Saxagliptin is a DPP-4 inhibitor widely used to manage type 2 diabetes, though genetic contributors to variability in response remain unclear. This study aimed to identify genetic variants associated with the efficacy and safety of saxagliptin using genome-wide association studies (GWAS) with data from randomized controlled trials. GWAS were conducted on glycated hemoglobin (HbA1c) levels and hypoglycemic events in saxagliptin-treated individuals from 12 phase II or III clinical trials. Additional exploratory traits included fasting glucose, weight, body mass index, systolic and diastolic blood pressure, heartbeat, and renal function. Primary analyses focused on participants of European ancestry, followed by analyses in the overall multi-ancestry cohort. Linear mixed models for continuous traits and Cox proportional hazards models for time-to-event analyses were used, including gene-by-sex and gene-by-treatment interactions. We also performed a transcriptome-wide association study (TWAS) to identify genes whose predicted expression is associated with the response to saxagliptin using the S-PrediXcan framework. Among 1 016 European and 1 826 multi-ancestry saxagliptin-treated participants, no genome-wide significant associations were observed. Suggestive associations included rs2168426:T > C near UBE2E1 for HbA1c (p = 1.0 × 10⁻⁷) and rs138558907:C > T near CCSER2 for hypoglycemia (p = 6.1 × 10⁻⁸). Treatment interaction analyses showed a stronger rs2168426 effect on HbA1c among saxagliptin users compared with non-users (pint = 9.2 × 10⁻³). The TWAS revealed three genes that significantly associated with HbA1c in human tissues: OGFOD3 in aorta (p = 8.3 × 10⁻⁷), TBC1D3D in coronary artery (p = 3.3 × 10⁻⁶), and TYSND1 in amygdala (p = 3.9 × 10⁻⁶). In the multi-ancestry cohort, KDM7A in whole blood was significantly associated with HbA1c (p = 3.6 × 10⁻⁶). Larger, diverse studies are needed to identify pharmacogenomic determinants of responses to saxagliptin.

Cognitive impairment can affect up to 50% of patients with chronic heart failure (CHF) and is associated with reduced treatment adherence, high mortality rates, and poor quality of life. Nonpharmacologic strategies, including cognitive intervention and physical exercise training, may help enhance cognition in patients with CHF. Recent studies in dementia prevention have shown that combining cognitive and exercise interventions could have synergistic effects on cognition, but scientific evidence for the benefits in CHF patients is lacking. Moreover, how men and women with heart failure may differ in their response to nonpharmacologic interventions is also unknown. This randomized controlled trial will investigate the effects of combining cognitive and exercise training, on cognition and cerebral blood flow regulation in men and women with CHF. To achieve this, 216 participants (50% female) with stable CHF regardless of etiology and left ventricular ejection fraction will be randomized to 1 of the 3 following arms: 1. combined cognitive and exercise training; 2. exercise training alone; 3. usual medical care with standard cardiovascular rehabilitation. The first 2 groups will engage in a 6-month intervention, whereas those in group 3 will take part in a standard 3-month cardiac rehabilitation program. The primary endpoint will be changes in cognitive performance from baseline to 6 months based on 4 cognitive composite scores (global cognitive functioning, memory, executive functions, processing speed). Secondary outcomes will include changes in cerebral blood flow regulation (neurovascular coupling, pulsatility, and autoregulation). Tertiary outcomes will include cardiorespiratory fitness, physical functioning, and quality of life. Clinical Trial Registration:NCT04970888.

Background Excess adiposity, most commonly indexed through body mass index (BMI), is strongly associated with the development of heart failure (HF). Weight loss therapies improve outcomes in patients with obesity and HF with preserved left ventricular ejection fraction (LVEF), but their effects in HF with reduced LVEF remain unclear. Objectives The aim of this work is to determine whether higher BMI is associated with adverse clinical outcomes in patients with HF and whether there is effect modification by LVEF subgroup. Methods Two-sample Mendelian randomization (MR) was used, with genome-wide significant loci associated with BMI as instrumental variables and outcome data from a genome-wide association study (GWAS) of time-to-event clinical outcomes in patients with HF. A total of 50,636 individuals of European ancestry with established HF from 22 cohorts were included in the genetic analysis: 12 HF trials, 1 prospective case-cohort study, 9 cohorts nested within non-HF cardiovascular trials, and 1 population-based cohort derived from the UK Biobank.The exposure was genetically predicted BMI and the outcome measures were all-cause mortality and a composite of cardiovascular mortality or HF hospitalization. Genetic associations for the outcomes were derived from our GWAS and MR was used to estimate the unbiased association of genetically predicted BMI with these clinical outcomes. Results The mean BMI was 29.2 ± 5.8 kg/m2. Over a median follow-up of 27.0 months, all-cause mortality occurred in 11,454 patients (23%), and 11,360 participants (22%) experienced the composite endpoint. Genetically predicted BMI was associated with an increased rate of both all-cause mortality (HR per SD [4.8 BMI units] 1.21; 95% CI: 1.13-1.29; P = 9 × 10-8) and the composite outcome (HR 1.29; 95% CI: 1.20-1.38; P = 8 × 10-13). Associations were consistent across LVEF ≤40% and >40%: for all-cause mortality, HR: 1.16 (95% CI: 0.99-1.37) and 1.20 (95% CI: 0.94-1.53); and for the composite outcome, HR: 1.30 (95% CI: 1.15-1.48) and 1.57 (95% CI: 1.29-1.91), respectively. Conclusions Among patients with HF, higher BMI was associated with increased all-cause mortality and cardiovascular death or HF hospitalization, supporting the potential role of weight-management strategies across the ejection fraction spectrum.

BACKGROUND AND AIMS:The 12-lead electrocardiogram (ECG) remains a cornerstone of cardiac diagnostics, yet existing artificial intelligence (AI) solutions for automated interpretation often lack generalizability, remain closed source, and are primarily trained using supervised learning (SL), which requires extensive labelled datasets and may limit adaptability across diverse clinical settings. Self-supervised learning (SSL) can potentially overcome these limitations by learning robust representations from unlabelled data. To address these challenges, this study developed and compared two open-source foundational ECG models: DeepECG-SL, a supervised multilabel ECG model, and DeepECG-SSL, a self-supervised model. METHODS:Both models were trained on over 1 million ECGs using a standardized preprocessing pipeline and automated free-text extraction from ECG reports to predict 77 cardiac conditions. DeepECG-SSL leveraged unlabelled data through self-supervised contrastive learning and masked lead modelling before fine-tuning for downstream tasks, while DeepECG-SL was trained directly on labelled diagnostic data in an end-to-end fashion. Performance was evaluated across seven private, multilingual healthcare systems and four public ECG repositories, with assessment of fairness by age and sex, and investigation of privacy vulnerabilities as well as memory and compute requirements. RESULTS:DeepECG-SSL achieved micro-averaged area under the receiver operating characteristic curves (AUROCs) across all 77 cardiac conditions for ECG interpretation of 0.990 [95% confidence interval (CI): 0.990, 0.990] on the internal dataset (MHI-ds), 0.981 (95% CI: 0.981, 0.981) on external public datasets (UKB, CLSA, MIMIC-IV and PTB), and 0.983 (95% CI: 0.983, 0.983) on external private datasets (UW, UCSF, JGH, NYP, MGH, CSH and CHUM), while DeepECG-SL demonstrated AUROCs of 0.992 (95% CI: 0.992, 0.992), 0.980 (95% CI: 0.980, 0.980), and 0.983 (95% CI: 0.983, 0.984), respectively. Fairness analyses revealed minimal disparities (true-positive rate and false-positive rate difference <0.1) across age and sex groups for both models. DeepECG-SSL demonstrated superior performance on limited-data digital biomarker tasks, with the largest improvements in long QT syndrome (LQTS) genotype classification (AUROC 0.931 vs 0.850, P = .026, n = 127 ECGs) and 5 year atrial fibrillation risk prediction (AUROC 0.742 vs 0.734, P < 0.001, n = 132 050 ECGs), while achieving superior performance in left ventricular ejection fraction ≤40% classification (AUROC 0.926 vs 0.917, P < 0.001, n = 25 252 ECGs) and comparable performance in LQTS detection (AUROC 0.767 vs 0.735, P = 0.117, n = 934 ECGs). CONCLUSIONS:This study establishes SSL as a promising paradigm for ECG analysis, particularly in settings with limited annotated data, enhancing accessibility, generalizability, and fairness in AI-driven cardiac diagnostics. By releasing model weights, preprocessing tools, and validation code, this work aims to support robust, data-efficient AI diagnostics across diverse clinical environments and questions.

Numerous pharmacogenomic (PGx) associations of pharmacokinetics and pharmacodynamics of commonly prescribed medications have been reported. Yet, the evidence supporting their association on survival remains underexamined. We sought to investigate the CYP2D6-metoprolol association and determine whether the established variability in drug exposure and hemodynamic response would translate into differences in mortality across metabolizer status in the Montreal Heart Institute Hospital Cohort. Single random plasma samples were collected from 996 patients receiving metoprolol tartrate. Bioanalytical quantification was performed via liquid chromatography-tandem mass spectrometry, while CYP2D6 metabolizer status was based on standardized classifications. Cox regression models adjusted for age, sex, cardiovascular history, concomitant medications, and CYP2D6 inhibitors were used to assess the time to death since enrolment. Overall, 24.3% (n = 242) of patients were deceased at follow-up (median 101.4 months). Higher metabolizer status at CYP2D6 was associated with lower risk of death (HR = 0.82, 95% CI: 0.67-0.99; p = 0.04). Metoprolol concentrations were no longer associated with mortality after adjusting for possible confounders. Our study suggests that CYP2D6-inferred metabolizer status can inform on mortality risk in patients treated with metoprolol. Larger initiatives are required to confirm the association of CYP2D6 metabolizer status with clinical events beyond pharmacokinetic and hemodynamic considerations.

Brain-Derived Neurotrophic Factor (BDNF), highly enriched in platelets, contributes to vascular integrity and thrombotic responses. The rs11030119 variant, located in a BDNF intronic enhancer, has been linked to stroke recovery, but its influence on peripheral BDNF dynamics and hemostasis remains unexplored. In this study, we aimed to understand the impact of rs11030119 on circulating levels of BDNF, and its subsequent effects on thrombus formation. Twenty-six healthy individuals homozygous for either the GG or AA genotype of rs11030119 were matched for age and sex. BDNF and proBDNF levels were quantified in serum, plasma, and platelets. Platelet aggregation, ATP secretion, thrombin generation, and clot firmness were assessed. Washed platelets were also tested with recombinant BDNF. AA carriers displayed significantly lower plasma and serum BDNF levels and diminished BDNF release upon activation, despite comparable platelet BDNF content. The expression of its receptor, TrkB, on the platelet surface was also reduced in AA carriers. However, platelet reactivity, thrombin generation, and viscoelastic clot properties were preserved across genotypes. In conclusion, rs11030119 modulates circulating and platelet-releasable BDNF without impairing hemostatic function, suggesting that BDNF bioavailability may not directly correlate with thrombotic potential. These findings uncover a novel genetic mechanism controlling platelet-derived neurotrophin output.

Sex differences, in terms of prevalence, symptoms and disease progression, are established in the aetiology of complex neurodegenerative diseases, including amyotrophic lateral sclerosis, Parkinson's disease and Alzheimer's disease, but the underlying biology driving these differences remains poorly understood. There is emerging evidence from genetic and functional analyses affirming the role of the immune system in such diseases, but a thorough assessment of sex differences in the link between the immune system and neurodegenerative diseases remains lacking.Here, we applied a robust causal inference approach, two-sample Mendelian randomization, to evaluate the causal effect of immune-related protein levels on three neurodegenerative diseases with large-scale sex-stratified genome-wide association data available: amyotrophic lateral sclerosis (females = 10 895 cases, 57 062 controls; males = 15 547 cases, 50 145 controls); Parkinson's disease (females = 7947 cases, 90 662 controls; males = 13 020 cases, 89 660 controls); and Alzheimer's disease (females = 18 822 cases, 281 415 controls; males = 17 293 cases, 213 339 controls). As exposures, we focused on 932 immune system-related proteins with significant protein cis-quantitative trait loci (false discovery rate cut-off < 0.01) from a large sex-combined plasma protein dataset (n = 33 477), for which corresponding genes were included in the Immunology Database and Analysis Portal gene list. We tested for a causal relationship between genetically predicted levels of each of these proteins and each neurodegenerative disease in sex-stratified and sex-combined data, followed by colocalization and estimation of sex-differential effects. We additionally performed exploratory analyses using sex-combined CSF protein cis-quantitative trait loci (n = 971) as exposures.We observed evidence for a sex-differential causal relationship between FCGR2A and Parkinson's disease and between CD2AP, MAMDC2, PCDH17 or CSF3 and Alzheimer's disease. We validated significant results using two independent protein cis-quantitative trait loci datasets for those plasma proteins available. After performing sensitivity analyses, we validated the potential causal relationships of OMG on Parkinson's disease and of GRN, SERPINF2 and TREM2 on Alzheimer's disease. Mendelian randomization with CSF protein cis-quantitative trait loci showed a potential causal effect of ADGRE2, GPNMB and COLEC11 on Parkinson's disease and of CD33 on Alzheimer's disease, without evidence of sex-differential effects. Finally, we substantiated our findings of protein-disease pairs using triangulation, specifically reporting independent supporting evidence from the literature and drug-related databases.Overall, our results point to potential causal effects of genetically predicted levels of immune system-related plasma and CSF proteins in Alzheimer's disease and Parkinson's disease, some of which may be considered as potential candidates for drug development.

Background:The 12-lead electrocardiogram (ECG) remains a cornerstone of cardiac diagnostics, yet existing artificial intelligence (AI) solutions for automated interpretation often lack generalizability, remain closed-source, and are primarily trained using supervised learning, limiting their adaptability across diverse clinical settings. To address these challenges, we developed and compared two open-source foundational ECG models: DeepECG-SSL, a self-supervised learning model, and DeepECG-SL, a supervised learning model. Methods:Both models were trained on over 1 million ECGs using a standardized preprocessing pipeline and automated free-text extraction from ECG reports to predict 77 cardiac conditions. DeepECG-SSL was pretrained using self-supervised contrastive learning and masked lead modeling. The models were evaluated on six multilingual private healthcare systems and four public datasets for ECG interpretation across 77 diagnostic categories. Fairness analyses assessed disparities in performance across age and sex groups, while also investigating fairness and resource utilization. Results:DeepECG-SSL achieved AUROCs of 0.990 (95%CI 0.990, 0.990) on internal dataset, 0.981 (95%CI 0.981, 0.981) on external public datasets, and 0.983 (95%CI 0.983, 0.983) on external private datasets, while DeepECG-SL demonstrated AUROCs of 0.992 (95%CI 0.992, 0.992), 0.980 (95%CI 0.980, 0.980) and 0.983 (95%CI 0.983, 0.983) respectively. Fairness analyses revealed minimal disparities (true positive rate & false positive rate difference<0.010) across age and sex groups. Digital biomarker prediction (Long QT syndrome (LQTS) classification, 5-year atrial fibrillation prediction and left ventricular ejection fraction (LVEF) classification) with limited labeled data, DeepECG-SSL outperformed DeepECG-SL in predicting 5-year atrial fibrillation risk (N=132,050; AUROC 0.742 vs. 0.720; Δ=0.022; P<0.001), identifying reduced LVEF ≤40% (N=25,252; 0.928 vs. 0.900; Δ=0.028; P<0.001), and classifying LQTS syndrome subtypes (N=127; 0.931 vs. 0.853; Δ=0.078; P=0.026). Conclusion:By releasing model weights, preprocessing tools, and validation code, we aim to support robust, data-efficient AI diagnostics across diverse clinical environments. This study establishes self-supervised learning as a promising paradigm for ECG analysis, particularly in settings with limited annotated data, enhancing accessibility, generalizability, and fairness in AI-driven cardiac diagnostics.

Supplementary Table 8: The effect of mosaic loss of X chromosome on death of cardiovascular disease causes, coronary artery disease causes, from cancer, and any cause of death

BACKGROUND:Acute coronary syndrome (ACS) survivors have heightened risk for subsequent cardiovascular events. METHODS:All baseline characteristics collected in both the Dal-Outcomes and Dal-GenE trials were considered as potential risk markers. A prediction index for subsequent fatal and non-fatal myocardial infarction (MI) following ACS was developed using Cox proportional hazards modeling on data from Dal-Outcomes placebo patients (n=7086). This prediction index was then applied in all Dal-GenE participants (n=5989) to determine whether the reduction in MI observed with dalcetrapib (versus placebo) in patients with the AA genotype at rs1967309 in the ADCY9 gene remained significant, independent of the other markers integrated into the prediction index. RESULTS:Of the 36 baseline variables considered as potential risk markers, 18 contributed to the prediction index with a Harrell's C-index of 0.72 (95% CI, 0.69-0.75) in Dal-Outcomes placebo patients. Prior history of coronary events, LDL-C, blood pressure, A1c, hs-CRP, smoking and age were contributors. The prediction index was strongly predictive when applied to the 5989 AA genotype patients from Dal-GenE, with a HR for MI of 1.92 (95%CI: 1.78-2.08) for each SD increase in score. When adjusting for the prediction index, the HR for dalcetrapib versus placebo was 0.77 (95% CI, 0.63-0.94) in Dal-GenE. CONCLUSION:Despite guideline directed therapy following ACS, history of prior coronary events and on-treatment LDL-C, A1c, hs-CRP and blood pressure remain determinants of future MI. In the Dal-GenE AA genotype patients, dalcetrapib reduced the rate of MI, independently of those variables. The Dal-GenE 2 trial is designed to confirm this pharmacogenetic hypothesis.

Supplementary Table 2: ICD-9 and ICD-10 diagnostic and procedure codes for inpatient cardiovascular-related endpoints and for the primary endpoints

Supplementary Table 13: Cox regression analyses for the effect of mosaic chromosomal alterations on the risk of death of coronary artery disease causes by cancer type

Supplementary Table 9: The effect of mosaic loss of Y chromosome on death of cardiovascular disease causes, coronary artery disease causes, from cancer and any cause of death

Supplementary Table 16: Cox regression analyses for the effect of mosaic chromosomal alterations on the risk of incident cardiovascular endpoints aged ≥65 years old (n=15,273)

Background & Aims The relationship between alcohol consumption and type 2 diabetes risk is often described as a J- or U-shaped curve, with moderate drinkers having a lower risk compared to non-drinkers and heavy drinkers. However, this protective effect appears to be more pronounced in women than in men, suggesting a potential interaction between sex-specific factors and alcohol metabolism. Methods We conducted an interaction genome-wide association study (GWAS) to identify genetic variants that modify the relationship between alcohol consumption and type 2 diabetes risk in a sex-specific manner. We utilized data from the UK Biobank in a case-control approach including 309,568 individuals to investigate the three-way interaction between genetic variants, alcohol consumption, and sex on type 2 diabetes risk. Results We identified genetic variant rs78681203, located between the FOXO6 and EDN2 genes, with a significant sex-specific interaction with alcohol consumption (interaction p = 2.85 x 10-8). The T allele of rs78681203 was associated with an increased risk of type 2 diabetes in women consuming 0 to <4 UK units/week (OR = 1.19, 95% CI: 1.07-1.34) but had a protective effect in women consuming 4 to <28 UK units/week (OR = 0.81, 95% CI: 0.69-0.95). Conversely, in men, the T allele was associated with a higher risk of type 2 diabetes in the 4 to <28 UK units/week group (OR = 1.19, 95% CI: 1.08-1.32) and had a protective effect in the 0 to <4 UK units/week group (OR = 0.85, 95% CI: 0.74-0.99). Conclusions Our findings suggest that genetic variation may play a role in the differential effects of alcohol consumption on type 2 diabetes risk between men and women. Further replication and mechanistic studies are needed to confirm and clarify the role of the identified genetic variant.

Supplementary Table 3: Sensitivity analyses by Cox regression considering additional covariates

Supplementary Table 1: Primary diagnostic codes for selected cancers

Common genetic variation detected by genome-wide association studies (GWAS) partially explains variability in the spectrum of cardiac phenotypes. In this work, we explore genetic correlations among 58 cardiac-related traits/diseases, detecting novel ones. We subsequently employ multi-trait analysis of GWAS (MTAG), which meta-analyzes genetically correlated traits, to improve genomic loci discovery and prediction in atrial fibrillation (AF), coronary artery disease (CAD), and heart failure (HF). We identify 19 novel loci specific for AF, 131 for CAD, and 141 for HF. Polygenic scores (PGS) in 15,177 Canadian individuals show similar results when PGS are derived from conventional GWAS versus MTAG summary statistics, although MTAG-PGS improve prediction and discrimination of CAD in females [∆R2 1.735

Statins are first-line treatments in the primary and secondary prevention of cardiovascular disease. Clinical studies show statins act independently of lipid-lowering mechanisms to decrease C-reactive protein (CRP), an inflammation marker. We aim to elucidate genetic loci associated with CRP statin response. CRP statin response is the change in log-CRP between off-treatment and on-treatment measurements. Cohort-level Genome-Wide Association Studies (GWAS) of CRP response were performed using 1000 Genomes imputed data, testing ∼10 million common genetic variants. GWAS meta-analysis combined results from seven cohorts and clinical trials totalling 14,070 statin-treated individuals of European ancestry within the GIST consortium. Secondary analyses included statin-by-placebo interaction analyses, and lookups in African ancestry cohorts. Our GWAS identified two genome-wide significant (P < 5e-8) loci: APOE and HNF1A for CRP statin response corrected for baseline CRP. The missense lead variant rs429358 at APOE, contributing to the APOE-E4 haplotype, is a risk locus for dyslipidaemia, Alzheimer's and coronary artery disease (CAD). The HNF1A locus is associated with diabetes, cholesterol levels, and CAD. Both loci are also associated with baseline CRP levels, and neither locus achieved a significant (P < 0.05) result from the statin v. placebo interaction meta-analysis using randomized clinical trial data. However, the interaction result (P-int=0.09) for APOE was suggestive and possibly underpowered. The APOE-E4 signal may therefore be associated with both CRP and LDL-cholesterol statin response. Combined with suggestions in the literature that APOE also leads to differential statin benefit in Alzheimer's, the APOE locus warrants further investigation for potential genetic effects on healthcare with statin treatment.

INTRODUCTION:The apolipoprotein E (APOE) ɛ4 allele is a well-established risk factor for neurocognitive impairment (NCI), with varying impacts between men and women. This study investigates the distinct roles of sex and gender in modifying APOE ɛ4-related NCI. METHODS:Biological sex was inferred from sex chromosomes, and a femininity score (FS) was used as a proxy for gender. We analyzed 276,596 UK Biobank participants without prior NCI to assess whether sex and FS modified the effect of APOE ɛ4 on NCI. RESULTS:NCI risk was higher in APOE ɛ4 carriers compared to non-carriers (hazard ratio [HR] = 2.48 in females; HR = 1.96 in males) with significant interaction by sex (P < 0.0001). FS was associated with an increased NCI risk after accounting for sex (HR = 1.07, 95% confidence interval: 1.04-1.10, P < 0.0001) with no significant differences by sex or APOE ɛ4 carrier status. DISCUSSION:Our findings show that APOE ɛ4 increases NCI risk more in females, while FS independently elevates risk across sexes. Highlights:Apolipoprotein E (APOE) ɛ4 increases neurocognitive impairment (NCI) risk, with a greater impact in females (hazard ratio [HR] = 2.48) than males (HR = 1.96).Sex significantly modifies the effect of APOE ɛ4 on NCI (P < 0.0001f).Femininity score increases NCI risk (HR = 1.07) independently of sex and APOE ɛ4.Understanding the distinct sex and gender contributions to APOE ɛ4-related NCI can improve interventions.