Browse Articles

HCLmNet: A unified hybrid continual learning strategy multimodal network for lung cancer survival prediction

Lung cancer survival prediction remains one of the most challenging tasks in modern healthcare, as accurate and adaptive prediction models are essential for improving patient outcomes. However, the continuous inflow of new patient data in hospital environments demands models that can update incrementally without losing prior knowledge a challenge known as catastrophic forgetting. This problem is compounded by the complexity of multimodal data integration, which combines heterogeneous sources such as CT and PET imaging, genomic (DNA) sequences, and clinical records. Traditional deep learning (DL) models, especially CNN-based systems, often fail to capture subtle patterns such as ground-glass opacities or multi-lesion tumors and cannot effectively adapt to new data streams. To overcome these challenges, this study proposes HCLmNet, a Hybrid Continual Learning (CL) Multimodal Network that integrates Elastic Weight Consolidation (EWC) with three complementary replay-based modules: Experience Replay (ER), Instance-Level Correlation Replay (EICR), and Class-Level Correlation Replay (ECCR). ER stabilizes learning through selective sample replay; EICR preserves fine-grained inter-instance relationships across modalities; and ECCR employs triplet-based contrastive learning to maintain class-level correlations. The architecture incorporates a Swin Transformer (SwinT) for extracting critical imaging features, XLNet for modeling DNA patterns, and a Fully Connected Network (FCN) for processing temporal clinical data. A cross-attention fusion layer integrates these modalities, while an FCN and Cox Proportional Hazards (CoxPH) model produce final 5-year survival predictions. Experimental results on multimodal lung cancer datasets show that traditional models such as CoxPH and DeepSurv achieved Concordance Index (C-index) scores of 0.65 and 0.70, respectively. The base multimodal model without CL achieves a C-index of 0.76 and a Mean Absolute Error (MAE) of 189 days. In contrast, the proposed HCLmNet, equipped with CL mechanisms, reaches a C-index of 0.84, representing a 7.7% improvement over the best baseline. Furthermore, the model reduces the MAE from 252 and 189 days to 140 days and minimizes catastrophic forgetting to 0.08. These improvements stem from the synergistic integration of the ER, EICR, and ECCR CL modules, which enable the model to retain prior knowledge while effectively adapting to new data. Overall, HCLmNet demonstrates superior stability, adaptability, and interpretability for lung cancer survival prediction in dynamic clinical environments.

Abstract TH868: Plasma Lipidome, Apolipoprotein-Defined High-Density Lipoprotein Subspecies, and Their Associations with Coronary Heart Disease Risk

Introduction: HDL subspecies defined by functional apolipoproteins (apo) exhibit different associations with coronary heart disease (CHD) risk; however, the underlying mechanism, particularly from a lipidomic perspective, is unclear. Hypothesis: Apo-defined HDL subspecies may be characterized by distinct lipid profiles that are associated with differential CHD risk. Methods: We measured the apoA1 concentrations of 15 apo-defined HDL subspecies and 176 plasma lipids among 446 women in Nurses’ Health Study (NHS). The proportion of each HDL subspecies was calculated as the ratio of apoA1 concentrations in HDL containing a specific protein to total plasma apoA1 concentration. Linear regression was used to identify individual lipids associated with the proportions of HDL subspecies, and elastic net regression was used to define predictive lipidomic signatures. Associations between lipidomic signatures and CHD risk were examined using conditional logistic regression in a nested case-control study (400 matched pairs) and Cox regression in 13,495 participants from NHS, NHSII, and Health Professionals Follow-Up Study. Results: HDL subspecies showed different associations with lipidomic features: proportions of HDL containing apoC1, apoE, apoL1, apoC3, and apoJ were associated with specific lipids, while few associations were found for the other ten subspecies. The lipid profile of HDL containing apoC1 closely resembled that of total apoA1, whereas HDL containing apoL1 showed a distinct pattern. Several TAGs and DAGs were negatively associated with apoC1 but positively with apoL1; conversely, certain phospholipids showed positive associations with apoC1 and negative associations with apoL1. The Pearson r between the lipidomic signatures and their corresponding HDL subspecies was 0.76 for total apoA1 (total HDL), 0.67 for apoC1, 0.63 for apoE, 0.55 for apoC3, 0.50 for apoL1, and 0.34 for apoJ. Multivariable-adjusted ORs (95% CIs) for CHD risk per 1-SD increase in lipidomic signatures were 0.85 (0.72, 1.00) for total apoA1, 0.77 (0.65, 0.91) for apoC1, 0.79 (0.66, 0.94) for apoE, 0.92 (0.79, 1.08) for apoC3, 1.30 (1.10, 1.53) for apoL1, and 0.97 (0.81, 1.15) for apoJ ( Figure ). Individual lipids positively associated with HDL containing apoL1, including specific TAGs, DAGs, and ceramides, were also related to a higher CHD risk in the cohorts. Conclusions: Lipid profiles vary across apo-defined HDL subspecies and may contribute to their different associations with CHD risk.

Abstract TH976: Premenopausal Depression and Risk of Incident Cardiovascular Disease in Postmenopausal Women: A TriNetX-Based Cohort Study

Objective: This study aimed to determine whether depression diagnosed prior to menopause independently predicts incident cardiovascular disease (CVD) after menopause, addressing an important gap in understanding the long-term cardiovascular consequences of midlife mental health. Methods: A retrospective cohort study was conducted using the TriNetX Global Collaborative Network, which includes data from 155 large healthcare organizations. Two cohorts of postmenopausal women were identified: cohort 1 was comprised of women with depression diagnosed before menopause and without preexisting CVD, premature menopause, or conventional CVD risk factors (smoking, diabetes mellitus, essential hypertension, dyslipidemia, overweight/obesity, or family history of ischemic heart disease); cohort 2 included comparable women without depression prior to menopause. Propensity score matching was performed to balance baseline characteristics, including age at menopause, current age, and ethnicity. Risk differences and risk ratios (RRs) with 95% confidence intervals (CIs) were calculated to compare CVD incidence (composite of myocardial infarction, coronary heart disease, stroke, heart failure, and cardiovascular death) between cohorts. Kaplan–Meier survival analysis with log-rank testing assessed time to incident CVD, and Cox proportional hazards regression estimated hazard ratios (HRs). Results: Initial queries identified 90,638 women in cohort 1 and 1,237,202 in cohort 2. After propensity score matching, 88,884 women were included in each cohort, with a mean (SD) age of 57.6 (11.4) years. Participants were observed for a maximum of 20 years after the onset of menopause, with a mean (median) follow-up 3.37 (2.23) years in cohort 1 and 3.31 (2.05) years in cohort 2. Compared with cohort 2, women with premenopausal depression had a significantly higher incidence of CVD after menopause (4.0% vs. 3.3%; RR = 1.23, 95% CI 1.17–1.29, p < .001). Premenopausal depression was associated with a significantly increased hazard of incident CVD over time (HR = 1.21, 95% CI 1.15–1.27, p < .001). Conclusion: Among women without preexisting CVD or traditional cardiovascular risk factors before menopause, depression diagnosed prior to menopause was associated with a higher risk of developing CVD later in life. These findings suggest that premenopausal depression may serve as an independent and early indicator of cardiovascular risk in midlife women.

Correction for Chen et al., RNA-binding activity of PHGDH drives amyloid-beta production in a human brain organoid model of sporadic Alzheimer’s disease

Correction: Oral health in patients scheduled for hematopoietic stem cell transplantation in the Orastem study

Abstract 38: Objectively-Measured Individual-Level Nighttime Light Exposure Was Associated With Increased Cardiovascular Disease Risk and Adverse Cardiovascular Imaging Phenotypes

Introduction: Circadian rhythm disruption has been implicated as a risk factor for cardiovascular disease (CVD). Yet, whether light, the primary circadian entrainer, affects CVD risk remains unclear. Prior studies relied primarily on satellite-derived estimates or self-reports that poorly approximate individual-level light exposure. Moreover, no study has evaluated whether light influences cardiovascular structural and functional phenotypes. Methods: We conducted a prospective cohort study of 73,286 adults in the UK Biobank who wore a wrist accelerometer with a light sensor for 7 days between 2013-2015 (baseline). Daytime light exposure >1000 lux and nighttime light exposure >3 lux were estimated. Incident myocardial infarction, stroke, heart failure, and atrial fibrillation was ascertained through hospital admissions and death registries to 2022, and CVD mortality through 2024. Participants also underwent cardiac magnetic resonance (CMR) imaging (n=11,071) and carotid ultrasound (n=13,937) during follow-up. Cox proportional hazards and linear regression models were used to estimate associations of light exposures with incident CVD and subclinical measures, respectively. Results: After adjusting for anthropometric, sociodemographic, lifestyle, environmental, and health-related factors, greater nighttime light exposure >3 lux (high vs none) was associated with a higher risk of myocardial infarction (hazard ratio [HR], 1.24 [95% CI, 1.07-1.44]), stroke (HR, 1.33 [95% CI, 1.11-1.59]), heart failure (HR, 1.29 [95% CI, 1.12-1.49]), atrial fibrillation (HR, 1.15 [95% CI, 1.04-1.27]), and CVD mortality (HR, 1.26 [95% CI, 1.05-1.52]) ( Fig.1 ). These associations were partly mediated by behavioral markers of circadian rhythms, including lower rest-activity rhythm relative amplitude and shorter sleep duration ( Fig.2 ). Consistently, greater nighttime light exposure was also associated with higher mean carotid intima-media thickness and adverse CMR-derived phenotypes, including larger left and right ventricular end-diastolic and end-systolic volumes, greater left ventricular mass, larger left atrial maximum volume, and lower left atrial ejection fraction (FDR-adjusted P <0.05) ( Fig.3 ). Daytime light exposure was not associated with CVD risk or subclinical markers. Conclusions: Maintaining dark nights may offer a novel and practical avenue to promote cardiovascular health at the population level.

Abstract WE417: Inflammation and Long-Term Progression of Cardiovascular-Kidney-Metabolic Syndrome Among Young Adults: Insights from the CARDIA Study

Background: Chronic, low-grade inflammation is increasingly recognized as an important mechanism underlying the development of cardiovascular, kidney, and metabolic (CKM) conditions. However, whether it is associated with CKM syndrome progression is uncertain. Methods: In the prospective Coronary Artery Risk Development in Young Adults (CARDIA) study, which recruited young adults (aged 18-30 years) between 1985-1986. For this study, baseline was defined at a follow-up visit in 1992 when high-sensitivity C-reactive protein (hsCRP) was measured. CKM syndrome stages were characterized according to adapted American Heart Association criteria at follow-up visits 3- and 28-years after hsCRP measurement. The association between hsCRP and progression of CKM syndrome (transition to ≥1 higher stages) over 25 years was examined. Results: Among 2,952 participants with available hsCRP and without established cardiovascular disease (mean age, 35±4 years), 1,005 (34%) had a hsCRP level ≥2 mg/dL and 740 (25%) had a hsCRP level ≥3 mg/dL. Among those alive at follow-up, 76% experienced any CKM syndrome progression and 40% experienced any regression. Participants with higher hsCRP levels had more advanced CKM syndrome stages at baseline (41% vs. 25% with CKM stages ≥2 if hsCRP ≥2 vs. <2 mg/dL, respectively) and at follow-up ( Figure, A ). After covariate adjustment, higher baseline hsCRP was incrementally associated with a higher rate of CKM syndrome progression (aHR per doubling of hsCRP, 1.04; 95% CI, 1.02-1.07; P =0.002). Baseline CKM syndrome stage (0 vs. ≥1) did not appear to modify the association between higher hsCRP and CKM syndrome progression ( P interaction =0.27) ( Figure, B ). Similar findings were observed when hsCRP was dichotomized at ≥3 vs. <3 mg/dL (aHR, 1.11, 95% CI, 1.00-1.24; P =0.05), but not at ≥2 vs. <2 mg/dL (aHR, 1.06, 95% CI, 0.96-1.16; P =0.28). Conclusions: Elevated hsCRP levels were common among young adults and modestly associated with CKM syndrome onset and progression over 25 years. These findings suggest low-grade inflammation may incrementally portend worsening CKM health over time.

Dislodging the drift barrier: Why do mutation rates vary?

Digital transformation, dynamic capabilities and new quality productive forces: Empirical data from listed Chinese manufacturing companies

In the context of the rapid growth of the digital economy, it is of great significance to explore how digital transformation can promote the development of new quality productive forces (Nqpf). Drawing on data from A share manufacturing listed companies in Shanghai and Shenzhen between 2011 and 2022, this study empirically analyzes the effect of digital transformation in the manufacturing industry on emerging Nqpf. The results reveal that digital transformation in the manufacturing industry has a significant positive impact on enhancing the level of Nqpf. Mechanism tests show that digital transformation in the manufacturing industry can promote the growth of Nqpf by strengthening companies’ innovation, absorptive, and adaptive capabilities. Heterogeneity tests further indicate that the influence of digital transformation on advancing Nqpf is particularly pronounced in Non-state-owned, High-tech, and Growth stage manufacturing companies. The research results uncover the underlying mechanisms through which digital transformation in the manufacturing industry influences Nqpf, and offering valuable theoretical and practical insights for manufacturing companies on how to leverage digital transformation to promote the development of Nqpf.

Abstract TH890: Urinary Metabolomics Signature for Cognitive Function among African American Patients with Chronic Kidney Disease

Introduction: Patients with chronic kidney disease (CKD) have a substantially higher risk of cognitive impairment. Metabolic dysregulation and accumulation of uremic toxins in CKD may contribute to neurocognitive decline through oxidative stress, inflammation, and blood–brain barrier (BBB) dysfunction. Because the kidney and BBB share solute transport mechanisms, urine metabolomics provides a noninvasive approach to capture systemic metabolic alterations relevant to brain health. However, urinary metabolomic correlates of cognitive function in CKD remain largely unexplored. This study examined associations between urine metabolites and cognitive function measures among African American patients with CKD. Methods: We analyzed data from 1,681 African American participants in the Chronic Renal Insufficiency Cohort (CRIC). Baseline metabolites were quantified from 24-hour urine samples using the untargeted Metabolon platform. Cognitive function was assessed using the Mini-Mental State Examination (MMSE) at baseline and annually over a median follow-up of 9.1 years. Multivariable linear mixed-effects models were used to estimate associations of each metabolite with repeatedly measured cognitive scores, adjusting for age, sex, education, study site in the base model and additionally for hypertension, diabetes, depression, LDL cholesterol, BMI, ACE inhibitors or ARB use, smoking, and alcohol use in the full model. False discovery rate (FDR) correction was applied. To evaluate the predictive value of the metabolite profile, we developed a composite metabolite score using LASSO regression with 10-fold cross-validation to select the optimal penalty parameter. We then compared two regression models: Model 1 included all covariates from the full model, and Model 2 additionally included the metabolite score. Results: Six metabolites were significantly associated with MMSE scores in the base model ( Table ). However, these associations were attenuated and not significant in the full model. Nineteen metabolites were selected by LASSO to construct the composite metabolite score. The score was strongly associated with MMSE decline (β = 0.031 per SD, P < 0.001) and modestly but significantly improved model fit, explaining an additional 3.9% of variance over the fully adjusted model (P < 0.001). Conclusions: A composite metabolite score integrating 19 urinary metabolites modestly but significantly improved the prediction of cognitive decline beyond conventional risk factors.

Abstract TU123: Indicators of Cardiovascular Risk Factors in Hispanic Children

Background: Childhood represents a critical window for the development of cardiovascular health. Early exposure to psychological stress and adverse experiences can alter autonomic, metabolic, and vascular regulation, increasing lifetime cardiovascular disease (CVD) risk. Hispanic children remain understudied despite disproportionate exposure to adversity. Identifying the early indicators of CVD risk factors and psychosocial factors in youth is essential to guide targeted prevention strategies in high-risk populations. Objective: To characterize the prevalence of pediatric CVD risk factors and psychosocial indicators in Puerto Rican children. Methods: In this ongoing cross-sectional study, children are recruited from the Puerto Rico Health Justice Center and pediatric community clinics. They complete two visits that include PROMIS Pediatric measures (stress, sleep, family relations), anthropometrics, blood pressure, fasting lipids/glucose, and blood collection for biomarkers. The current analysis includes 38 participants (mean age 12.1 ± 2.3 y; 73% female). Results: More than half (68.4%) of participants presented one or more CVD risk factors. Prevalence by category was: overweight/obesity = 23.7%, hypertension = 21.6%, high cholesterol = 21.6%, low HDL = 21.6%, and high fasting glucose = 10.8%. Risk-factor clustering (two or more) was observed in 8/38 (21.0%), with 18/38 (47.4%) having one factor, and 12/38 (31.6%) none. Psychosocial indicators revealed that 34.2% reported elevated stress, 23.7% poor sleep, and 47.4% below-average physical activity, reflecting widespread lifestyle and emotional vulnerabilities in this population. Conclusions: Nearly one in four Hispanic children already exhibits early indicators of multiple cardiovascular risk factors, and more than two-thirds have at least one. These findings underscore the importance of early, culturally sensitive screening and intervention programs in pediatric care and community settings to mitigate long-term cardiovascular risk.

Biosphere expansion drives Earth’s secular oxygenation while tectonics modulate oxygen variability revealed by machine learning

The rise of atmospheric oxygen fundamentally transformed Earth’s surface environment and enabled the evolution of complex life. However, the processes driving long-term oxygen fluctuations remain poorly resolved, partly from limited proxy resolution and temporal coverage. Trace element (TE) concentrations in sedimentary pyrite offer a robust archive of redox conditions in ancient oceans and their linkage to atmospheric oxygen levels. Here we integrate high-resolution geochemical data from pyrite grains spanning 3.5 billion years with machine learning to reconstruct atmospheric oxygen evolution. We identify two coherent TE groups representing redox-sensitive and hydrothermal influences. Our results reveal that the long-term, secular trend of atmospheric oxygen is tightly coupled with biosphere expansion, whereas superimposed short-term fluctuations are influenced by tectonic events, including supercontinent assembly and breakup. Specifically, we show that primary oxygenation events (GOE and NOE) correlate strongly with biological expansion. Episodes of prolonged oxygenation broadly overlap with continental assembly, reflecting enhanced weathering, nutrient fluxes, and organic carbon burial, whereas supercontinent breakup phases are commonly associated with more reducing conditions, likely linked to increased volcanic emissions and diminished net biospheric oxygen. This reconstruction not only refines the temporal dynamics of Earth’s redox evolution but also highlights the interconnected roles of biological productivity, tectonics, ocean chemistry, and Earth-system processes in shaping planetary habitability. These findings provide a comprehensive framework for understanding Earth’s atmospheric evolution and inform models of environmental change on early Earth and other habitable planets.

Postgraduate students’ perceptions of artificial intelligence integration in research: A cross-sectional study

Background Generative artificial intelligence (AI) tools such as ChatGPT are increasingly used in academic research, yet evidence on postgraduate students’ perceptions remains limited in non-Western and health-professional contexts. Understanding how students perceive AI’s benefits, risks, and ethical implications is essential for informing institutional research policies. Methods This cross-sectional case study surveyed 267 master’s students enrolled in nursing and health profession programs at Northern Border University in Arar, Saudi Arabia. Data were collected between October 1 and November 15, 2025, using a validated 54-item questionnaire that assessed perceived benefits, perceived risks, privacy concerns, mistrust in AI, performance anxiety, social bias, regulatory matters, liability issues, and intention to adopt AI tools. Multiple linear regression with heteroscedasticity-robust (HC3) standard errors was used to identify predictors of AI adoption intention. Results Most participants (85.0%) reported prior use of AI tools, predominantly ChatGPT. Perceived benefits were the strongest predictor of intention to adopt AI for research purposes (β = 0.588, p < 0.001). Privacy concerns were positively associated with adoption intention (β = 0.230, p < 0.001), suggesting informed and critical engagement rather than resistance. Female students reported higher adoption intention than males (β = 0.137, p = 0.002), while prior publication experience was negatively associated with intention (β = −0.089, p = 0.036). Demographic variables such as age, specialty, and marital status were not significant predictors. The adoption-intention model demonstrated moderate explanatory power (adjusted R 2  = 0.560). Conclusions Among nursing and health profession master’s students at a regional Saudi university, findings indicate pragmatic optimism toward AI integration in academic research, driven primarily by perceived benefits alongside heightened ethical and privacy awareness. Privacy concerns appear to reflect critical literacy rather than barriers to adoption.

Abstract TH903: Health Literacy and Lifestyles in the Italian Adult Population Aged Under and Over 65 Years: a National Survey

Introduction: Health literacy (HL) refers to people’s ability to find, understand, judge, and use health information to take health-related decisions. Hypothesis: This study aims to evaluate the association between HL and lifestyles in the Italian adult population, comparing under (U65) and over (O65) 65 years age groups. Methods: In 2021, within the framework of the WHO network Measuring Population and Organizational Health Literacy (M-POHL), the Italian National Institute of Health conducted a national survey on a representative sample of the general population aged 18+ years (n=3,500; 2,548 U65, 952 O65). A validated 47-item questionnaire was used to collect core HL information, along with 31 items on related factors, including lifestyles. HL items were rated on a 4-point Likert scale (very easy/easy/difficult/very difficult). Percentage of very easy/easy responses was categorized as Inadequate (0–50%), Problematic (51–66%), Sufficient (67–80%), Excellent (81–100%). Inadequate/Problematic groups were merged into ‘Limiting’, Sufficient/Excellent into ‘Not limiting’. Multivariate logistic regression models were elaborated in U65 and O65 to assess the association of dichotomous HL with obesity (BMI≥30 kg/m2), physical activity-PA (intense (5–7 days/week)/not intense (0–4 d/w)), fruit and vegetable diet-FVD (daily/not daily), smoking habit (yes/no), adjusted by sex, age, region, education-ED, and financial deprivation-FD. Results: Prevalence of Limiting HL was significantly higher in O65 than in U65 (68.1% vs 63.3%). In individuals with Limiting HL, smoking, not daily FVD, not intense PA were significantly more frequent in U65 compared to O65 (34.4% vs 20.3%, 61.4% vs 36.5%, 77.1% vs 67.7%, respectively). Low ED was significantly higher in O65 than in U65 (65.9% vs 3.3%). Obesity and severe FD were similar in O65 and U65 (15.1% vs 14.3% and 20.2% vs 21.0%). Not intense PA and not daily FVD were more strongly associated with Limiting HL in U65 (OR=1.45 and 1.42) than in O65 (OR=1.25 and 0.92). Severe FD remained a strong determinant in both groups (OR=3.38 U65 and 3.45 O65), Low ED was associated with Limiting HL only in O65 (OR=1.38). Conclusions: Unhealthy lifestyles (not intense PA, not daily FVD) showed stronger associations with Limiting HL among U65, whereas socioeconomic determinants were more relevant in O65. These findings suggest different vulnerability profiles by age and socio-economic conditions, calling for targeted strategies to strengthen HL.

Abstract TU259: Increased Mean Arterial Pressure Was Prospectively Associated With a Higher 45-year Mortality Risk From Cardiovascular Disease In Men: The NHLBI Twin Study

Saam Honarvar, D.O. ’28, MPH 1 ; Craig Clark, D.O. 2 ; Jun Dai, M.D., M.Sc., Ph.D. 3 Affiliations: 1 College of Osteopathic Medicine, Des Moines University, West Des Moines, Iowa 2 Master of Science in Physician Assistant, College of Health Sciences, Des Moines University, West Des Moines, Iowa 3 Department of Public Health, College of Health Sciences, Des Moines University, West Des Moines, Iowa Background: Mean arterial pressure (MAP) is a cardiovascular risk factor and better reflects vascular load and perfusion than systolic (SBP) or diastolic blood pressure (DBP) alone. However, it is unclear if MAP is prospectively associated with cardiovascular death risk independent of genetic and common environmental factors. Method: A total of 639 white male twins [129 monozygotic and 116 dizygotic pairs, and 64 monozygotic and 85 dizygotic unpaired twins] aged 42–55 years at baseline (1969–1973), free of cardiovascular disease and hypertension, were followed up until December 31, 2010, or 2014 in the National Heart, Lung, and Blood Institute (NHLBI) Twin Study. The baseline MAP was calculated from baseline SBP and DBP. Frailty survival models were used to evaluate the association of MAP with cardiovascular death risk and to estimate hazard ratios (HRs) per 10 mm Hg increment in MAP. Results: Baseline MAP ranged from 70 to 115 (median: 93.5) mm Hg. After adjustment for caloric intake, alcohol intake, education, body mass index, marital status, and modified Framingham risk score, the overall association was statistically significant [HR 1.09 (95% CI: 1.01–1.17; p =0.03) for follow-up through both 2010 and 2014]. For within-pair associations, the fully adjusted HR was 1.17 (95% CI: 1.02–1.34; p =0.03) through 2010 and 1.16 (95% CI: 1.02–1.33; p =0.03) through 2014. Conclusion: Mean arterial pressure is positively associated with cardiovascular death risk, with a similar magnitude, during both 41 and 45 years of follow-up, independent of known cardiovascular risk factors, genetic, and shared environmental factors. Funding Support: This study was funded by the U.S. National Institutes of Health (grant number HL51429 to the NHLBI Twin Study) and the Mentored Student Research Program at Des Moines University.

Reply to Vankov et al.: Reasoning traces are linked to accuracy and capture key dimensions of problem complexity

How do we tread? Differences in stability-related foot placement control between overground and treadmill walking in young adults

Step-by-step foot placement control, accommodating for natural variations in center-of-mass state, ensures stability during steady-state gait. Current understanding of this foot placement mechanisms is primarily based on findings during treadmill walking. However, contextual differences might hamper generalization of these findings towards overground walking, and ultimately daily life gait. This study investigated whether foot placement control manifests itself differently during overground as compared to treadmill walking in healthy young adults. 14 young adults walked at comfortable walking speed, both on the treadmill and overground in a figure-8 path. During overground walking we found a significant relationship between the step width/step length and the center-of-mass state during the swing phase of walking, capturing foot placement control with the same linear model as during treadmill walking. Contrary to what was hypothesized, we found a significant lower foot placement precision during overground walking for the step width model with center-of-mass state at the start of the swing phase as predictor, complemented by a wider average step width. Moreover, during overground walking participants responded less strongly to a deviation in center-of-mass position, yet significantly stronger to deviations in center-of-mass velocity at the end of the step for both the step width/step length models. Exploratory analysis showed a larger relative contribution of velocity feedback during overground walking as compared to treadmill walking. These differences warrant caution in generalizing foot placement findings during treadmill walking to overground walking and might be promising for the estimation of foot placement control in daily life gait.

Abstract TH853: Performance of Machine Learning-based Weighted Life’s Essential 8 Scores in Predicting Cardiovascular and All-Cause Mortality

Background: The American Heart Association's Life's Essential 8 (LE8) metric uses equal weighting across eight components to predict CVD risk and mortality. However, there is evidence suggesting variability in the associations between individual components and mortality. Objective: Compare the performance of machine learning-based, data-driven component weighted LE8 scores to an equal-weighted LE8 score in predicting cardiovascular and all-cause mortality. Methods: We analyzed 55,879 adults aged ≥18 years from NHANES 1999-2018 survey waves with mortality linkage through 2019 (mean follow-up 9.8 years). Using multiple imputation (m=10) with predictive mean matching, we performed a machine learning-based adaptive lasso variable selection across imputed datasets to identify LE8 components predictive of all-cause and CVD mortality. Component weights were derived from selection frequencies: components selected in 100% of imputations received proportionately higher weights, while those selected less frequently received lower weights. We created outcome-specific weighted LE8 scores and compared the discriminative performance against the equal-weighted LE8 using Cox proportional hazards models adjusted for age, sex, and race/ethnicity. Performance was evaluated using C-statistics and hazard ratios. Results: For all-cause mortality, seven components were consistently selected (100% selection: diet quality, nicotine exposure, sleep health, body mass index, blood lipids and glucose, blood pressure); physical activity was never selected. For CVD mortality, five components showed 100% selection, sleep health 70%, nicotine exposure 30%, and physical activity 0% ( Figure ). The weighted LE8 scores (five and seven components, respectively) and the equal-weighted LE8 score performed identically for all-cause mortality (C-statistic weighted 0.856 vs equal-weighted 0.856, 95% CI: 0.852-0.860) and nearly identically for CVD mortality (C-statistic 0.884 vs 0.885; Table 1 ). Hazard ratios per 10-point increases in LE8 score demonstrated similar protective effects (all-cause mortality: HR 0.877 vs 0.872; CVD mortality: HR 0.869 vs 0.827; Table 2 ). Risk reclassification occurred in 15.1% (all-cause) and 17.9% (CVD) of participants without improving discrimination. Conclusion: A subset of five and seven machine learning–selected LE8 components predicted CVD and all-cause mortality as well as the full equal-weighted LE8 score.

Abstract TU240: Depression and 10-Year Cardiovascular Disease Risk in a Nationally Representative Sample using AHA PREVENT Equation

Background: Cardiovascular disease (CVD) remains the leading cause of death in the United States. In 2023, the American Heart Association (AHA) introduced the PREVENT equation to precisely estimate 10-year risk for heart failure (HF), atherosclerotic CVD (ASCVD), and total CVD across diverse populations. PREVENT equation provides more accurate risk estimation than the previously used equations. Depression, a well-recognized yet under-quantified determinant of cardiovascular health, may influence estimated risk. We aim to examine the association between depression severity and predicted 10-year cardiovascular risk using the AHA PREVENT equations. Hypothesis: We hypothesized that higher depression severity would be associated with greater predicted 10-year cardiovascular risk, demonstrating a graded, dose-response relationship across depression categories. Methods: We analyzed data from the 2017-2018 National Health and Nutrition Examination Survey. Adults aged 30-70 years without self-reported or clinically diagnosed CVD who completed the Patient Health Questionnaire (PHQ-9) and had complete data for calculating PREVENT-based 10-year risk estimates were included (n ≈ 700). Per the PHQ-9, depression severity was classified as minimal/none (0–4), mild (5–9), moderate (10–14), moderately severe (15–19), and severe (20–27). Predicted 10-year risk for HF, ASCVD, and total CVD was calculated using the PREVENT equations. Non-parametric Kruskal-Wallis tests, followed by Dunn pairwise comparisons, were conducted to asses group differences. Results: Depression severity was significantly associated with predicted 10-year risk for HF (p < 0.001), ASCVD (p < 0.001), and total CVD (p < 0.001). Predicted risk increased progressively across depression categories for all outcomes, consistent with a dose-response pattern. Individuals with higher depression scores demonstrated the greatest elevation in predicted HF risk, followed by total CVD and ASCVD. Individuals with mild to severe depressive symptoms had significantly higher estimated 10-year risk than those with minimal or no symptoms (all p < 0.05). Stepwise differences were observed between moderate, moderately severe, and severe depression categories, suggesting a graded association between symptom burden and cardiovascular risk. Conclusions: These findings highlight the importance of incorporating mental health assessment into cardiovascular risk stratification and prevention efforts.

Pharmaco-behavioral profiling identifies suppressors of autism gene–associated phenotypes in zebrafish

Pharmaco-behavioral screens in scalable in vivo systems have critical advantages for drug discovery relevant to large-effect autism spectrum disorder (ASD) genes. Here, we establish a database and open-source website of the behavioral signatures of 520 US Food and Drug Administration (FDA)-approved drugs using high-throughput assays of basic sensory processing and arousal behaviors in larval zebrafish. By leveraging the behavioral profiles of 9 large-effect ASD gene mutants, we identify enrichment of pharmacological mechanisms that anticorrelate with subgroups of ASD genes with shared behavioral phenotypes. Screening of anticorrelating drugs in mutants of two ASD genes, SCN2A and DYRK1A , uncovers compounds that suppress mutant behavioral phenotypes. We identify estropipate, an estrogen receptor agonist, and paclitaxel, a microtubule inhibitor, as the top suppressors in scn1lab and dyrk1a mutants, respectively, and levocarnitine (LEVO), a mitochondrial modulator and carnitine supplement, as a top suppressor of both mutant behavioral phenotypes. Finally, we find that LEVO rescues regional brain activity deficits and dysregulated lipid metabolic pathways in mutants, as well as signaling deficits in human pluripotent stem cell–derived glutamatergic neurons carrying mutations in SCN2A and DYRK1A , demonstrating conservation of drug rescue across systems. Therefore, our study establishes a pharmaco-behavioral resource for precision medicine-based drug discovery, illuminating targets relevant to large-effect ASD genes.