was verified in the test set. Further analyses showed stable prognostic performance in the stratified cohorts. In conclusion, the new predictive DNA methylation signature proposed in this study may be used as an independent biomarker to assess the overall survival of LUAD patients and provide bioinformatics information for development of targeted therapy.

Identification of a costimulatory molecule-based signature for predicting prognosis risk and immunotherapy response in patients with lung adenocarcinoma

ABSTRACT Background Costimulatory molecules play significant roles in mounting anti-tumor immune responses, and antibodies targeting these molecules are recognized as promising adjunctive cancer immunotherapies. Here, we aim to conduct a first full-scale exploration of costimulatory molecules from the B7-CD28 and TNF families in patients with lung adenocarcinoma (LUAD) and generated a costimulatory molecule-based signature (CMS) to predict survival and response to immunotherapy. Methods We enrolled 1549 LUAD cases across 10 different cohorts and included 502 samples from TCGA for discovery. The validation set included 970 cases from eight different Gene Expression Omnibus (GEO) datasets and 77 frozen tumor tissues with qPCR data. The underlying mechanisms and predictive immunotherapy capabilities of the CMS were also explored. Results A five gene-based CMS (CD40LG, TNFRSF6B, TNFSF13, TNFRSF13C, and TNFRSF19) was initially constructed using the bioinformatics method from TCGA that classifies cases as high- vs. low-risk groups per OS. Multivariable Cox regression analysis confirmed that the CMS was an independent prognostic factor. As expected, CMS exhibited prognostic significance in the stratified cohorts and different validation cohorts. Additionally, the prognostic meta-analysis revealed that CMS was superior to the previous signature. Samples in high- and low-risk groups exhibited significantly different tumor-infiltrating leukocytes and inflammatory activities. Importantly, we found that the CMS scores were closely related to multiple immunotherapy biomarkers. Conclusion We conducted the first and most comprehensive costimulatory molecule landscape analysis of patients with LUAD and built a clinically feasible CMS for prognosis and immunotherapy response prediction, which will be helpful for further optimize immunotherapies for cancer.

Calcium phosphate injection of symptomatic bone marrow lesions of the knee: what is the current clinical evidence?

BackgroundChronic bone marrow lesions (BML) in the weight-bearing portions of the knee are often associated with symptomatic degenerative arthritis resulting in pain and dysfunction. Injection of bone substitute material like calcium phosphate has been described. Whilst some studies have reported encouraging results others have shown limited benefit of this technique.AimThe aim was to collate the available evidence on the injection of calcium phosphate and systematically evaluate the results to answer the questions encountered in clinical decision making: (1) does it provide effective long-lasting pain relief to avoid further surgical intervention? (2) which factors (patient/surgical) significantly influence the outcome? and (3) does it adversely affect the outcomes of subsequent arthroplasty?MethodsA literature search was performed to identify the studies describing the clinical outcomes of calcium phosphate injection for treatment of BML. We evaluated the reported clinical outcomes with respect to pain, function and complications. Isolated case reports and studies with no objective assessment of clinical outcomes were excluded.ResultsWe noted 46 articles in the current literature of which 8 described clinical outcomes of calcium phosphate injection. Mean (plus/minus SD) score on the visual analog scale (VAS) has been reported to improve from 7.90 (± 0.38) to 2.76 (± 0.90), whereas the International Knee Documentation Committee (IKDC) score improved from 30.5 (SD not reported (NR)) to 53.0 (SD NR). Pre and post procedure Short form survey (SF-12) scores were 29.8 (SD NR) and 36.7 (SD NR), respectively. In one study, scores on the Tegner Lysholm knee scoring scale improved in 12 out of 22 patients, whereas the remainder had no change in symptoms. Extravasation of calcium phosphate into the joint was the most common complication, whereas no adverse effect has been reported on subsequent arthroplasty.ConclusionLimited data from the published studies would suggest that calcium phosphate injection of BML may potentially improve pain and function. However, no evidence is currently available to clearly identify patient/surgical factors that may influence the long-term outcomes of this procedure. Hence pragmatic, prospective studies with stratified patient cohorts and robust reporting of outcome measures are essential to improve the understanding of the indications and clinical effectiveness of this novel procedure.

Systematic construction and validation of an immune prognostic model for lung adenocarcinoma.

Lung adenocarcinoma (LUAD), the most common non‐small‐cell lung cancer, is characterized by a dense lymphocytic infiltrate, which indicates that the immune system plays an active role in the development and growth of this cancer. However, no investigations to date have proposed robust models for predicting survival outcome for patients with LUAD in terms of tumour immunology. A total of 761 LUAD patients were included in this study, in which the database of The Cancer Genome Atlas (TCGA) was utilized for discovery, and the Gene Expression Omnibus (GEO) database was utilized for validation. Bioinformatics analysis and R language tools were utilized to construct an immune prognostic model and annotate biological functions. Lung adenocarcinoma showed a weakened immune phenotype compared with adjacent normal tissues. Immune‐related gene sets were profiled, an immune prognostic model based on 2 immune genes (ANLN and F2) was developed with the TCGA database to distinguish cases as having a low or high risk of unfavourable prognosis, and the model was verified with the GEO database. The model was prognostically significant in stratified cohorts, including stage I‐II, stage III‐IV and epidermal growth factor receptor (EGFR) mutant subsets, and was considered to be an independent prognostic factor for LUAD. Furthermore, the low‐ and high‐risk groups showed marked differences in tumour‐infiltrating leucocytes, tumour mutation burden, aneuploidy and PD‐L1 expression. In conclusion, an immune prognostic model was proposed for LUAD that is capable of independently identifying patients at high risk for poor survival, suggesting a relationship between local immune status and prognosis.

Systematic profiling of a novel prognostic alternative splicing signature in hepatocellular carcinoma.

Alternative splicing (AS) is a pervasive and vital mechanism involved in the progression of cancer by expanding genomic encoding capacity and increasing protein complexity. However, the systematic analysis of AS in hepatocellular carcinoma (HCC) is lacking and urgently required. In the present study, genome-wide AS events with corresponding clinical information were profiled in 290 patients with HCC from the Cancer Genome Atlas and SpliceSeq software. Functional enrichment analyses revealed the pivotal biological process of AS regulation. Univariate Cox regression analyses were performed, followed by stepwise forward multivariate analysis to develop the prognostic signatures. Spearman's correlation analyses were also used to construct potential regulatory network between the AS events and aberrant splicing factors. A total of 34,163 AS events were detected, among which 1,805 AS events from 1,314 parent genes were significantly associated with the overall survival (OS) of patients with HCC, and their parent genes serve crucial roles in HCC-related oncogenic processes, including the p53 signaling pathway, AMPK signaling pathway and HIF-1 signaling pathway. A prognostic AS signature was established that was found to be an independent prognostic factor for OS in stratified cohorts, harboring a noteworthy ability to distinguish between the distinct prognoses of patients with HCC (high-risk vs. low-risk, 827 vs. 3,125 days, P<2e-16). Time-dependent receiver-operator characteristic curves confirmed its robustness and clinical efficacy, with the area under the curves maintained >0.9 for short-term and long-term prognosis prediction. The splicing correlation network suggested a trend in the interactions between splicing factors and prognostic AS events, further revealing the underlying mechanism of AS in the oncogenesis of HCC. In conclusion, the present study provides a comprehensive portrait of global splicing alterations involved in the progression and HCC in addition to valuable prognostic factors for patients, which may represent as underappreciated hallmark and provide novel clues of therapeutic targets in HCC.