https://orcid.org/0000-0003-2555-1266
Abstract
Several biomarkers are associated with clinical outcomes in patients with atrial fibrillation (AF), but a causal relationship has not been established. This study aimed to evaluate angiopoietin-2, a novel candidate biomarker of endothelial inflammation and vascular remodelling, in patients with AF.
Angiopoietin-2 was measured in plasma obtained from patients with AF treated with aspirin monotherapy (exploration cohort, n = 2987) or with oral anticoagulation (validation cohort, n = 13 079). Regression models were built to assess the associations between angiopoietin-2, clinical characteristics, and outcomes. In both cohorts, plasma angiopoietin-2 was independently associated with AF on the baseline electrocardiogram and persistent/permanent AF, age, history of heart failure, female sex, tobacco use/smoking, body mass index, renal dysfunction, diabetes, and N-terminal pro-B-type natriuretic peptide (NT-proBNP). Angiopoietin-2 was independently associated with subsequent hospitalization for heart failure after adjusting for age, creatinine, and clinical characteristics in the exploration cohort [c-index 0.79, 95% confidence interval (CI) 0.75–0.82; third vs. first quartile, hazard ratio (HR) 1.74, 95% CI 1.26–2.41] and in the validation cohort (c-index 0.76, 95% CI 0.74–0.78; HR 1.58, 95% CI 1.37–1.82). In both cohorts, the association persisted when also adjusting for NT-proBNP (P ≤ 0.001). In full multivariable models also adjusted for NT-proBNP, angiopoietin-2 did not show statistically significant associations with ischaemic stroke, cardiovascular and all-cause death, or major bleeding that were consistent across the two cohorts.
In patients with AF, plasma levels of angiopoietin-2 were independently associated with subsequent hospitalization for heart failure and provided incremental prognostic value to clinical risk factors and NT-proBNP.
This study evaluated angiopoietin-2, a candidate biomarker of endothelial inflammation and vascular remodelling, in patients with atrial fibrillation (AF).
In both an exploration cohort of patients enrolled in ACTIVE A and AVERROES (n = 2987) and in a validation cohort of patients enrolled in ARISTOTLE (n = 13 079), angiopoietin-2 was independently associated with AF on the baseline electrocardiogram and persistent/permanent AF, age, history of heart failure, female sex, tobacco use/smoking, body mass index, renal dysfunction, diabetes, and N-terminal pro-B-type natriuretic peptide (NT-proBNP).
Angiopoietin-2 was independently associated with subsequent hospitalization for heart failure after adjusting for age, creatinine, and clinical characteristics in the exploration cohort [c-index 0.79, 95% confidence interval (CI) 0.75–0.82; third vs. first quartile, hazard ratio (HR) 1.74, 95% CI 1.26–2.41] and in the validation cohort (c-index 0.76, 95% CI 0.74–0.78; HR 1.58, 95% CI 1.37–1.82).
In both cohorts, the association with heart failure persisted when also adjusting for NTproBNP (P ≤ 0.001).
Introduction
Risk assessment may inform treatment decisions in patients with atrial fibrillation (AF). Historically, risk prediction modelling in AF has mainly been based on patient age and other clinical variables.1,2 More recently, cardiac biomarkers measured in plasma have been shown to be associated with clinical outcomes beyond traditional risk factors. Studies in large, prospective cohorts of patients with AF treated with oral anticoagulation identified N-terminal pro-B-type natriuretic peptide (NT-proBNP) and cardiac troponin as powerful predictors of stroke and other adverse outcomes.3–5 In addition, growth–differentiation factor 15 (GDF-15) is a potent predictor of bleeding and death in this population.6,7 Subsequently, NT-proBNP, cardiac troponin, and GDF-15 were incorporated in the ABC (age, biomarkers, and clinical history)-AF scores for biomarker-based assessment of the risks of stroke, bleeding, and death in patients with AF.8–12 However, despite its strong prognostic potential, Mendelian randomization studies suggest that a causal relationship of NT-proBNP with outcomes in patients with cardiovascular disease is unlikely.13,14
We have used multiplex screening of biomarkers and started a specific programme to identify novel biomarkers with incremental prognostic value.15–17 Moreover, we are exploring biomarkers that may reflect mechanisms of underlying pathophysiological processes. Angiopoietin-2 is an angiogenic growth factor that promotes endothelial inflammation and vascular remodelling and has been associated with adverse clinical outcomes in other conditions.18,19 In this study, we evaluated angiopoietin-2 as a novel candidate biomarker and explored and validated its associations with clinical characteristics and outcomes in two large, prospective cohorts of patients with AF, treated and not treated with oral anticoagulation.
Methods
The primary objective of this study was to assess the associations of plasma levels of circulating angiopoietin-2 with clinical outcomes in patients with AF. Secondary objectives included the evaluation of the associations of angiopoietin-2 with clinical characteristics and other biomarkers. We used an exploration cohort of patients on aspirin monotherapy (i.e. not treated with oral anticoagulation). Thereafter, we repeated the same analyses in a larger validation cohort of patients treated with oral anticoagulation. We only considered findings that were consistent across the two cohorts as validated associations between angiopoietin-2, clinical characteristics, and outcomes.
We measured levels of angiopoietin-2 in plasma samples obtained at baseline from patients with AF who were enrolled in three large clinical trials evaluating antithrombotic therapy, Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events (ACTIVE A), Apixaban Versus Acetylsalicylic Acid (ASA) to Prevent Stroke in Atrial Fibrillation Patients Who Have Failed or Are Unsuitable for Vitamin K Antagonist Treatment (AVERROES), and Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation (ARISTOTLE).20–22
Study population
Exploration cohort: patients with atrial fibrillation not treated with oral anticoagulation
This cohort included patients randomized to receive aspirin monotherapy in the ACTIVE A and AVERROES trials. ACTIVE A was a trial testing clopidogrel vs. placebo, in addition to aspirin, in patients with AF who were considered to be unsuitable for oral anticoagulation with a vitamin K antagonist.21 Between 2003 and 2006, a total of 7554 patients were enrolled. A plasma sample obtained at baseline was available for 1092 patients receiving aspirin, of which 956 had analysable levels of angiopoietin-2. AVERROES evaluated the factor Xa inhibitor apixaban vs. aspirin in patients with AF who were considered to be unsuitable for oral anticoagulation with a vitamin K antagonist.22 Between 2007 and 2009, a total of 5599 patients were enrolled. A plasma sample obtained at baseline was available for 2092 patients receiving aspirin, of which 2031 had analysable levels of angiopoietin-2. For this study evaluating angiopoietin-2, patients randomized to receive clopidogrel in addition to aspirin in ACTIVE A and apixaban in AVERROES were excluded.
Validation cohort: patients with atrial fibrillation treated with oral anticoagulation
This cohort included patients enrolled in ARISTOTLE, a randomized non-inferiority trial evaluating apixaban vs. warfarin.20 Between 2006 and 2010, a total of 18 201 patients with AF and at least one additional risk factor for stroke were enrolled. A plasma sample obtained at baseline was available for 14 980 patients. Angiopoietin-2 was analysable in 6546 and 6533 samples of patients randomized to warfarin and apixaban, respectively.
Biomarker analysis
Plasma levels of angiopoietin-2 were analysed using a prototype Elecsys electrochemiluminescence immunoassay developed by Roche Diagnostics. This assay employs a quantitative sandwich principle, with a first monoclonal antibody specifically binding angiopoietin-2 as capture antibody, and a ruthenylated second monoclonal antibody binding angiopoietin-2 as detection antibody. Recombinant angiopoietin-2 is used to normalize measurements across runs with a high grade of accuracy. All angiopoietin-2–related analyses were performed at the Uppsala Clinical Research Center Laboratory, Uppsala, Sweden. Coefficients of variation for angiopoietin-2 were 5.2% and 4.0% at concentrations of 0.98 and 3.40 ng/mL, respectively. The lower limit of detection was 0.03 ng/mL; 4 patients (0.1%) in the exploration cohort and 26 patients (0.2%) in the validation cohort had angiopoietin-2 concentrations at or below the lower limit of detection. Levels above 10 ng/mL were determined after sample dilution.
Plasma levels of several established biomarkers in patients with AF (NT-proBNP, cardiac troponin T, and GDF-15) were analysed using Elecsys electrochemiluminescence immunoassays (Roche Diagnostics) at the Clinical Research Laboratory and Biobank, Hamilton, Ontario, Canada (ACTIVE A and AVERROES),12 and at the Uppsala Clinical Research Center Laboratory, Uppsala, Sweden (ARISTOTLE).4,5,7
Outcomes
We assessed the associations of plasma levels of circulating angiopoietin-2 with subsequent hospitalization for heart failure, ischaemic stroke, cardiovascular and all-cause death, and major bleeding. Outcome definitions were those used in the ACTIVE A, AVERROES, and ARISTOTLE trials.20–22 In ACTIVE A, outcomes were centrally adjudicated, including hospitalization for heart failure. In AVERROES and ARISTOTLE, hospitalization and the main reason for hospitalization were reported by the local investigator, without central adjudication of source documents.
Statistical analysis
All analyses were performed in both the exploration and the validation cohort. Baseline characteristics were presented, overall and according to quartile groups of plasma angiopoietin-2. Categorical variables were presented as counts and percentages. Continuous variables were summarized as median (25th–75th percentile). Multivariable linear regression models were used to assess the association of clinical predictor variables with plasma levels of angiopoietin-2 (outcome variable). These models included cardiac rhythm on the baseline electrocardiogram (ECG) (AF vs. other rhythms) and type of AF (persistent/permanent vs. paroxysmal AF). Continuous variables were entered as restricted cubic splines with three knots placed at the 10th, 50th, and 90th percentiles. The coefficient of determination (R2) indicates the proportion of variance of angiopoietin-2 that was explained by the full set of predictor variables included in the model. A plot listing the partial R2, indicating the relative importance of each individual predictor variable, was constructed. The bivariate associations between log-transformed angiopoietin-2 and NT-proBNP, cardiac troponin T, and GDF-15 were assessed using the Spearman correlation coefficient. Median total follow-up duration was estimated by the Kaplan–Meier method as the time from randomization until the last day of follow-up while censoring for death. Outcome-specific follow-up duration was similarly estimated as the time from randomization until end of study while censoring for the specific event or death. Incidence rates were calculated as per 100 patient-years, and a corresponding 95% confidence interval (CI) was estimated using a gamma distribution. The Kaplan–Meier method was used to estimate the cumulative incidence rate of each outcome according to quartile groups of angiopoietin-2. An unadjusted smooth association curve was constructed to illustrate the crude association of angiopoietin-2 with the 1-year risk of each outcome. Cox proportional hazards models were used to assess the associations of plasma levels of angiopoietin-2 with clinical outcomes. Here, all biomarker concentrations were transformed using the natural logarithm. To account for potential non-linear associations, angiopoietin-2 was entered as a restricted cubic spline with three knots placed at the 10th, 50th, and 90th percentiles. An unadjusted analysis (Model 1) and several pre-specified adjusted analyses were employed. Multivariable models that adjusted for age, creatinine, and clinical variables (Model 2) and that adjusted for age, creatinine, clinical variables, and NT-proBNP (GDF-15 for major bleeding) (Model 3) are presented in this report. As angiopoietin-2 was allowed to be non-linear in the models, a single hazard ratio (HR) would not be representative of the association with clinical outcomes. Nevertheless, an example HR for a comparison of the third vs. first sample quartile of angiopoietin-2 with corresponding 95% CI was presented. Discrimination of each of the models was measured using Harrell’s c-index. For hospitalization for heart failure, a plot listing the relative importance of each predictor variable (defined here as the partial χ2 minus the predictor degrees of freedom) was constructed for Models 2 and 3.
All statistical analyses were performed using R, version 4.1.1, at the Uppsala Clinical Research Center, Uppsala, Sweden.
Ethical considerations
This study conforms to the principles outlined in the Declaration of Helsinki. All patients had provided written informed consent, and approval from local ethics committees had been obtained prior to initiation of ACTIVE A, AVERROES, and ARISTOTLE trials.20–22
Results
Patient characteristics and plasma levels of angiopoietin-2 at baseline
The analysis included 2987 patients in the exploration cohort treated with aspirin monotherapy (i.e. not treated with oral anticoagulation) and 13 079 patients in the validation cohort treated with oral anticoagulation (Table 1). Median (25th–75th percentile) plasma levels of angiopoietin-2 were 3.12 (2.15–4.88) and 3.08 (2.09–4.70) ng/mL, respectively.
Baseline characteristics
| Exploration cohort, n = 2987 | Validation cohort, n = 13 079 | |
|---|---|---|
| Antithrombotic regimen | Aspirin monotherapy | Apixaban or warfarin |
| Age (years) | 71.0 (63.0–77.0) | 70.0 (63.0–76.0) |
| Female sex | 41.8% (1250) | 36.2% (4730) |
| Height (cm) | 168.0 (160.0–175.3) [6] | 169.0 (161.0–176.0) [58] |
| Weight (kg) | 79.0 (68.0–92.0) [2] | 82.0 (70.0–95.7) [39] |
| BMI (kg/m²) | 27.8 (24.7–31.6) [6] | 28.6 (25.3–32.7) [60] |
| Current tobacco use/smoking | 7.3% (219) [1] | 7.8% (1023) [12] |
| Regular alcohol consumption | 26.2% (782) | 2.5% (331) |
| Type of atrial fibrillation | [2] | [3] |
| Paroxysmal | 28.7% (858) | 15.4% (2018) |
| Persistent/permanent | 71.3% (2127) | 84.6% (11 058) |
| Atrial fibrillation on baseline ECG | 66.9% (1998) [2] | 86.0% (11 210) [44] |
| Heart failure | 35.0% (1045) | 31.1% (4072) |
| Hypertension | 86.7% (2589) | 87.6% (11 462) |
| Diabetes mellitus | 19.8% (591) | 24.7% (3230) |
| Prior stroke or transient ischaemic attack | 12.8% (382) | 18.6% (2433) |
| Peripheral artery disease | 3.4% (103) | 4.9% (640) [1] |
| Coronary artery disease | 14.1% (422) [3] | Not reported |
| Prior myocardial infarction | 8.6% (258) [3] | 12.6% (1652) [1] |
| Biomarkers, in plasma | ||
| Angiopoietin-2 (ng/mL) | 3.12 (2.15–4.88) | 3.08 (2.09–4.70) |
| Cardiac troponin-T (ng/L) | 13.2 (9.1–20.2) | 10.8 (7.4–16.4) |
| Creatinine (µmol/L) | 89.0 (77.0–106.0) [146] | 89.3 (76.9–105.2) [5] |
| GDF-15 (ng/L) | 1593.0 (1094.0–2377.5) | 1358.0 (965.0–2025.0) |
| NT-proBNP (ng/L) | 706.2 (277.0–1384.4) | 702.0 (358.0–1235.5) |
| Exploration cohort, n = 2987 | Validation cohort, n = 13 079 | |
|---|---|---|
| Antithrombotic regimen | Aspirin monotherapy | Apixaban or warfarin |
| Age (years) | 71.0 (63.0–77.0) | 70.0 (63.0–76.0) |
| Female sex | 41.8% (1250) | 36.2% (4730) |
| Height (cm) | 168.0 (160.0–175.3) [6] | 169.0 (161.0–176.0) [58] |
| Weight (kg) | 79.0 (68.0–92.0) [2] | 82.0 (70.0–95.7) [39] |
| BMI (kg/m²) | 27.8 (24.7–31.6) [6] | 28.6 (25.3–32.7) [60] |
| Current tobacco use/smoking | 7.3% (219) [1] | 7.8% (1023) [12] |
| Regular alcohol consumption | 26.2% (782) | 2.5% (331) |
| Type of atrial fibrillation | [2] | [3] |
| Paroxysmal | 28.7% (858) | 15.4% (2018) |
| Persistent/permanent | 71.3% (2127) | 84.6% (11 058) |
| Atrial fibrillation on baseline ECG | 66.9% (1998) [2] | 86.0% (11 210) [44] |
| Heart failure | 35.0% (1045) | 31.1% (4072) |
| Hypertension | 86.7% (2589) | 87.6% (11 462) |
| Diabetes mellitus | 19.8% (591) | 24.7% (3230) |
| Prior stroke or transient ischaemic attack | 12.8% (382) | 18.6% (2433) |
| Peripheral artery disease | 3.4% (103) | 4.9% (640) [1] |
| Coronary artery disease | 14.1% (422) [3] | Not reported |
| Prior myocardial infarction | 8.6% (258) [3] | 12.6% (1652) [1] |
| Biomarkers, in plasma | ||
| Angiopoietin-2 (ng/mL) | 3.12 (2.15–4.88) | 3.08 (2.09–4.70) |
| Cardiac troponin-T (ng/L) | 13.2 (9.1–20.2) | 10.8 (7.4–16.4) |
| Creatinine (µmol/L) | 89.0 (77.0–106.0) [146] | 89.3 (76.9–105.2) [5] |
| GDF-15 (ng/L) | 1593.0 (1094.0–2377.5) | 1358.0 (965.0–2025.0) |
| NT-proBNP (ng/L) | 706.2 (277.0–1384.4) | 702.0 (358.0–1235.5) |
BMI, body mass index; ECG, electrocardiogram; GDF-15, growth–differentiation factor-15; NT-proBNP, N-terminal pro-B-type natriuretic peptide.
Reported values correspond to % (count) or median (25th–75th percentile). Numbers in [brackets] denote number of missing values.
Baseline characteristics
| Exploration cohort, n = 2987 | Validation cohort, n = 13 079 | |
|---|---|---|
| Antithrombotic regimen | Aspirin monotherapy | Apixaban or warfarin |
| Age (years) | 71.0 (63.0–77.0) | 70.0 (63.0–76.0) |
| Female sex | 41.8% (1250) | 36.2% (4730) |
| Height (cm) | 168.0 (160.0–175.3) [6] | 169.0 (161.0–176.0) [58] |
| Weight (kg) | 79.0 (68.0–92.0) [2] | 82.0 (70.0–95.7) [39] |
| BMI (kg/m²) | 27.8 (24.7–31.6) [6] | 28.6 (25.3–32.7) [60] |
| Current tobacco use/smoking | 7.3% (219) [1] | 7.8% (1023) [12] |
| Regular alcohol consumption | 26.2% (782) | 2.5% (331) |
| Type of atrial fibrillation | [2] | [3] |
| Paroxysmal | 28.7% (858) | 15.4% (2018) |
| Persistent/permanent | 71.3% (2127) | 84.6% (11 058) |
| Atrial fibrillation on baseline ECG | 66.9% (1998) [2] | 86.0% (11 210) [44] |
| Heart failure | 35.0% (1045) | 31.1% (4072) |
| Hypertension | 86.7% (2589) | 87.6% (11 462) |
| Diabetes mellitus | 19.8% (591) | 24.7% (3230) |
| Prior stroke or transient ischaemic attack | 12.8% (382) | 18.6% (2433) |
| Peripheral artery disease | 3.4% (103) | 4.9% (640) [1] |
| Coronary artery disease | 14.1% (422) [3] | Not reported |
| Prior myocardial infarction | 8.6% (258) [3] | 12.6% (1652) [1] |
| Biomarkers, in plasma | ||
| Angiopoietin-2 (ng/mL) | 3.12 (2.15–4.88) | 3.08 (2.09–4.70) |
| Cardiac troponin-T (ng/L) | 13.2 (9.1–20.2) | 10.8 (7.4–16.4) |
| Creatinine (µmol/L) | 89.0 (77.0–106.0) [146] | 89.3 (76.9–105.2) [5] |
| GDF-15 (ng/L) | 1593.0 (1094.0–2377.5) | 1358.0 (965.0–2025.0) |
| NT-proBNP (ng/L) | 706.2 (277.0–1384.4) | 702.0 (358.0–1235.5) |
| Exploration cohort, n = 2987 | Validation cohort, n = 13 079 | |
|---|---|---|
| Antithrombotic regimen | Aspirin monotherapy | Apixaban or warfarin |
| Age (years) | 71.0 (63.0–77.0) | 70.0 (63.0–76.0) |
| Female sex | 41.8% (1250) | 36.2% (4730) |
| Height (cm) | 168.0 (160.0–175.3) [6] | 169.0 (161.0–176.0) [58] |
| Weight (kg) | 79.0 (68.0–92.0) [2] | 82.0 (70.0–95.7) [39] |
| BMI (kg/m²) | 27.8 (24.7–31.6) [6] | 28.6 (25.3–32.7) [60] |
| Current tobacco use/smoking | 7.3% (219) [1] | 7.8% (1023) [12] |
| Regular alcohol consumption | 26.2% (782) | 2.5% (331) |
| Type of atrial fibrillation | [2] | [3] |
| Paroxysmal | 28.7% (858) | 15.4% (2018) |
| Persistent/permanent | 71.3% (2127) | 84.6% (11 058) |
| Atrial fibrillation on baseline ECG | 66.9% (1998) [2] | 86.0% (11 210) [44] |
| Heart failure | 35.0% (1045) | 31.1% (4072) |
| Hypertension | 86.7% (2589) | 87.6% (11 462) |
| Diabetes mellitus | 19.8% (591) | 24.7% (3230) |
| Prior stroke or transient ischaemic attack | 12.8% (382) | 18.6% (2433) |
| Peripheral artery disease | 3.4% (103) | 4.9% (640) [1] |
| Coronary artery disease | 14.1% (422) [3] | Not reported |
| Prior myocardial infarction | 8.6% (258) [3] | 12.6% (1652) [1] |
| Biomarkers, in plasma | ||
| Angiopoietin-2 (ng/mL) | 3.12 (2.15–4.88) | 3.08 (2.09–4.70) |
| Cardiac troponin-T (ng/L) | 13.2 (9.1–20.2) | 10.8 (7.4–16.4) |
| Creatinine (µmol/L) | 89.0 (77.0–106.0) [146] | 89.3 (76.9–105.2) [5] |
| GDF-15 (ng/L) | 1593.0 (1094.0–2377.5) | 1358.0 (965.0–2025.0) |
| NT-proBNP (ng/L) | 706.2 (277.0–1384.4) | 702.0 (358.0–1235.5) |
BMI, body mass index; ECG, electrocardiogram; GDF-15, growth–differentiation factor-15; NT-proBNP, N-terminal pro-B-type natriuretic peptide.
Reported values correspond to % (count) or median (25th–75th percentile). Numbers in [brackets] denote number of missing values.
Baseline characteristics of patients in the two cohorts according to quartile groups of plasma levels of angiopoietin-2 are shown in Supplementary material online, Table S1. In the exploration cohort, angiopoietin-2, in multivariable analyses, was associated with AF rhythm on the baseline ECG, advanced age, history of heart failure, persistent/permanent AF, female sex, tobacco use, higher body mass index (BMI) (all P < 0.001), creatinine (P = 0.001), and diabetes (P = 0.01) (see Supplementary material online, Figure S1A). In the validation cohort, angiopoietin-2 was similarly associated with female sex, history of heart failure, higher BMI, advanced age, persistent/permanent AF, AF rhythm on the baseline ECG, creatinine, smoking, and diabetes (all P < 0.001), but also prior myocardial infarction, peripheral artery disease (both P < 0.001), alcohol consumption (P = 0.007), and absence of hypertension (P = 0.04) (see Supplementary material online, Figure S1B).
In both the exploration and the validation cohort, plasma levels of angiopoietin-2 showed significant correlation with those of NT-proBNP (r = 0.61 and r = 0.51), cardiac troponin (r = 0.38 and r = 0.26), and GDF-15 (r = 0.35 and r = 0.25) (all P < 0.001).
Association between angiopoietin-2 and hospitalization for heart failure
In the exploration cohort, the incidence of a first hospitalization for heart failure was 4.1 (95% CI 3.6–4.7) per 100 patient-years. The incidence was increased with higher plasma levels of angiopoietin-2 (Figures 1A and 2A). Adjusting for age, renal function, and clinical characteristics (Model 2, c-index 0.79, 95% CI 0.75–0.82), plasma levels of angiopoietin-2 were associated with hospitalization for heart failure (third vs. first quartile, HR 1.74, 95% CI 1.26–2.41, P < 0.001) (Table 2). The observed association remained statistically significant after additionally adjusting the model for the established biomarker NT-proBNP (HR 1.40, 95% CI 1.02–1.92, P = 0.001) (Model 3). Improvement in c-index through addition of angiopoietin-2 was 0.02 and 0.01 for multivariable Models 2 and 3, respectively. Variable contribution plots showed that angiopoietin-2 was the second most important predictor (after a history of heart failure) of hospitalization for heart failure in Model 2 and ranked fourth (after a history of heart failure, BMI, and NT-proBNP) in Model 3 (Figure 3A).
Hospitalization for heart failure. Kaplan–Meier curves according to quartile groups of angiopoietin-2. The incidence of hospitalization for heart failure was increased with higher plasma angiopoietin-2 in the exploration and in the validation cohort. (A) Exploration cohort. (B) Validation cohort. HF, heart failure.
Hospitalization for heart failure. Unadjusted smooth association curves according to plasma levels of angiopoietin-2. (A) Exploration cohort. (B) Validation cohort. ANG2, angiopoietin-2; HF, heart failure.
Variable importance plots. Plasma angiopoietin-2 at baseline was among the most important predictors for subsequent hospitalization for heart failure. (A) Exploration cohort. (B) Validation cohort. AF, atrial fibrillation; ANG2, angiopoietin-2; BMI, body mass index; CAD, coronary artery disease; HF, heart failure; MI, myocardial infarction; PAD, peripheral artery disease; TIA, transient ischaemic attack.
Association of plasma levels of angiopoietin-2 with hospitalization for heart failure
| Exploration cohort | Validation cohort | |||||
|---|---|---|---|---|---|---|
| Model | HR (95% CI) | P | C-index (95% CI) | HR (95% CI) | P | C-index (95% CI) |
| Model 1 | 2.67 (2.00–3.56) | <0.001 | 0.72 (0.68–0.76) | 1.85 (1.61–2.13) | <0.001 | 0.70 (0.67–0.72) |
| Model 2 | 1.74 (1.26–2.41) | <0.001 | 0.79 (0.75–0.82) | 1.58 (1.37–1.82) | <0.001 | 0.76 (0.74–0.78) |
| Model 3 | 1.40 (1.02–1.92) | 0.001 | 0.80 (0.77–0.83) | 1.20 (1.05–1.37) | <0.001 | 0.79 (0.77–0.81) |
| Exploration cohort | Validation cohort | |||||
|---|---|---|---|---|---|---|
| Model | HR (95% CI) | P | C-index (95% CI) | HR (95% CI) | P | C-index (95% CI) |
| Model 1 | 2.67 (2.00–3.56) | <0.001 | 0.72 (0.68–0.76) | 1.85 (1.61–2.13) | <0.001 | 0.70 (0.67–0.72) |
| Model 2 | 1.74 (1.26–2.41) | <0.001 | 0.79 (0.75–0.82) | 1.58 (1.37–1.82) | <0.001 | 0.76 (0.74–0.78) |
| Model 3 | 1.40 (1.02–1.92) | 0.001 | 0.80 (0.77–0.83) | 1.20 (1.05–1.37) | <0.001 | 0.79 (0.77–0.81) |
Model 1 Unadjusted.
Model 2 Adjusted for age, creatinine, sex, body mass index, smoking status, regular alcohol consumption, type of AF, cardiac rhythm at baseline, heart failure, hypertension, diabetes, prior stroke or transient ischaemic attack, peripheral artery disease, coronary artery disease (only in the exploration cohort), and prior myocardial infarction.
Model 3 Model 2 + additionally adjusted for NT-proBNP.
Table shows an example HR of the third vs. first quartile of angiopoietin-2. In the exploration cohort, all multivariable models were additionally adjusted for study (ACTIVE A or AVERROES). In the validation cohort, all multivariable models were additionally adjusted for randomized treatment (warfarin or apixaban).
CI, confidence interval; HR, hazard ratio; NT-proBNP, N-terminal pro-B-type natriuretic peptide.
Association of plasma levels of angiopoietin-2 with hospitalization for heart failure
| Exploration cohort | Validation cohort | |||||
|---|---|---|---|---|---|---|
| Model | HR (95% CI) | P | C-index (95% CI) | HR (95% CI) | P | C-index (95% CI) |
| Model 1 | 2.67 (2.00–3.56) | <0.001 | 0.72 (0.68–0.76) | 1.85 (1.61–2.13) | <0.001 | 0.70 (0.67–0.72) |
| Model 2 | 1.74 (1.26–2.41) | <0.001 | 0.79 (0.75–0.82) | 1.58 (1.37–1.82) | <0.001 | 0.76 (0.74–0.78) |
| Model 3 | 1.40 (1.02–1.92) | 0.001 | 0.80 (0.77–0.83) | 1.20 (1.05–1.37) | <0.001 | 0.79 (0.77–0.81) |
| Exploration cohort | Validation cohort | |||||
|---|---|---|---|---|---|---|
| Model | HR (95% CI) | P | C-index (95% CI) | HR (95% CI) | P | C-index (95% CI) |
| Model 1 | 2.67 (2.00–3.56) | <0.001 | 0.72 (0.68–0.76) | 1.85 (1.61–2.13) | <0.001 | 0.70 (0.67–0.72) |
| Model 2 | 1.74 (1.26–2.41) | <0.001 | 0.79 (0.75–0.82) | 1.58 (1.37–1.82) | <0.001 | 0.76 (0.74–0.78) |
| Model 3 | 1.40 (1.02–1.92) | 0.001 | 0.80 (0.77–0.83) | 1.20 (1.05–1.37) | <0.001 | 0.79 (0.77–0.81) |
Model 1 Unadjusted.
Model 2 Adjusted for age, creatinine, sex, body mass index, smoking status, regular alcohol consumption, type of AF, cardiac rhythm at baseline, heart failure, hypertension, diabetes, prior stroke or transient ischaemic attack, peripheral artery disease, coronary artery disease (only in the exploration cohort), and prior myocardial infarction.
Model 3 Model 2 + additionally adjusted for NT-proBNP.
Table shows an example HR of the third vs. first quartile of angiopoietin-2. In the exploration cohort, all multivariable models were additionally adjusted for study (ACTIVE A or AVERROES). In the validation cohort, all multivariable models were additionally adjusted for randomized treatment (warfarin or apixaban).
CI, confidence interval; HR, hazard ratio; NT-proBNP, N-terminal pro-B-type natriuretic peptide.
In the validation cohort, the incidence of a first hospitalization for heart failure was 2.0 (95% CI 1.8–2.2) per 100 patient-years. Corresponding Kaplan–Meier estimates according to quartile groups of angiopoietin-2 (Figure 1B) and the unadjusted smooth association of the biomarker with the 1-year risk of hospitalization for heart failure (Figure 2B) were consistent with the findings from the exploration cohort. In Model 2 (c-index 0.76, 95% CI 0.74–0.78), plasma levels of angiopoietin-2 were associated with hospitalization for heart failure (HR 1.58, 95% CI 1.37–1.82, P < 0.001), which remained statistically significant after additionally adjusting for NT-proBNP in Model 3 (HR 1.20, 95% CI 1.05–1.37, P < 0.001) (Table 2). Improvement in c-index through addition of angiopoietin-2 was 0.04 and 0.01 for multivariable Models 2 and 3, respectively. Results suggested a similar importance of angiopoietin-2 in the multivariable models in the validation cohort, ranking first in Model 2 and fourth in Model 3 (after NT-proBNP, history of heart failure, and prior myocardial infarction), respectively (Figure 3B).
Associations between angiopoietin-2 and other outcomes
Ischaemic stroke
Although plasma levels of angiopoietin-2 were associated with ischaemic stroke in unadjusted analyses and after adjusting for age, renal function, and clinical characteristics in the exploration cohort, this association no longer remained statistically significant when additionally adjusting for NT-proBNP (P = 0.23). Consistent findings were observed in the validation cohort (P = 0.57, Supplementary material online, Table S2 and Figure S2).
Cardiovascular and all-cause death
In the exploration cohort, plasma levels of angiopoietin-2 were associated with cardiovascular and all-cause death in unadjusted analyses and after adjusting for age, renal function, and clinical characteristics. However, these associations remained no longer statistically significant when additionally adjusting for NT-proBNP (P = 0.16 and P = 0.35, respectively). In the validation cohort, unadjusted analyses and models that adjusted for age, renal function, and clinical characteristics yielded consistent findings. Furthermore, in this larger cohort of patients treated with oral anticoagulation, the association between angiopoietin-2 and fatal outcomes remained statistically significant after additionally adjusting for NT-proBNP (P = 0.01 and <0.001, respectively, Supplementary material online, Table S2 and Figures S3 and S4).
Major bleeding
In the exploration cohort, plasma levels of angiopoietin-2 did not show a statistically significant association with major bleeding in adjusted analyses. Analyses in the validation cohort suggested an association of angiopoietin-2 with major bleeding after adjusting for age, renal function, and clinical characteristics (P = 0.04). After additionally adjusting for the established biomarker GDF-15, the association no longer remained statistically significant (P = 0.13) (see Supplementary material online, Table S2 and Figure S5).
Discussion
We investigated plasma levels of angiopoietin-2, a novel candidate biomarker of endothelial inflammation and vascular remodelling, in patients with AF. In both an exploration and a validation cohort of patients treated and not treated with oral anticoagulation, plasma levels of angiopoietin-2 were associated with AF on the baseline ECG and persistent/permanent AF, increasing age, history of heart failure, female sex, tobacco use, increasing BMI, renal dysfunction, diabetes, and plasma levels of the established biomarker NT-proBNP. After adjusting for these and other clinical characteristics, a significant association between angiopoietin-2 and subsequent hospitalization for heart failure remained, which was consistent across both cohorts. Therefore, angiopoietin-2 provided incremental prognostic value to clinical risk factors and NT-proBNP.
Constant activation of the endothelial tyrosine kinase receptor Tie2 promotes vascular stabilization, maintaining anticoagulant properties and suppressing pro-inflammatory pathways.19,23,24 Angiopoietin-1 is a Tie2 agonist, promoting phosphorylation of the receptor that then activates downstream pathways. In contrast, previous work suggests a complex, context-dependent role of angiopoietin-2, which mainly acts as a Tie2 antagonist, particularly under pathological conditions.18,19,25,26 By competing with angiopoietin-1, angiopoietin-2 decreases the extent of Tie2 phosphorylation, making the endothelium more responsive to inflammatory stimuli.18,25 Increased plasma levels of angiopoietin-2 have been demonstrated in patients with several solid tumours and diseases involving inflammation, such as acute respiratory distress syndrome or sepsis.19,27,28 Similarly, angiopoietin-2 has been associated with worse clinical outcomes in cohorts of patients with a variety of cardiovascular diseases,29–34 but not in large, well-characterized cohorts of patients with AF. Furthermore, previous observational studies have linked angiopoietin-2 with heart failure events but were smaller and not limited to patients with AF.35,36
In the present study, plasma levels of angiopoietin-2 were strongly and independently associated with subsequent hospitalization for heart failure, even when adjusting for plasma levels of the established prognostic biomarker NT-proBNP. While causality cannot be determined from this type of observational study, our findings are consistent with the hypothesis that increased expression of angiopoietin-2 and inactivation of the stabilizing angiopoietin-1/Tie2 pathway might amplify pro-inflammatory processes and contribute to heart failure in individuals with AF. Previous studies have shown that systemic markers of inflammation, such as C-reactive protein and interleukin-6, are associated with subsequent hospitalization for heart failure in patients with AF.37,38 Whether angiopoietin-2 is associated with hospitalization for heart failure beyond systemic biomarkers of inflammation or whether other associated factors (like plasma levels of the Tie2 agonist angiopoietin-1) are more strongly associated needs evaluation in future studies. Our results also allow speculations about a role of pro-inflammatory processes promoting AF that may result in worsening of cardiac output. Furthermore, inflammatory signalling in atrial cardiomyocytes has been linked to incident AF as well as AF following cardiac surgery.39,40 Future studies should assess a potential involvement of angiopoietin-2 in these settings.
One potential clinical application of angiopoietin-2 may be the early identification of patients at particularly high risk for incident or recurrent hospitalization for heart failure. Lastly, our findings raise the questions as to a role of inflammation that has implications beyond risk stratification alone.37,41 The angiopoietin-Tie2 axis has emerged as a target for novel therapies in other fields of medicine, including cancer and diabetic retinopathy.42,43 Randomized trials may be warranted to determine whether interventions aimed at reducing inflammation by reducing the expression of angiopoietin-2, or, activation of Tie2, thereby amplifying the stabilizing angiopoietin-1/Tie2 pathway, could improve outcomes in patients with AF.
In both cohorts, plasma levels of angiopoietin-2 were associated with subsequent ischaemic stroke, as well as cardiovascular and all-cause death after adjusting for age, kidney function, and clinical characteristics. However, the association with ischaemic stroke was no longer statistically significant when also adjusting for NT-proBNP, suggesting that angiopoietin-2 may be of lesser importance in the pathophysiology of AF-related stroke. In the exploration cohort, plasma levels of angiopoietin-2 were not associated with cardiovascular and all-cause death in the full models adjusted for NT-proBNP. However, in the larger validation cohort, the association between angiopoietin-2 and fatal outcomes persisted even after additional adjustment for this established biomarker. The observed associations of angiopoietin-2 with fatal outcomes in the validation cohort, if confirmed by other studies, may add to findings from other research that has associated biomarkers of inflammation with cardiovascular death.37,44–46 Heart failure, which seems to be closely linked to inflammatory processes, potentially with the influence of the angiopoietin-Tie2 axis, is a major contributor to mortality risk in patients with AF.47,48 At the same time, mechanisms of death in patients with AF are complex and variable, and angiopoietin-2 may be a risk marker that is not specific to AF.28,29,31 Finally, although the angiopoietin-Tie2 system has been characterized as a main regulator of vascular stability,19 angiopoietin-2 did not show relevant associations with major bleeding. This suggests that bleeding in patients with AF, irrespective of the use of oral anticoagulation, is promoted by other mechanisms.
Plasma levels of circulating angiopoietin-2 showed positive associations with several clinical characteristics, including AF on the baseline ECG, increasing age, history of heart failure, persistent/permanent AF, female sex, tobacco use, increasing BMI, renal dysfunction, and diabetes. While there was some variation in the exact order of importance, 4 out of the 5 variables that showed the strongest association with plasma levels of angiopoietin-2 were the same in the exploration and validation cohort.
In both cohorts of patients treated and not treated with oral anticoagulation, plasma levels of angiopoietin-2 were strongly correlated with those of NT-proBNP, a powerful risk marker in cardiovascular disease, including AF.5,49,50 After additional adjustment for NT-proBNP, all observed associations of angiopoietin-2 with clinical outcomes were markedly attenuated. However, Mendelian randomization studies suggest that although NT-proBNP is a strong risk marker for adverse outcomes in AF and atherothrombotic disease, the association is unlikely to be causal.13,14 It might be hypothesized that one contributing cause of the prognostic value of NT-proBNP is its association with angiopoietin-2 and inactivation of the stabilizing angiopoietin-1/Tie2 pathway.
Strengths and limitations
The main strengths of this study include the evaluation of a novel candidate biomarker in two large cohorts of well-characterized patients with AF, adjusted for the established biomarker NT-proBNP, and the consistency of the main findings irrespective of oral anticoagulation use. There are also several limitations. First, although we performed rigorous adjustment for known prognostic factors in patients with AF, the observed associations of angiopoietin-2 with clinical outcomes in several independent trials and different treatment settings may be subject to residual confounding. Second, due to the observational study design, we are unable to prove that the relationships are causal. Third, the angiopoietin-Tie2 system is responsive to external and transient stimuli, especially through the rapid release of angiopoietin-2 from endothelial cells.51 In this study, we did not assess changes in plasma levels of angiopoietin-2 over time.
Conclusions
In patients with AF, plasma levels of angiopoietin-2 were independently associated with subsequent hospitalization for heart failure and provided incremental prognostic value to clinical characteristics and NT-proBNP. A potential pathophysiologic role of angiopoietin-2–associated pathways and the value of angiopoietin-2 plasma levels for decision support in this population warrant further exploration.
Supplementary material
Supplementary material is available at Europace online.
Funding
The ACTIVE A trial was supported by grants from Sanofi-Aventis and Bristol-Myers Squibb. The AVERROES trial was funded by Bristol-Myers Squibb and Pfizer. Biomarker analyses were supported by Roche Diagnostics. The ARISTOTLE trial was funded by Bristol-Myers Squibb and Pfizer. Z.H. receives research support from the Swedish Society for Medical Research (S17–0133), the Swedish Heart-Lung Foundation (20200722), and Uppsala University Hospital, Sweden.
Conflict of interest: A.P.B. has nothing to disclose. Z.H. reports lecture fees from Boehringer Ingelheim, Bristol-Myers Squibb, and Pfizer; consulting fees from Boehringer Ingelheim, Bristol-Myers Squibb, Merck Sharp & Dohme, and Pfizer; and fees paid to his institution for advisory boards and lectures from Roche Diagnostics. J.L. reports an institutional research grant from Roche Diagnostics. S.J.C. reports institutional research grants and honoraria from Boehringer-Ingelheim, Portola/Alexion Pharmaceuticals, Bristol-Myers Squibb/Pfizer, Bayer, and Daiichi Sankyo. J.W.E. reports institutional research grants and honoraria from AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb/Pfizer, Daiichi-Sankyo, Eli Lilly, Glaxo Smith Kline, Janssen, and Sanofi. P.K. reports employment with Roche Diagnostics. A.Z. reports employment with Roche Diagnostics. J.O. reports institutional research grants and fees paid to his institution for advisory boards and lectures from Roche Diagnostics and fees paid to his institution for advisory boards, study steering committees, safety committees, and/or lectures from AstraZeneca, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Daiichi-Sankyo, Novartis, Portola, Pfizer, and Sanofi. A.S. reports institutional research grants from AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb/Pfizer, GlaxoSmithKline, Roche Diagnostics; consultancy fees from Olink Proteomics. C.B.G. reports personal fees from Bayer and Boston Scientific; grants and personal fees from Boehringer Ingelheim, Bristol Myers Squibb, Janssen, and Pfizer; and grants from Daiichi-Sankyo during the conduct of the study; personal fees from AbbVie, Espero, Medscape, Medtronic Inc., Merck, the National Institutes of Health, Novo Nordisk, Roche, Rho Pharmaceuticals, CeleCor, Correvio, Philips, Abiomed, and Anthos Therapeutics; grants from Akros, AstraZeneca, the US Food and Drug Administration, Glaxo Smith Kline, Medtronic Foundation, and Apple; and grants and personal fees from Novartis and The Medicines Company outside the submitted work. R.D.L. reports personal fees from Bayer, Boehringer Ingelheim, Bristol Myers Squibb, Daiichi-Sankyo, GlaxoSmithKline, Medtronic, Merck, Pfizer, Portola, and Sanofi and grants from Bristol Myers Squibb, GlaxoSmithKline, Medtronic, Pfizer, and Sanofi outside the submitted work. J.H.A. reports grants from Bayer and XaTek; grants and personal fees from Bristol Myers Squibb and CryoLife; and personal fees from Janssen, Pfizer, and Portola outside the submitted work. L.W. reports institutional research grants from AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb/Pfizer, GlaxoSmithKline, Roche Diagnostics, Merck & Co; consulting fees from Abbott and holds two patents involving GDF-15 licensed to Roche Diagnostics (EP2047275B1 and US8951742B2).
Data availability
Data and methods used in the analysis may be made available for other researchers upon reasonable request to the corresponding author.
