Bruce A. Barton

Motivation: Mediation analysis plays a crucial role in understanding how exposure variables influence health outcomes via intermediate variables, or mediators, in environmental studies. When analyzing a large number of environmental exposures, such as chemical mixtures or pollutants, together with multiple potential mediators such as metabolites, advanced methodologies are necessary to accurately separate direct and indirect effects. This paper proposes a novel mediation analysis method based on Sparse Canonical Correlation Analysis (SCCA), designed specifically for settings where both exposures and mediators are high-dimensional. The effectiveness of the proposed method is evaluated through simulation studies and an application to real-world data. Results: The proposed SCCA-based mediation framework improved identification of relevant mediators and pathways in simulation studies, particularly in high-dimensional and noisy settings. The two-step screening extension further enhanced feature selection while maintaining stable estimation. In the real-data application, the method identified interpretable exposure–metabolite pathways associated with MELD score, with several pathways showing moderate selection stability and robustness to potential unmeasured confounding. Availability: The R code for implementing the proposed method and the simulation studies is available at https://github.com/MaggieLi2001/HDM-SCCA2. Supplementary information: Supplementary data are available at Bioinformatics online.

Background: The Dietary Intervention Study in Children (DISC) showed that a cholesterol-lowering dietary intervention can improve pediatric cardiovascular risk factors. Building on advances in longitudinal modeling, we re-analyzed DISC data to (1) reassess intervention effects using linear and spline mixed-effects models, (2) evaluate treatment effects within key subgroups, and (3) examine the impact of compliance on percent calories from fats. Methods: We conducted a secondary analysis of the randomized DISC trial, comparing a dietary intervention group with a usual care group. Linear mixed-effects and B-spline mixed-effects models characterized trajectories of LDL-cholesterol (LDL-C) and nutrient intake over time. We evaluated subgroup differences by sex and parental education and assessed compliance using session attendance and adherence to dietary goals for saturated fat, cholesterol, and total fat. Results: Compared with usual care, the intervention group showed a significant initial reduction in LDL-C and sustained decreases in total and saturated fat intake. B-spline models identified non-linear trends in LDL-C and dietary fat over follow-up. Higher attendance at intervention sessions was associated with reduced fat intake, with this effect less pronounced among girls. The intervention group also showed higher adherence to dietary goals, and higher parental education was related to better attendance. Conclusion: This reanalysis indicates that a family-based dietary intervention initiated in childhood can produce meaningful reductions in LDL-C and sustained improvements in dietary fat intake. The findings highlight the importance of adherence, parental involvement, and targeted support for specific subgroups in promoting long-term cardiovascular health in pediatric populations. Trial registration: NCT00000459.

Background: Increased circulating pro-inflammatory cytokine levels correlate with mortality in severe alcohol-associated hepatitis (AH), but their kinetics during disease progression or response to treatment remain unexplored. Objective: We longitudinally assessed circulating biomarkers in a severe AH cohort enrolled in a multicenter, double-blind clinical trial. Design: Eighty-nine patients with severe AH (MELD≥20) from 4 US sites were randomly assigned to treatment with IL-1receptor antagonist (IL-1Ra) + pentoxifylline + zinc (anakinra: 47) or methylprednisolone + placebo (steroid: 42). Plasma levels of 43 indicators of AH pathology (inflammation, bacterial translocation, liver regeneration, tissue remodeling, and cell-activation) were assessed on days 0, 7, 28, 90 and 180 after enrollment, and in 27 healthy controls. Results: Baseline characteristics and cytokine levels were similar between treatment groups but significantly different compared to healthy controls. Consistent with anakinra administration, day 7 IL-1Ra levels were significantly increased in the anakinra group accompanied by elevated levels of several other cytokines. IL-1β levels were increased in the steroid group on day 28. Strong correlations were observed between IL-1α and IL-17A as well as IL-1β and IL-13 both prior to and during treatment. Although the dysregulated biomarkers demonstrated improving trends in survivors, most did not normalize by day 180. Markers associated with 90-day mortality were distinct between the treatment groups with few exceptions (IL-13—absolute level; Sonic hedgehog and sTNFR-1—level change). Conclusions: Circulating cytokine and immune biomarkers dynamically change in AH during treatment, disease progression and/or resolution. Our results highlight the importance of treatment-specific biomarkers in future clinical trials. Export

In the early stages of atrial fibrillation (AF), most cases are paroxysmal (pAF), making identification only possible with continuous and prolonged monitoring. With the advent of wearables, smartwatches equipped with photoplethysmographic (PPG) sensors are an ideal approach for continuous monitoring of pAF. There have been numerous studies demonstrating successful capture of pAF events, especially using deep learning. However, deep learning requires a large amount of data and independent testing on diverse datasets, to ensure the generalizability of the model, and most prior studies did not meet these requirements. Moreover, most prior studies using wearable-based PPG sensor data collection were limited either to controlled environments, to minimize motion artifacts, or to short duration data collection. Most importantly, frequent premature atrial and ventricular contractions (PAC/PVC) can confound most AF detection algorithms. This has not been well studied, largely due to limited datasets containing these rhythms. Note that the recent deep learning models show 97% AF detection accuracy, and the sensitivity of the current state-of-the-art technique for PAC/PVC detection is only 75% on minimally motion artifact corrupted PPG data. Our study aims to address the above limitations using a recently completed NIH-funded Pulsewatch clinical trial which collected smartwatch PPG data over two weeks from 106 subjects. For our approach, we used multi-modal data which included 1D PPG, accelerometer, and heart rate data. We used a computationally efficient 1D bi-directional Gated Recurrent Unit (1D-Bi-GRU) deep learning model to detect three classes: normal sinus rhythm, AF, and PAC/PVC. Our proposed 1D-Bi-GRU model’s performance was compared with two other deep learning models that have reported some of the highest performance metrics, in prior work. For three-arrhythmia-classification, testing data for all deep learning models consisted of using independent data and subjects from the training data, and further evaluations were performed using two independent datasets that were not part of the training dataset. Our multimodal model achieved an unprecedented 83% sensitivity for PAC/PVC detection while maintaining a high accuracy of 97.31% for AF detection. Our model was computationally more efficient (14 times more efficient and 2.7 times faster) and outperformed the best state-of-the-art model by 20.81% for PAC/PVC sensitivity and 2.55% for AF accuracy. We also tested our models on two independent PPG datasets collected with a different smartwatch and a fingertip PPG sensor. Our three-arrhythmia-classification results show high macro-averaged area under the receiver operating characteristic curve values of 96.22%, and 94.17% for two independent datasets, demonstrating better generalizability of the proposed model.

Background: The US Preventive Services Task Force recommends providers offer individualized healthy behavior interventions for all adults, independent of their risk of cardiovascular disease. While strong evidence exists to support disease-specific programs designed to improve multiple lifestyle behaviors, approaches to adapting these interventions for a broader population are not well established. Digital behavior change interventions (DBCIs) hold promise as a more generalizable and scalable approach to overcome the resource and time limitations that traditional behavioral intervention programs face, especially within an occupational setting. Objective: We aimed to evaluate the efficacy of a multimodal DBCI on (1) self-reported behaviors of physical activity, nutrition, sleep, and mindfulness; (2) cardiometabolic biomarkers; and (3) chronic disease–related medical expenditure. Methods: We conducted a 2-arm randomized controlled trial for 12 months among employees of an academic health care facility in the United States. The intervention arm received a scale, a smartphone app, an activity tracker, a video library for healthy behavior recommendations, and an on-demand health coach. The control arm received standard employer-provided health and wellness benefits. The primary outcomes of the study included changes in self-reported lifestyle behaviors, cardiometabolic biomarkers, and chronic disease–related medical expenditure. We collected health behavior data via baseline and quarterly web-based surveys, biometric measures via clinic visits at baseline and 12 months, and identified relevant costs through claims datasets. Results: A total of 603 participants were enrolled and randomized to the intervention (n=300, 49.8%) and control arms (n=303, 50.2%). The average age was 46.7 (SD 11.2) years, and the majority of participants were female (80.3%, n=484), White (85.4%, n=504), and non-Hispanic (90.7%, n=547), with no systematic differences in baseline characteristics observed between the study arms. We observed retention rates of 86.1% (n=519) for completing the final survey and 77.9% (n=490) for attending the exit visit. Conclusions: This study represents the largest and most comprehensive evaluation of DBCIs among participants who were not selected based on their underlying condition to assess its impact on behavior, cardiometabolic biomarkers, and medical expenditure. Trial Registration: ClinicalTrials.gov NCT04712383; https://clinicaltrials.gov/study/NCT04712383 International Registered Report Identifier (IRRID): RR1-10.2196/50378

Background and Aims: Alcohol-associated hepatitis (AH) is a clinically severe, acute disease that afflicts only a fraction of patients with alcohol use disorder. Genomic studies of alcohol-associated cirrhosis (AC) have identified several genes of large effect, but the genetic and environmental factors that lead to AH and AC, and their degree of genetic overlap, remain largely unknown. This study aims to identify genes and genetic variations that contribute to the development of AH. Approach and Results: Exome-sequencing of patients with AH (N=784) and heavy drinking controls (N=951) identified an exome-wide significant association for AH at patalin-like phospholipase domain containing 3, as previously observed for AC in genome-wide association study, although with a much lower effect size. Single nucleotide polymorphisms (SNPs) of large effect size at inducible T cell costimulatory ligand (ICOSLG) (Chr 21) and TOX4/RAB2B (Chr 14) were also exome-wide significant. ICOSLG encodes a co-stimulatory signal for T-cell proliferation and cytokine secretion and induces B-cell proliferation and differentiation. TOX high mobility group box family member 4 (TOX4) was previously implicated in diabetes and immune system function. Other genes previously implicated in AC did not strongly contribute to AH, and the only prominently implicated (but not exome-wide significant) gene overlapping with alcohol use disorder was alcohol dehydrogenase 1B (ADH1B). Polygenic signals for AH were observed in both common and rare variant analysis and identified genes with roles associated with inflammation. Conclusions: This study has identified 2 new genes of high effect size with a previously unknown contribution to alcohol-associated liver disease and highlights both the overlap in etiology between liver diseases and the unique origins of AH.

Background: Understanding changes in diagnostic performance after symptom onset and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) exposure within different populations is crucial to guide the use of diagnostics for SARS-CoV-2. Methods: The Test Us at Home study was a longitudinal cohort study that enrolled individuals across the United States between October 2021 and February 2022. Participants performed paired antigen-detection rapid diagnostic tests (Ag-RDTs) and reverse-transcriptase polymerase chain reaction (RT-PCR) tests at home every 48 hours for 15 days and self-reported symptoms and known coronavirus disease 2019 exposures immediately before testing. The percent positivity for Ag-RDTs and RT-PCR tests was calculated each day after symptom onset and exposure and stratified by vaccination status, variant, age category, and sex. Results: The highest percent positivity occurred 2 days after symptom onset (RT-PCR, 91.2%; Ag-RDT, 71.1%) and 6 days after exposure (RT-PCR, 91.8%; Ag-RDT, 86.2%). RT-PCR and Ag-RDT performance did not differ by vaccination status, variant, age category, or sex. The percent positivity for Ag-RDTs was lower among exposed, asymptomatic than among symptomatic individuals (37.5% (95% confidence interval [CI], 13.7%–69.4%) vs 90.3% (75.1%–96.7%). Cumulatively, Ag-RDTs detected 84.9% (95% CI, 78.2%–89.8%) of infections within 4 days of symptom onset. For exposed participants, Ag-RDTs detected 94.0% (95% CI, 86.7%–97.4%) of RT-PCR–confirmed infections within 6 days of exposure. Conclusions: The percent positivity for Ag-RDTs and RT-PCR tests was highest 2 days after symptom onset and 6 days after exposure, and performance increased with serial testing. The percent positivity of Ag-RDTs was lowest among asymptomatic individuals but did not differ by sex, variant, vaccination status, or age category.

Diagnostic challenges continue to impede development of effective therapies for successful management of alcohol-associated hepatitis (AH), creating an unmet need to identify noninvasive biomarkers for AH. In murine models, complement contributes to ethanol-induced liver injury. Therefore, we hypothesized that complement proteins could be rational diagnostic/prognostic biomarkers in AH. Here, we performed a comparative analysis of data derived from human hepatic and serum proteome to identify and characterize complement protein signatures in severe AH (sAH). The quantity of multiple complement proteins was perturbed in liver and serum proteome of patients with sAH. Multiple complement proteins differentiated patients with sAH from those with alcohol cirrhosis (AC) or alcohol use disorder (AUD) and healthy controls (HCs). Serum collectin 11 and C1q binding protein were strongly associated with sAH and exhibited good discriminatory performance among patients with sAH, AC, or AUD and HCs. Furthermore, complement component receptor 1-like protein was negatively associated with pro-inflammatory cytokines. Additionally, lower serum MBL associated serine protease 1 and coagulation factor II independently predicted 90-day mortality. In summary, meta-analysis of proteomic profiles from liver and circulation revealed complement protein signatures of sAH, highlighting a complex perturbation of complement and identifying potential diagnostic and prognostic biomarkers for patients with sAH.

Systemic thromboxane A₂ generation, assessed by quantifying the concentration of stable thromboxane B₂ metabolites (TXB₂-M) in the urine adjusted for urinary creatinine, is strongly associated with mortality risk. We sought to define optimal TXB₂-M cutpoints for aspirin users and nonusers and determine if adjusting TXB₂-M for estimated glomerular filtration rate (eGFR) in addition to urinary creatinine improved mortality risk assessment. Urinary TXB₂-M were measured by competitive ELISA in 1363 aspirin users and 1681 nonusers participating in the Framingham Heart Study. Cutpoints were determined for TXB₂-M and TXB₂-M/eGFR using log-rank statistics and used to assess mortality risk by Cox proportional hazard modeling and restricted mean survival time. Multivariable models were compared using the Akaike Information Criterion (AIC). A cohort of 105 aspirin users with heart failure was used for external validation. Optimized cutpoints of TXB₂-M were 1291 and 5609 pg/mg creatinine and of TXB₂-M/eGFR were 16.6 and 62.1 filtered prostanoid units (defined as pg·min/creatinine·mL·1.73 m²), for aspirin users and nonusers, respectively. TXB₂-M/eGFR cutpoints provided more robust all-cause mortality risk discrimination than TXB₂-M cutpoints, with a larger unadjusted hazard ratio (2.88 vs 2.16, AIC P < 0.0001) and greater differences in restricted mean survival time between exposure groups (1.46 vs 1.10 years), findings that were confirmed in the external validation cohort of aspirin users. TXB₂-M/eGFR cutpoints also provided better cardiovascular/stroke mortality risk discrimination than TXB₂-M cutpoints (unadjusted hazard ratio 3.31 vs 2.13, AIC P < 0.0001). Adjustment for eGFR strengthens the association of urinary TXB₂-M with long-term mortality risk irrespective of aspirin use.