Aptamer proteomics for biomarker discovery in heart failure with reduced ejection fraction

Background

Though current heart failure (HF) biomarkers are highly prognostic, systematically characterizing associations between circulating proteins and risk of subsequent events may improve clinical risk prediction and illuminate new biological pathways. Large-scale assays measuring thousands of proteins now enable unbiased proteomic investigation in clinical trials.

Purpose

To identify and replicate serum proteins associated with HF events in patients with chronic HF with reduced ejection fraction (HFrEF), and to develop and validate a proteomic risk score.

Methods

Serum levels of 4076 proteins were measured at baseline in the ATMOSPHERE (n=1261, 487 events over 6 years) and PARADIGM-HF (n=1257, 287 events over 4 years) trials of chronic HFrEF using a modified aptamer-based proteomics assay. Proteins associated with the primary endpoint, HF hospitalization or cardiovascular death, were identified in the ATMOSPHERE discovery cohort (false discovery rate<0.05) by Cox regression adjusted for age, sex, treatment arm, and anticoagulant use, and replicated in PARADIGM-HF (Bonferroni-corrected p<0.05). A proteomic risk score was derived in ATMOSPHERE using Cox LASSO penalized regression and evaluated in PARADIGM-HF compared to the MAGGIC clinical risk score and N-terminal pro-B-type natriuretic peptide (NT-proBNP). For proteins associated with the primary endpoint, pathway analysis was conducted using Ingenuity Pathway analysis and an exploratory two-sample Mendelian randomization was performed using genetic and outcome data from both trials and protein quantitative trait loci from deCODE to infer which identified proteins contribute to HF prognosis.

Results

We identified 377 serum proteins associated with the primary endpoint in ATMOSPHERE and replicated 167 in PARADIGM-HF. Prognostic proteins included known HF biomarkers Growth Differentiation Factor 15, NT-proBNP, and Angiopoietin-2, and also a previously unrecognized HF biomarker: Sushi, Von Willebrand Factor Type A, EGF And Pentraxin Domain Containing 1 (SVEP1) (HR 1.60 [95% CI 1.44–1.79] per standard deviation [SD], p=2x10–17) (Table 1). Proteins related to hepatic fibrosis, granulocyte adhesion, and inhibition of matrix metalloproteinases were over-represented. A 64-protein risk score derived in ATMOSPHERE predicted clinical events in PARADIGM-HF with greater discrimination (c-statistic 0.70) than the MAGGIC clinical score (c-statistic 0.61), NT-proBNP (c-statistic 0.65), or both (c-statistic 0.66) (Figure 1). Genetically predicted levels of NT-proBNP, WISP2, FSTL1, and CTSS were associated with the primary endpoint by Mendelian randomization.

Conclusions

We identify SVEP1, an extracellular matrix protein known to cause inflammation in vascular smooth muscle cells, as a previously unrecognized HF biomarker. A 64-protein score improved risk discrimination compared with NT-proBNP and may assist in identifying high-risk patients for clinical trials or disease management programs.

Funding Acknowledgement

Type of funding sources: Private company. Main funding source(s): The ATMOSPHERE and PARADIGM-HF trials were sponsored by Novartis

Effects of dapagliflozin on cardiovascular and kidney events by baseline eGFR and UACR in patients with type 2 diabetes mellitus: a patient-level pooled analysis of DECLARE-TIMI 58 and DAPA-CKD trials

Background

The sodium glucose co-transporter 2 (SGLT2) inhibitor dapagliflozin reduced the risk of hospitalization for heart failure (HHF) or cardiovascular death (CVD) and the risk of kidney events in patients type 2 diabetes mellitus (T2DM) and high cardiovascular risk or chronic kidney disease in the DECLARE-TIMI 58 and DAPA-CKD trials. These events are more common at lower levels of kidney function. Combining data from the two trials creates an opportunity to examine the effect of dapagliflozin across the spectrum of baseline kidney function.

Purpose

To determine the effects of dapagliflozin on HHF/CVD and kidney endpoints across a broad range of kidney function in the combined dataset.

Methods

We conducted a post hoc analysis of pooled patient-level data from DECLARE and DAPA-CKD. The effects of dapagliflozin compared with placebo on HHF/CVD and kidney endpoints (defined as sustained eGFR decrease ≥40%, end-stage kidney disease, or renal death) were assessed in the combined cohorts and in subgroups of baseline eGFR (<45, 45-<60, 60-<90, ≥90 mL/min/1.73 m2) and urinary albumin:creatinine ratio (UACR) (<30, 30-<300, 300-<1000, ≥1000 mg/g).

Results

A total of 19,748 patients with T2DM were included. Median (IQR) follow up time was 4.1 (3.7–4.4) years. Median eGFR was 85 (65–95) mL/min/1.73 m2 and UACR 18.2 (7–135) mg/g. Overall, dapagliflozin reduced the risk of HHF/CVD by 18% (HR 0.82, 95% CI 0.73–0.92, p<0.001) and kidney endpoints by 40% (HR 0.60, 95% CI 0.52–0.69, p<0.001). Overall rates of HHF/CVD and kidney endpoints were higher with lower eGFR (p<0.001) and with higher UACR (p<0.001). There were consistent relative risk reductions in HHF/CVD and kidney events with dapagliflozin across eGFR (p-interaction 0.25 and 0.32, respectively, Figure 1) and UACR (p-interaction 0.29 and 0.83, respectively, Figure 2) subgroups. The absolute rate difference (ARD) with dapagliflozin for CVD/HHF ranged from 0.1 events per 1000 patient years in patients in normal categories of eGFR and UACR to 1.0–1.7 events in patients in the most abnormal categories. Likewise, the ARD for kidney events ranged from 0.2 events per 1000 patient years in the normal eGFR and UACR groups to 2.5–4.3 events in patients in the most abnormal categories.

Conclusion

In this pooled analysis of pts with T2DM, there was higher risk of HHF/CVD and kidney events with lower eGFR and higher UACR. Dapagliflozin consistently reduced these events regardless of baseline eGFR and UACR, with large absolute risk reductions in patients with lower eGFR and higher UACR.

Funding Acknowledgement

Type of funding sources: None.

Prognostic implications of NYHA class and NT-proBNP levels in mild heart failure: a PARADIGM-HF analysis

Background

Treatment recommendations for heart failure (HF) with reduced ejection fraction are primarily centered on New York Heart Association (NYHA) classification, such that apparently asymptomatic patients might not be eligible for disease-modifying therapies. NYHA classification, however, may be particularly limited to discriminate mild forms of HF.

Purpose

The present study aimed to determine the relationship between NYHA classification and an objective measure of HF severity (N-terminal pro–B-type natriuretic peptide [NT pro-BNP]), and their association with long-term prognosis in the PARADIGM-HF trial.

Methods

We compared PARADIGM-HF patients classified as NYHA class I, II, and III at randomization (NYHA class IV patients or with unavailable NYHA class were excluded [n=73]). We present kernel density estimation (KDE) plots–a non-parametric way to describe the underlying distribution of a variable–to compare NT-proBNP levels across NYHA classes. Logistic regression and the area under the receiver operating characteristic curve (AUC) were used to assess the ability to predict a patient's NYHA class using NT-proBNP levels. Time-to-event data were calculated with Kaplan–Meier estimates and NYHA class were further stratified by median baseline NT-proBNP (< or ≥1600 pg/ml). The primary outcome was cardiovascular death or first HF hospitalization.

Results

8326 patients were included in this analysis (median age, 64 years; women, 22%; and median left ventricular ejection fraction, 30%). Of 389 patients classified as NYHA class I at randomization, 228 (59%) changed functional class during the first year after randomization. For log-transformed NT-proBNP, KDE overlapped substantially across NYHA classes (Figure 1A). NT-proBNP levels were a poor predictor of NYHA classification: for NYHA class I vs. II, AUC (95% confidence interval [CI]) was 0.51 (0.48–0.54); for NHYA I vs. III, 0.57 (0.54–0.60); and for NYHA II vs. III, 0.56 (0.54–0.57). NYHA class III patients displayed a distinctively higher rate of cardiovascular deaths or first HF hospitalizations (Figure 1B). NYHA class I and II patients revealed lower event rates that were not significantly different (NYHA II vs. I, HR 1.24 [0.97–1.58]). Stratification by NT-proBNP levels identified subgroups with distinctive risk, such that NYHA I patients with high NT-proBNP levels (n=175) had a higher event rate than patients with low NT-proBNP with any NYHA class (Figure 1C).

Conclusion

NYHA class I and II patients overlapped substantially in objective HF measures and long-term prognosis. NYHA classification remains a powerful predictor of cardiovascular events but might be limited to differentiate mild forms of HF, as apparently asymptomatic patients based on physician-defined functional class might become symptomatic within a year and conceal subjects at substantial risk for adverse outcomes.

Funding Acknowledgement

Type of funding sources: None.

Machine learning to aid in the diagnosis of acute heart failure in the emergency department

Background

B-type natriuretic peptide (BNP) and mid-regional pro-atrial natriuretic peptide (MRproANP) testing are recommended to aid in the diagnosis of acute heart failure. However, the application of these biomarkers for optimal diagnostic performance is uncertain.

Methods

We performed a systematic review and harmonised individual patient-level data to evaluate the diagnostic performance of BNP and MRproANP for the diagnosis of acute heart failure using random-effects meta-analysis. We subsequently developed and externally validated a decision-support tool called CoDE-HF for both BNP and MRproANP that combines the natriuretic peptide concentrations with clinical variables using machine learning to report the probability of acute heart failure for an individual patient.

Results

Fourteen studies from 12 countries provided individual patient-level data in 8,493 patients for BNP and 3,847 patients for MRproANP, in whom, 48.3% (4,105/8,493) and 41.3% (1,611/3899) had an adjudicated diagnosis of acute heart failure, respectively. The negative and positive predictive values of guideline-recommended thresholds for BNP (100 pg/mL) and MR-proANP (120 pg/mL) were 93.6% (95% confidence interval 88.4–96.6%) and 68.8% (62.9–74.2%), and 95.6% (92.2–97.6%) and 64.8% (56.3–72.5%), respectively. However, we observed significant heterogeneity in the diagnostic performance across important patient subgroups (Figure 1). In the external validation cohort, CoDE-HF was well calibrated with excellent discrimination in those without prior acute heart failure for both BNP and MRproANP (area under the curve of 0.946 [0.933–0.958] and 0.943 [0.921–0.964], and Brier scores of 0.105 and 0.073, respectively). CoDE-HF performed consistently across all subgroups for both BNP and MRproANP, and identified 30% and 65.7% at low-probability (negative predictive value of 99.1% [98.8–99.3%] and 99.1% [98.8–99.4%]), and 30% and 17.3% at high-probability (positive predictive value of 91.3% [90.7–91.9%] and 70.0% [68.5–71.4%]) in those without prior heart failure, respectively (Figure 2).

Conclusion

In an international collaborative analysis, we observed that guideline-recommended thresholds for BNP and MRproANP to diagnose acute heart failure varied significantly across patient subgroups. A decision-support tool using machine learning to combine natriuretic peptides as a continuous measure and other clinical variables provides a more accurate and individualised approach.

Funding Acknowledgement

Type of funding sources: Other. Main funding source(s): Medical Research Council and British Heart Foundation Figure 1. NPV of BNP threshold (100 pg/mL)Figure 2. NPV of the CoDE-HF rule-out score

Elevated markers of liver function are associated with poorer outcomes in HFREF: an analysis of DAPA-HF

Background

Abnormalities of liver tests in patients with heart failure with reduced ejection fraction (HFrEF) is a well-recognised phenomenon. We examined the prognostic value of measures of liver function in a large contemporary cohort of patients with HFrEF enrolled in the Dapagliflozin and Prevention of Adverse Outcomes in Heart Failure (DAPA-HF) trial

Methods

In this post-hoc analysis of the DAPA-HF trial we studied 4625 patients with liver function tests available at baseline. Cox proportional hazards models were used to assess the association between liver tests (total bilirubin, alkaline phosphatase [ALP], alanine transaminase [ALT], aspartate transaminase [AST]) and the Model for End-stage Liver Disease excluding INR (MELD-XI) score (calculated as 5.11 Ln [total bilirubin as mg/dL] + 11.76 Ln [creatinine as mg/dL] + 9.44), and the risk of the primary composite endpoint (hospitalisation or urgent visit for heart failure or cardiovascular death). Models were adjusted for age, sex, race, region, systolic blood pressure, heart rate, LVEF, eGFR, log-transformed NT-proBNP, NYHA class, history of hypertension, stroke, myocardial infarction, atrial fibrillation, heart failure aetiology and randomized treatment to dapagliflozin and stratified by diabetic status at baseline. An interaction term between each measure and the effect of treatment on the primary composite outcome was tested as a fractional polynomial.

Results

Total bilirubin, ALP, and MELD-XI score were associated with a higher risk of all the primary outcome (Figure 1) but not ALT or AST. These relationships persisted after adjustment: total bilirubin per log unit increase (HR=1.46; 95% CI 1.28 – 1.67, p<0.001), ALP per log unit increase (HR=1.39; 95% CI 1.15 – 1.66, p<0.001), MELD-XI per 1 SD increase (HR 1.27; 95% CI 1.13 – 1.42, p<0.001). The effect of dapagliflozin on the primary outcome was not modified by the baseline levels of either total bilirubin, ALP or MELD-XI score (Figure 2)

Conclusions

Higher total bilirubin, ALP and MELD-XI score were independently associated with a higher risk of cardiovascular death or worsening HF and may be useful routinely available biomarkers to assess prognosis. The efficacy of dapagliflozin was the not modified by baseline levels of any of these markers.

Funding Acknowledgement

Type of funding sources: None. Figure 2

127Prevalence of coronary artery disease and coronary microvascular dysfunction in heart failure with preserved ejection fraction

Background/Introduction

The prevalence of epicardial coronary artery disease (CAD) and coronary microvascular dysfunction (CMD) have not been studied systematically in an unselected cohort of patients with heart failure and preserved ejection fraction (HFpEF). Both types of coronary disease may play an important role in the pathophysiology and prognosis of HFpEF.

Methods

This prospective multi-centre observational study enrolled near-consecutive patients hospitalized with HFpEF. Patients underwent invasive coronary angiography. Where possible, patients also had guidewire-based assessment of fractional flow reserve, coronary flow reserve (CFR) and index of microcirculatory resistance (IMR) performed, followed by vasoreactivity testing with intracoronary acetylcholine.

Results

A total of 75 patients underwent invasive coronary angiography. Guidewire-based assessment of FFR/CFR/IMR was performed in 62 patients, and vasoreactivity testing was possible in 41 patients. Obstructive epicardial CAD was identified in 38 patients (51%). CMD (defined as a CFR <2.0 and/or IMR ≥25) was present in 66% of patients assessed and was similarly prevalent in those with and without obstructive epicardial disease (62% vs. 69%, p 0.52). During vasoreactivity testing, 24% of those assessed had evidence of coronary microvascular endothelial dysfunction. Patients with obstructive CAD were more often male (63% vs. 38%, p 0.028), and had a history of CAD (50% vs. 19%, p 0.005), diabetes mellitus (63% vs. 41%, p 0.05), and a higher E/e' on echocardiography (median 14.4 vs. 12.3, p 0.044) than those without obstructive coronary disease. Patients with CMD had higher B-type natriuretic peptide levels (median 569 vs. 197 pg/ml, p 0.036) than those without microvascular dysfunction.

Selected baseline characteristics No obstructive CAD (n=37) Obstructive CAD (n=38) p-value No CMD (n=21) CMD (n=41) p-value Age (mean, years) 72 73 0.4 74 72 0.41 Female, n (%) 23 (62%) 14 (37%) 0.028 11 (52%) 22 (54%) 0.92 CAD history, n (%) 7 (19%) 19 (50%) 0.005 7 (33%) 12 (29%) 0.74 Diabetes mellitus, n (%) 15 (41%) 24 (63%) 0.05 11 (52%) 22 (54%) 0.92 BNP (median, pg/ml) 323 315 0.9 197 569 0.036 Ejection fraction (median, %) 59 58 0.35 60 56 0.064 E/e' (median) 12.3 14.4 0.044 14.2 12.4 0.74

Study flow diagram

Conclusion

Both epicardial CAD and CMD are common in HFpEF and each may be a therapeutic target in this condition. Although it has been hypothesized that CMD may be due to endothelial dysfunction, our findings suggest that CMD is predominantly due to structural abnormalities in HFpEF.

Acknowledgement/Funding

Chief Scientist Office

Ontological modeling of electronic health information exchange

INTRODUCTION

Investments of resources to purposively improve the movement of information between health system providers are currently made with imperfect information. No inventories of system-level electronic health information flows currently exist, nor do measures of inter-organizational electronic information exchange.

METHODS

Using Protégé 4, an open-source OWL Web ontology language editor and knowledge-based framework, we formalized a model that decomposes inter-organizational electronic health information flow into derivative concepts such as diversity, breadth, volume, structure, standardization and connectivity.

RESULTS

The ontology was populated with data from a regional health system and the flows were measured. Individual instance's properties were inferred from their class associations as determined by their data and object property rules. It was also possible to visualize interoperability activity for regional analysis and planning purposes. A property called Impact was created from the total number of patients or clients that a health entity in the region served in a year, and the total number of health service providers or organizations with whom it exchanged information in support of clinical decision-making, diagnosis or treatment. Identifying providers with a high Impact but low Interoperability score could assist planners and policy-makers to optimize technology investments intended to electronically share patient information across the continuum of care. Finally, we demonstrated how linked ontologies were used to identify logical inconsistencies in self-reported data for the study.

Modelling system level health information exchange: an ontological approach

Investment of resources to purposively improve the movement of information between health system providers is currently made with imperfect information. No inventories of system-level digital information flows currently exist, nor do measures of inter-organizational electronic information. exchange (HIE). Using Protégé 4, an open-source OWL Web ontology language editor and knowledge-based framework we formalized a model that decomposes inter-organizational electronic health information flow into derivative concepts such as diversity, breadth, volume, structure, standardization and connectivity. Self-reported data from a regional health system is used to measure HIE; the ontology identifies providers with low and high HIE, useful for planners, and using a related database is used to monitor data quality.