Prospective Testing Research Articles

BIOM-05. FUSION TRANSCRIPTOME LANDSCAPE IN GLIOBLASTOMA

Abstract BACKGROUND For most patients with glioblastoma (GBM), no truly effective salvage therapies exist. However, there are compelling examples in oncology where gene fusions lead to functioning oncoproteins that are highly effective targets (ie BCR-ABL1 in CML). Using a large genomic database, we sought to determine the prevalence of fusion transcripts in GBMs. METHODS A total of 4392 GBMs were tested at Caris Life Sciences (Phoenix, AZ). Mutations were detected by NextGen sequencing of DNA (592-gene panel or whole exome sequencing); gene fusions were detected by Whole Transcriptome Sequencing. All tumors were wild type for IDH1/2. Significance was tested by Fisher-Exact and ChiSquare and adjusted by Benjamini-Hochberg (q <0.05). RESULTS Overall, pathogenic fusions were found in 428 (9.7%). The most prevalent were FGFR3 [N=159 (37%), highest FGFR3:TACC3, N=134], MET [N=92 (21%), highest PTPRZ1:Met, N=31; ST7:MET N=25; CAPZA2: MET, N=23), EGFR [N=87(20%), highest EGFR:SETP14, N=21 and SEC61G: EGFR, N=19], NTRK2 (N=27), PDGFRA (N=23), ROS1 (N=14) and BRAF (N=10). Other actionable fusions including ALK (N=3), NTRK3 (N=3), RAF1(N=3), RET (N=3), NTRK1(N=2) were seen. Tumors with fusions were more likely to have pathogenic gene amplifications including: MET (7.5% vs. 0.7%), FGFR3 (5% vs. 0.2%), CDK4 (17% vs. 11%) and MDM2 (12% and 7.5%) but less frequent EGFR mutation (6.1% vs. 18%), amplification (6.1% vs. 18%) and EGFRvIII mutation (11.9% vs. 22.5%), RB1 (2.9% vs. 11.5%) and TP53mutation (22% vs. 30%) (all q<0.05). EGFR fusions are enriched with EGFR amplification (92%) and EGFRvIII mutation (51%); cMET fusion with MET amplification (40%). CONCLUSIONS Comprehensive molecular profiling reveals that approximately 10% of IDH WT GBMs carry oncogenic fusions that may be therapeutic targets. Broad spectrum of observed fusions underscores the need for novel clinical trial designs to allow efficient enrollment for prospective testing of potential targeted agents.

Tumor regression after neoadjuvant hormonal therapy in high risk prostate cancer: pathological outcomes from a randomized phase II trial.

High-risk localized prostate cancer (HRLPC) commonly progresses to metastatic disease after local treatment. Neoadjuvant androgen deprivation therapy (nADT) before radical prostatectomy (RP) has recently been suggested to improve early oncological outcomes in HRLPC. We aimed to perform an exploratory analysis of the pathological outcomes from a prospective trial testing nADT before RP. Prospective, single-centered, phase II, randomized trial performed between October 2018 and July 2021. Random assignment (1:1) for nADT modalities: goserelin (10.8mg) plus abiraterone acetate (1000mg/d) plus prednisone (5mg/d), with or without apalutamide (240mg/d) for 12 weeks, followed by RP (within 30 days) and extended lymph node dissection. Baseline clinical and pathological variables were assessed in needle biopsies before nADT. Tumor regression was histologically evaluated in surgical specimens using the residual cancer burden index (RCB). Sixty-two patients reached the surgical phase. Good response (RCB ≤ 0.25cm³) was achieved in 14 patients (22.5%). Overall stage migration rate between baseline status (MRI before nADT) and final status (after surgery) was 27.4%. Late stage detection (high tumor burden, perineural invasion) and altered PTEN/ERG immunostatus showed significant association with poor response in univariate analysis. Higher baseline tumor burden was the only independent factor related to poor response in multivariate analysis. There are subgroups of patients, such as those with low baseline cancer burden and PTEN/ERG wild-type status, more likely to achieve good response with nADT. In the case of long term oncological benefit to be proven, nADT might be an additional therapeutic resource for these patients.

Wastewater Surveillance for Poliovirus in Selected Jurisdictions, United States, 2022-2023.

Wastewater testing can inform public health action as a component of polio outbreak response. During 2022-2023, a total of 7 US jurisdictions (5 states and 2 cities) participated in prospective or retrospective testing of wastewater for poliovirus after a paralytic polio case was identified in New York state. Two distinct vaccine-derived poliovirus type 2 viruses were detected in wastewater from New York state and New York City during 2022, representing 2 separate importation events. Of those viruses, 1 resulted in persistent community transmission in multiple New York counties and 1 paralytic case. No poliovirus was detected in the other participating jurisdictions (Connecticut, New Jersey, Michigan, and Illinois and Chicago, IL). The value of routine wastewater surveillance for poliovirus apart from an outbreak is unclear. However, these results highlight the ongoing risk for poliovirus importations into the United States and the need to identify undervaccinated communities and increase vaccination coverage to prevent paralytic polio.

Beta-Lactam Antibiotic Exposure During Pediatric Extracorporeal Membrane Oxygenation: Retrospective Cohort Analysis of Drug Levels Using Standard Dosing, 2018–2020

Objectives: Children on extracorporeal membrane oxygenation (ECMO) are at high risk of infection that may worsen prognosis. Even though treatment with beta-lactam antibiotics is frequent, dosing is not adapted to altered pharmacokinetic and pharmacodynamic characteristics of children on ECMO. There is, therefore, a risk of inadequate drug levels when using standard dosing. In this study, we aimed to describe beta-lactam exposures of children on ECMO using current dosing and to identify factors associated with inadequate exposure. The optimal pharmacokinetic/pharmacodynamic target was considered as a plasma concentration four times above the minimum inhibitory concentration throughout the dosing interval target. Design: Two-center retrospective cohort study. Setting: Two PICUs in Paris, France. Patients: Children (from birth to 18 yr) undergoing venovenous or venoarterial ECMO, from 2018 to 2020. Interventions: None. Measurements and Main Results: There were 57 patients who received 11 different beta-lactams, with 226 plasma concentrations analyzed. A total of 32 infections were documented. Overall, 133 of 226 concentrations (58.8%) were insufficient, primarily in samples from children younger than 28 days (p = 0.035), with low body weight (p = 0.013), or in instances of hypoalbuminemia (p = 0.011) and increased renal clearance (p = 0.032). Supratherapeutic concentrations were observed in 25 of 226 samples (11.1%), associated with being taken from patients with renal impairment (p < 0.01). Conclusions: In this retrospective cohort of pediatric ECMO cases, there is an associated risk of underexposure when prescribing conventional dosing of beta-lactams, which are likely associated with renal impairment and fluid overload. Prospective testing of therapeutic drug monitoring combined with pharmacokinetic/pharmacodynamic models should be tested as a risk-reduction strategy in this vulnerable population.

Changes in Bladder Volume During Radiotherapy for Prostate Cancer.

Patients irradiated for prostate cancer may experience urinary toxicity, particularly if the bladder volume is small. A mobile application (app) that reminds the patients to drink water prior to each radiation fraction may help avoid small volumes. This study investigating bladder volumes during a radiotherapy course is a prerequisite for a prospective trial testing such a reminder app. Frequency of bladder volumes <200 ml and seven potential risk factors were retrospectively evaluated in 72 patients receiving external beam radiotherapy for non-metastatic prostate cancer. The mean and median values of the numbers of radiation fractions with bladder volumes <200 ml were 17.8 (standard deviation=12.0) and 16.5 (interquartile range Q1-Q3=7.5-29.5) fractions, respectively. Higher numbers of fractions with volumes <200 ml were significantly associated with pre-radiotherapy bladder volumes <200 ml (p<0.001) and high-risk prostate cancer (p=0.014). The proportion of bladder volumes <200 ml during the radiotherapy course was high and needs to be decreased. Pre-radiotherapy bladder volume and risk level of prostate cancer were significant risk factors for higher numbers of fractions with volumes <200 ml. These results are important for designing a prospective trial.

Construction of platelet count-optical method reflex test rules using Micro-RBC#, Macro-RBC%, "PLT clumps?" flag, and "PLT abnormal histogram" flag on the Mindray BC-6800plus hematology analyzer in clinical practice.

Utilizing RBC or PLT-related parameters to establish rules for the PLT-O reflex test can assist laboratories in quickly identifying specimens with interfered PLT-I that require PLT-O retesting. Prospective PLT-I and PLT-O testing was performed on 6857 EDTA-anticoagulated whole blood samples, split randomly into training and validation cohorts at a 2:3 ratio. Reflex and non-reflex groups were distinguished based on the differences between PLT-I and PLT-O results. By comparing RBC and PLT parameter differences and flags in the training set, we pinpointed factors linked to PLT-O reflex testing. Utilizing Lasso regression, then refining through univariate and multivariate logistic regression, candidate parameters were selected. A predictive nomogram was constructed from these parameters and subsequently validated using the validation set. ROC curves were also plotted. Significant differences were observed between the reflex and non-reflex groups for 19 parameters including RBC, MCV, MCH, MCHC, RDW-CV, RDW-SD, Micro-RBC#, Micro-RBC%, Macro-RBC#, Macro-RBC%, MPV, PCT, P-LCC, P-LCR, PLR,"PLT clumps?" flag, "PLT abnormal histogram" flag, "IDA Anemia?" flag, and "RBC abnormal histogram" flag. After further analysis, Micro-RBC#, Macro-RBC%,"PLT clumps?", and "PLT abnormal histogram" flag were identified as candidate parameters to develop a nomogram with an AUC of 0.636(95 %CI: 0.622-0.650), sensitivity of 42.9 % (95 %CI: 37.8-48.1 %), and specificity of 90.5 % (95 %C1: 89.6-91.3 %). The established rules may help laboratories improve efficiency and increase accuracy in determining platelet counts as a supplement to ICSH41 guidelines.

1838P AI-based smart oncology follow-up system: Prospective application testing and enhancement of clinical efficacy

1838P AI-based smart oncology follow-up system: Prospective application testing and enhancement of clinical efficacy

Evaluation of a BioFire multiplex PCR panel for detection of joint infections using retrospective and prospectively collected specimens.

The BioFire Joint Infection Panel (JI panel) is a newly FDA-approved multiplex PCR assay for detection of common bone and joint pathogens with 39 targets which include select Gram-positive and Gram-negative bacteria, yeast, and antimicrobial resistance genes. We evaluated the performance of the JI panel in detecting joint infections in our patient population. Sixty-three frozen, residual joint fluid specimens were retrospectively tested using the JI panel. An additional 104 residual joint fluid specimens were de-identified and prospectively tested within 1 week of collection. Results from routine bacterial cultures were used as the reference standard, which included inoculation to agar plates and blood culture bottles. For the frozen specimens, the JI panel showed a positive percent agreement (PPA) of 92.8% and a negative percent agreement (NPA) of 97.1%. PPA was 71.4% and NPA was 94.8% for fresh specimens. A total of 12 discrepancies were observed among the 167 specimens tested. The JI panel demonstrated good overall agreement with routine culture for the detection of joint infections and may improve timely diagnosis when used in conjunction with bacterial culture. However, potential false-positive and false-negative results were observed in both retrospective and prospective testing of specimens.IMPORTANCEThe BioFire JI panel is a new commercially available multiplex PCR assay for detecting common pathogens causing bone and joint infections. The test is performed directly on joint fluids with a fast turnaround time of 1 hour. Our study shows that while the JI panel overall shows good agreement with routine culture, discrepancies were observed in 7% of cases and results should be interpreted with appropriate clinical context.

Social-Ecological Correlates of Social Well-Being in Transgender and Gender Diverse Adults in the USA: Implications for Policy, Theory, and Research

Abstract Introduction Emerging transgender and gender diverse (TGD) health models focus on social well-being. The present study contributes to this movement through investigation of hate crime statutes, state-level political and anti-TGD attitudes, and multiple marginalized identities as correlates of social well-being (i.e., outness, community connection, and workplace and familial support). Methods The current study merged data from three sources: (a) the 2015 United States Transgender Survey (USTS), (b) the 2014 and 2018 Cooperative Congressional Election Studies (CCES), and (c) the 2018 Anti-Defamation League’s (ADL) Hate Crime Map. Bivariate (e.g., chi-square) and regression analyses were used to identify correlates of social well-being among TGD adults in the USA. Results All types of hate crime laws showed bivariate protective associations for outness and family support. Regression models showed that several marginalized identities (e.g., having a disability) were the most consistent and largest correlates of social well-being. Higher state-level anti-TGD attitudes were linked with lower familial support and outness. The presence of state-level hate crime law civil action and police training statutes was positively associated with outness. Conclusions The results extend the limited research on correlates of TGD social well-being. Findings support TGD health models addressing resilience and social safety. Future research should address hate crime law policy analysis, police training evaluation, and prospective testing of TGD social well-being. Policy Implications Hate crime laws, in particular civil action and police training statutes, may be associated with better outcomes for transgender and gender diverse (TGD) persons. Evaluation of hate crime law effectiveness may need to focus on implementation as opposed to mere presence of laws. Policies specific to police training should include guidelines for training content and rigorous evaluation.

O-131 A novel approach to predicting the potential of a blastocyst using untargeted spectral analysis of spent culture medium and machine learning

Abstract Study question Is it possible to effectively predict negative outcomes of intrauterine embryo transfer using modified untargeted Raman spectroscopy of the spent culture medium? Summary answer The proposed spectroscopic approach, combined with machine learning methods, allows identifying a group of embryos that are almost unable to implant and develop ongoing pregnancy. What is known already Embryo deselection is usually based on the karyotype analysis of the embryo and originates from the assumption that an abnormal PGT result is associated with negative transfer outcomes. Embryo selection methods aimed at identifying the most capable for implantation embryo are based on morphological, morphometric, and metabolic approaches. However, due to the dependence of the phenomenon (pregnancy) on several factors, predicting the outcome of the transfer solely through variables derived from the embryo does not correspond to modern views on implantation. At the same time, predicting negative transfer outcomes through the information originating from the embryo appears quite rational. Study design, size, duration 424 samples of spent culture medium used, collected from microdrops with individually cultured blastocyst from the internal institutional biosamples repository. Medium samples were collected from July 1, 2021, to June 1, 2023. The class attribute for each sample was the absence or presence of ongoing pregnancy at the 12-14 week of gestational age. The positive class (ongoing pregnancy) proportion in the presented dataset is 33%, and the negative (no ongoing pregnancy) class proportion is 67%. Participants/materials, setting, methods The study material was spent culture medium from blastocysts with known transfer outcomes by the end of first trimester. All media samples were stored at -86 °С until the moment of spectroscopy. Before the study, samples were thawed at room temperature for 15 minutes. Untargeted Raman spectroscopy was done on a signal amplification substrate using a green laser (532 nm). The architecture selection and machine learning model training were performed using the Python library mcfly. Main results and the role of chance After building the model the following results for validation and test sets were obtained: So it is possible to identify a group of embryos that are almost incapable of producing ongoing pregnancy. This group constitutes about 40% of the total number of negative outcomes. The error of determining an embryo as being incapable for implantation and normal pregnancy development is about 6%, what is relatively low number. Limitations, reasons for caution The presented data are not multicenter; they were obtained from samples from one laboratory and using a limited set (only one manufacturer) of culture media. This may lead to reduced or nonexistent relevance of the model for data from other sources and conditions. Wider implications of the findings It is necessary to check the proposed method on an additional set of culture media from other manufactures and laboratories. This may lead to refining the parameters of the built neural network. The assumption of significant clinical value of the developed model needs to be checked through prospective testing. Trial registration number fpc-2023-ivf-01

Usability testing of the Set Brave Goals app, a continence goal-selection app for children with spina bifida

IntroductionThe Set Brave Goals app is the first digital health app (DHA) aimed at helping children with spina bifida (SB) aged 8-17 years old to select and track their urinary and fecal continence goals. Developed by children, parents, providers and researchers, its usability, or appropriateness to a purpose (“user-friendliness”), remains unknown. In an alpha testing stage, our aim was to determine the usability of the app prior to clinical use. MethodsWe recruited children with SB and their parents in clinic and via social media. A recruitment goal of 10 children exceeded industry standard of 5 participants (10 participants are expected to identify approximately 96% of usability problems). Participants downloaded and used the app for a week. They completed a questionnaire, including the System Usability Scale (SUS) and closed and open-ended questions. The SUS is a 10-item validated usability questionnaire most widely used in DHA usability testing (scores range from 0 to 100, 100 representing greatest usability). Median/mean SUS scores for DHAs are 68. Mann-Whitney-U and t-tests were used. ResultsTen children with SB participated (median age: 14 years old, 6 female, 8 shunted, 8 using bladder catheterizations). Twenty parents participated (17 mothers, median age: 42 years old). Median SUS score was 77.5 for children and 73.8 for parents, corresponding to “good” to “excellent” usability (Summary Figure). There were no significant differences between child/parent scores (p=0.69) or those for DHAs (p=0.11). It took a median 5-10 minutes to go through the app (2 parents felt it took too much time). All participants felt the app was easy to understand and use (100%). Most participants would recommend it to children and parents (children: 89%, parents: 80%), and believed it would be valuable to children (90%, 75%) and urologists (80%, 80%). Free text responses related to more varied color schemes, keeping free-text notes within the app and more flexible goal/alarm setting. These changes were incorporated into the app. CommentUsability testing prior to clinical launch identified areas for app improvement. Although this study sample met industry standards, findings are limited by a small group of participants. The app will undergo further refinement during prospective beta testing. ConclusionsThe Set Brave Goals app for children with SB has acceptable usability parameters. This justifies proceeding with wider use to formalize children’s continence goal setting/tracking and testing its value to children with SB and their healthcare providers. Internal review board approval1707207063

Real-world tumor BRCA testing patterns and PARP inhibitor use in HER2-negative metastatic breast cancer.

1113 Background: Testing for BRCA1/BRCA2 mutations (BRCAm) aids systemic treatment decisions for patients (pts) with metastatic breast cancer (mBC). This retrospective study explored real-world germline/tumor (g/t) BRCA testing patterns and PARP inhibitor (PARPi) use in US pts with mBC. Methods: Data from pts ≥18 years old diagnosed with HER2-negative mBC from 2014–2022 were captured in the nationwide Flatiron Health electronic health record-derived deidentified database. Patterns and timing of prospective or incidental BRCA testing were compared across mBC subtypes (hormone receptor-positive [HR+ mBC]/triple-negative mBC [mTNBC]) and disease stage (recurrent/ de novo). BRCAm prevalence was compared by testing modality and mBC subtype. PARPi initiation was assessed in pts with mBC diagnosed from 2018 onwards and a positive BRCAm test result. Clinical characteristics were compared by receipt of a BRCA test, testing modality, and PARPi initiation. Real-world overall survival (OS) by PARPi use was estimated using the Kaplan-Meier method. Results: Of 15,006 pts (including 12,349 with HR+ mBC; 1998 with mTNBC), 8178 (54.5%) did not receive any type of BRCA test, 4654 (31.0%) received a gBRCA test, 3075 (20.5%) received a tumor tissue DNA (ttDNA) test, and 1418 (9.4%) received a circulating tumor DNA (ctDNA) test. BRCA testing prevalence was numerically higher in pts with mTNBC (n=1164; 58.3%) than in pts with HR+ mBC (n=5323; 43.1%). As expected, testing prevalence was higher with ttDNA than with ctDNA (mTNBC: n=569 [28.5%] vs n=137 [6.9%]; HR+ mBC: n=2339 [18.9%] vs n=1224 [9.9%]). Over 93% of pts (recurrent and de novo) who received a tBRCA test were tested after an mBC diagnosis, and &gt;1/3 were tested after first-line therapy, with median times from diagnosis to a ttDNA and ctDNA test of 127 and 431 days, respectively. Both gBRCA and tBRCA testing rates increased from 2014–2022, and testing occurred earlier in pts with recurrent than de novo disease. Among pts with tBRCAm, 66.0% had HR+ mBC, and 25.6% had mTNBC. BRCA1 mutations were more common in mTNBC (59.0%) than in HR+ mBC (22.9%) tumors. Of 219 pts with tBRCAm tumors who were diagnosed with mBC from 2018 onwards, PARPi was initiated in 39/74 (52.7%) pts with gBRCAm, 15/71 (21.1%) pts who were gBRCAm-negative, and 19/74 (25.7%) pts with gBRCAm status unknown. Among pts with g or tBRCAm and available OS data, median OS was numerically longer in pts who received PARPi (32.3 months [95% CI: 24.2, 47.4], n=128) versus pts who did not (29.0 months [95% CI: 21.9, 31.9], n=204; median follow-up time 23.2 months). Conclusions: While NCCN guidelines recommend gBRCA testing for all pts with mBC, more than half of pts with HER2-negative mBC (56.9% of pts with HR+ mBC) did not receive any BRCA test. Among pts with tBRCAm, ~2/3 had HR+ disease. Despite the potential benefits of PARPi for pts with BRCAm mBC, utilization remained low from 2018–2022.

#2067 Prediction of gastrointestinal bleeding hospitalization risk in hemodialysis using machine learning techniques

Abstract Background and Aims INitiativeS on advancing Patients’ outcomes In REnal disease (INSPIRE) is an academia and industry collaboration set forth to identify critical investigations needed to advance the practice of medicine in nephrology. Gastrointestinal bleeding (GIB) is one of the most common types of bleeding events in the kidney dialysis population [1]. We aimed to develop a model to predict GIB hospitalization risk in a hemodialysis (HD) patient within 180 days. We evaluated advanced machine learning algorithm (XGBoost) and compared it to traditional machine learning algorithm (logistic regression) modeling techniques. Method We used data from a kidney care network from 2017 through 2020. We included data from adult dialysis patients (age ≥18 years). GIB related hospitalization was defined based on international classification of diseases (ICD) diagnosis codes recorded as the primary, secondary, or tertiary discharge reason for hospitalization. Two distinct models were created using XGBoost and logistic regression algorithms and using the same dataset. Both models were evaluated using metrics such as area under the receiver operating curve (AUROC), accuracy, sensitivity, and specificity. Missing data was imputed using mean for quantitative data and mode for qualitative data. The dataset then was randomly divided into 60% training, 20% validation and 20% test dataset. The test data, comprising unseen patients and data was used to evaluate the performance of the model. This means that the model had never encountered the test data during its learning phase, which included training the model and then validating the model. Results The incidence of 180-day GIB hospitalization was 1.12% in the HD population (n = 5 116 with GIB hospitalization/ n = 451 653 without GIB hospitalization), and consistent in the unseen test dataset (n = 465 with GIB hospitalization/ n = 38586 without GIB hospitalization). The XGBoost model showed higher predictive performance compared to logistic regression (Table). The AUROC was 0.72 (95% confidence interval (CI) 0.69, 0.74) for the XGBoost model. In comparison, the AUROC was 0.615 (95% CI 0.60, 0.64) for the logistic regression model. Specificity was 67% versus 58% for the XGBoost model versus logistic regression model respectively. However, sensitivity for both the models was 65%. The top predictors for major GIB were consistent in both models for many factors, yet some factors showed different importance on outcome prediction (Fig. 1). Higher risk of GIB hospitalization was associated with older age, lower ferritin levels, and recent all-cause hospitalizations in both models. The XGBoost showed high importance on outcome prediction for lower hemoglobin and higher serum 25 hydroxy (25OH) vitamin D values, and the logistic regression model showed high importance for higher amounts of saline delivered during an HD treatment and lower intact parathyroid hormone (iPTH) levels. Conclusion We found advanced machine learning prediction modeling (XGBoost) appears suitable for identifying a HD patient at risk for a GIB hospitalization in the next 180 days, and outperforms traditional machine learning (logistic regression) modeling techniques. Although both models showed identical sensitivity, the XGBoost model had higher specificity. Prospective testing is needed to confirm the model's performance. The association between bone mineral metabolism markers and GIB hospitalization risk is unexpected and warrants further investigation.

Predicting the recurrence of spontaneous intracerebral hemorrhage using a machine learning model.

Recurrence can worsen conditions and increase mortality in ICH patients. Predicting the recurrence risk and preventing or treating these patients is a rational strategy to improve outcomes potentially. A machine learning model with improved performance is necessary to predict recurrence. We collected data from ICH patients in two hospitals for our retrospective training cohort and prospective testing cohort. The outcome was the recurrence within one year. We constructed logistic regression, support vector machine (SVM), decision trees, Voting Classifier, random forest, and XGBoost models for prediction. The model included age, NIHSS score at discharge, hematoma volume at admission and discharge, PLT, AST, and CRP levels at admission, use of hypotensive drugs and history of stroke. In internal validation, logistic regression demonstrated an AUC of 0.89 and precision of 0.81, SVM showed an AUC of 0.93 and precision of 0.90, the random forest achieved an AUC of 0.95 and precision of 0.93, and XGBoost scored an AUC of 0.95 and precision of 0.92. In external validation, logistic regression achieved an AUC of 0.81 and precision of 0.79, SVM obtained an AUC of 0.87 and precision of 0.76, the random forest reached an AUC of 0.92 and precision of 0.86, and XGBoost recorded an AUC of 0.93 and precision of 0.91. The machine learning models performed better in predicting ICH recurrence than traditional statistical models. The XGBoost model demonstrated the best comprehensive performance for predicting ICH recurrence in the external testing cohort.

Validation of a risk score to differentiate autoimmune and viral encephalitis: a Nationwide Cohort Study in Denmark

BackgroundA score to differentiate autoimmune (AE) and viral encephalitis (VE) early upon admission has recently been developed but needed external validation. The objective of this study was to evaluate the performance of the score in a larger and more diagnostically diverse patient cohort.MethodsWe conducted a retrospective nationwide and population-based cohort study including all adults with encephalitis of definite viral (2015–2022) or autoimmune aetiology (2009–2022) in Denmark. Variables included in the score-model were extracted from patient records and individual risk scores were assessed. The performance of the score was assessed by receiver-operating characteristics (ROC) curve analyses and calculation of the area under the curve (AUC).ResultsA total of 496 patients with encephalitis [AE n = 90, VE n = 287 and presumed infectious encephalitis (PIE) n = 119] were included in the study. The score was highly accurate in predicting cases of AE reaching an AUC of 0.94 (95% CI 0.92–0.97). Having a score ≥ 3 predicted AE with a PPV of 87% and an NPV of 91%. The risk score was found to perform well across aetiological subgroups and applied to the PIE cohort resulted in an AUC of 0.88 (95% CI 0.84–0.93).ConclusionThe excellent performance of the score as reported in the development study was confirmed in this significantly larger and more diverse cohort of patients with encephalitis in Denmark. These results should prompt further prospective testing with wider inclusion criteria.

A non-invasive method to determine core temperature for cats and dogs using surface temperatures based on machine learning

BackgroundRectal temperature (RT) is an important index of core temperature, which has guiding significance for the diagnosis and treatment of pet diseases.ObjectivesDevelopment and evaluation of an alternative method based on machine learning to determine the core temperatures of cats and dogs using surface temperatures.Animals200 cats and 200 dogs treated between March 2022 and May 2022.MethodsA group of cats and dogs were included in this study. The core temperatures and surface body temperatures were measured. Multiple machine learning methods were trained using a cross-validation approach and evaluated in one retrospective testing set and one prospective testing set.ResultsThe machine learning models could achieve promising performance in predicting the core temperatures of cats and dogs using surface temperatures. The root mean square errors (RMSE) were 0.25 and 0.15 for cats and dogs in the retrospective testing set, and 0.15 and 0.14 in the prospective testing set.ConclusionThe machine learning model could accurately predict core temperatures for companion animals of cats and dogs using easily obtained body surface temperatures.

Reliable detection of generalized convulsive seizures using an off-the-shelf digital watch: A multisite phase 2 study.

The aim of this study was to develop a machine learning algorithm using an off-the-shelf digital watch, the Samsung watch (SM-R800), and evaluate its effectiveness for the detection of generalized convulsive seizures (GCS) in persons with epilepsy. This multisite epilepsy monitoring unit (EMU) phase 2 study included 36 adult patients. Each patient wore a Samsung watch that contained accelerometer, gyroscope, and photoplethysmographic sensors. Sixty-eight time and frequency domain features were extracted from the sensor data and were used to train a random forest algorithm. A testing framework was developed that would better reflect the EMU setting, consisting of (1) leave-one-patient-out cross-validation (LOPO CV) on GCS patients, (2) false alarm rate (FAR) testing on nonseizure patients, and (3) "fixed-and-frozen" prospective testing on a prospective patient cohort. Balanced accuracy, precision, sensitivity, and FAR were used to quantify the performance of the algorithm. Seizure onsets and offsets were determined by using video-electroencephalographic (EEG) monitoring. Feature importance was calculated as the mean decrease in Gini impurity during the LOPO CV testing. LOPO CV results showed balanced accuracy of .93 (95% confidence interval [CI] = .8-.98), precision of .68 (95% CI = .46-.85), sensitivity of .87 (95% CI = .62-.96), and FAR of .21/24 h (interquartile range [IQR] = 0-.90). Testing the algorithm on patients without seizure resulted in an FAR of .28/24 h (IQR = 0-.61). During the "fixed-and-frozen" prospective testing, two patients had three GCS, which were detected by the algorithm, while generating an FAR of .25/24 h (IQR = 0-.89). Feature importance showed that heart rate-based features outperformed accelerometer/gyroscope-based features. Commercially available wearable digital watches that reliably detect GCS, with minimum false alarm rates, may overcome usage adoption and other limitations of custom-built devices. Contingent on the outcomes of a prospective phase 3 study, such devices have the potential to provide non-EEG-based seizure surveillance and forecasting in the clinical setting.

Abstract PO1-08-04: Universal Germline Genetic Testing for Breast Cancer: Implementation in a Rural Practice and Impact on Shared Decision Making

Abstract Background: Prior work from our group has shown a high rate of familial breast cancer (BC) among rural BC patients (pts), with 33% reporting &amp;gt; 1 first-degree family members with BC and 55% meeting NCCN germline genetic testing (GGT) criteria (Shelton CH et al. NC Med. J; 2022). In 2018, we increased GGT to include all BC pts and observed: 1) many pts were out-of-criteria per NCCN guidelines but had actionable pathogenic germline variants (PGVs) and 2) knowledge of GGT factored significantly into shared decision making by providers and patients, including for positive and negative results. Beginning in 2019, concurrent with the recommendation from the American Society of Breast Surgeons, we expanded this model of universal GGT to an IRB-approved study to measure the impact of prospective testing on shared decision-making. Herein we report the real-world implementation of universal GGT for patients with BC and genetics-informed treatment decision-making in a rural community practice. Methods: From 2019-2022 all BC pts at a small, rural community hospital were offered GGT. Demographics, clinical features, and GGT results were collected along with clinician-reported clinical recommendations for treatment based on GGT results. Descriptive statistics and two-tailed Fisher’s exact test were employed and significance was set at p&amp;lt; 0.05. Results: 210 BC pts were offered GGT and 192 (91%) underwent testing [97% female, 95% non-Hispanic White, mean age at diagnosis of 62, 75% Stage I], with all but 7 pts completing GGT before primary treatment decisions. 104 (54%) of patients were in-criteria (IC) and 88 (46%) out-of-criteria (OOC) with NCCN GGT guidelines. PGVs were identified in 25 pts (13%) and 15 genes, most commonly BRCA1/2 (5 pts), PALB2 (3), monoallelic MUTYH (3) and ATM, BARD1, CHEK2 (2 pts each). PGV frequencies were 15% in IC and 11% in OOC patients (p=0.495). 46 (24%) pts had only a variant of uncertain significance (VUS) results. 70% of patients had &amp;gt; 1 clinical management change based on their GGT results. Changes in breast surgery, uptake of risk-reducing bilateral salpingo-oophorectomy, surveillance and clinical follow-up were significantly higher in patients with PGV compared to those with VUS/negative results (p &amp;lt; 0.0001). Compared to those with negative results, patients with PGVs in BRCA1 and BRCA2 were significantly more likely to undergo bilateral mastectomy (BLM) while those with PGVs in other genes or with VUS were not significantly more likely to elect for BLM than those with negative results. Clinicians reported that return of GGT reinforced their recommendation for 73/138 (53%) of patients who underwent breast-conserving therapy. GGT results aided in radiation therapy strategy for 54% of pts, most often resulting in dose de-escalation (accelerated fractionation) for negative/VUS pts (Table). Conclusions: Universal GGT for patients with BC was successfully implemented in a rural community practice with &amp;gt;90% uptake. Treatment was enhanced in those with clinically actionable PGVs and de-escalated in those without PGVs. Universal GGT for patients with BC is feasible to implement as the standard of practice within rural populations, enabling optimization of clinical care to patients’ genetic profile and may reduce unnecessary healthcare resource utilization. Universal Germline Genetic Testing for Breast Cancer: Implementation in a Rural Practice and Impact on Shared Decision Making Changes in Radiation Therapy Strategy based on GGT Citation Format: Charles Shelton, Sarah Nielsen, Antonio Ruiz, Lauren Shelton, Karen Freas, Sarah Poll, Brandie Heald, Daniel Pineda-Alvarez, Edward Esplin, Rachel Ellsworth, Hannah Montgomery. Universal Germline Genetic Testing for Breast Cancer: Implementation in a Rural Practice and Impact on Shared Decision Making [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-08-04.

Prediction of mechanical ventilation in Guillain-Barré syndrome at admission: Construction of a nomogram and comparison with the EGRIS model

BackgroundRespiratory failure requiring mechanical ventilation (MV) is a common and severe complication of Guillain-Barré syndrome (GBS) with a reported incidence ranging from 20 % to 30 %. Thus, we aim to develop a nomogram to evaluate the risk of MV in patients with GBS at admission and tailor individualized care and treatment. MethodsA total of 633 patients with GBS (434 in the training set, and 199 in the validation set) admitted to the First Hospital of Jilin University, Changchun, China from January 2010 to January 2021 were retrospectively enrolled. Subjects (n = 71) from the same institution from January 2021 to May 2022 were prospectively collected and allocated to the testing set. Multivariable logistic regression analysis was applied to build a predictive model incorporating the optimal features selected in the least absolute shrinkage and selection operator (LASSO) in the training set. The predictive model was validated using internal bootstrap resampling, an external validation set, and a prospective testing set, and the model's performance was assessed by using the concordance index (C-index), calibration curves, and decision curve analysis (DCA). Finally, we established a multivariable logistic model by using variables of the Erasmus GBS Respiratory Insufficiency Score (EGRIS) and did the same analysis to compare the performance of our predictive model with the EGRIS model. ResultsVariables in the final model selected by LASSO included time from onset to admission, facial and/or bulbar weakness, Medical Research Council sum score at admission, neutrophil-to-lymphocyte ratio, and platelet-lymphocyte ratio. The model presented as a nomogram displaying favorable discriminative ability with a C-index of 0.914 in the training set, 0.903 in the internal validation set, 0.953 in the external validation set, and 0.929 in the testing set. The model was well-calibrated and clinically useful as assessed by the calibration curve and DCA. As compared with the EGRIS model, our predictive model displayed satisfactory performance. ConclusionsWe constructed a nomogram for early prediction of the risk of MV in patients with GBS. This model had satisfactory performance and appeared more efficient than the EGRIS model in Chinese patients with GBS.

Rapid 2D 23Na MRI of the calf using a denoising convolutional neural network

Purpose23Na MRI can be used to quantify in-vivo tissue sodium concentration (TSC), but the inherently low 23Na signal leads to long scan times and/or noisy or low-resolution images. Reconstruction algorithms such as compressed sensing (CS) have been proposed to mitigate low signal-to-noise ratio (SNR); although, these can result in unnatural images, suboptimal denoising and long processing times. Recently, machine learning has been increasingly used to denoise 1H MRI acquisitions; however, this approach typically requires large volumes of high-quality training data, which is not readily available for 23Na MRI. Here, we propose using 1H data to train a denoising convolutional neural network (CNN), which we subsequently demonstrate on prospective 23Na images of the calf. Methods1893 1H fat-saturated transverse slices of the knee from the open-source fastMRI dataset were used to train denoising CNNs for different levels of noise. Synthetic low SNR images were generated by adding gaussian noise to the high-quality 1H k-space data before reconstruction to create paired training data. For prospective testing, 23Na images of the calf were acquired in 10 healthy volunteers with a total of 150 averages over ten minutes, which were used as a reference throughout the study. From this data, images with fewer averages were retrospectively reconstructed using a non-uniform fast Fourier transform (NUFFT) as well as CS, with the NUFFT images subsequently denoised using the trained CNN. ResultsCNNs were successfully applied to 23Na images reconstructed with 50, 40 and 30 averages. Muscle and skin apparent TSC quantification from CNN-denoised images were equivalent to those from CS images, with <0.9 mM bias compared to reference values. Estimated SNR was significantly higher in CNN-denoised images compared to NUFFT, CS and reference images. Quantitative edge sharpness was equivalent for all images. For subjective image quality ranking, CNN-denoised images ranked equally best with reference images and significantly better than NUFFT and CS images. ConclusionDenoising CNNs trained on 1H data can be successfully applied to 23Na images of the calf; thus, allowing scan time to be reduced from ten minutes to two minutes with little impact on image quality or apparent TSC quantification accuracy.