GDF‐15 at admission predicts cardiovascular death, heart failure, and bleeding outcomes in patients with CAD

Abstract

Aims

We aimed to investigate the independent associations between growth differentiation factor 15 (GDF‐15) level at admission and cardiovascular (CV) death, thrombotic events, heart failure (HF), and bleeding outcomes in patients with coronary artery disease (CAD).

Methods and results

We measured the plasma concentrations of GDF‐15 centrally in patients from the BIomarker‐based Prognostic Assessment for patients with Stable angina and acute coronary Syndrome (BIPass) registry, which consecutively enrolled patients with CAD from November 2017 to September 2019 at five tertiary hospitals in China. The outcomes included CV death, thrombotic events [myocardial infarction (MI) and ischaemic stroke], HF events [acute HF during hospitalization and hospitalization for HF post‐discharge (A/H HF) and cardiogenic shock], and bleeding outcomes [non‐coronary artery bypass grafting‐related major bleeding and clinically significant bleeding (CSB)] during the 12 month follow‐up period after hospitalization. Among 6322 patients with CAD {65.4% male, median age 63.7 [inter‐quartile range (IQR)] 56.0–70.1 years}, the median concentration of plasma GDF‐15 at admission was 1091 (IQR 790.5–1635.0) ng/L. Higher concentrations of GDF‐15 were associated with an increased risk of CV death [hazard ratio (HR) 1.98, 95% confidence interval (CI) 1.35–2.88, P < 0.001], A/H HF (HR 2.69, 95% CI 1.92–3.77, P < 0.001), cardiogenic shock (HR 1.46, 95% CI 1.04–2.05, P = 0.029), and CSB (HR 1.48, 95% CI 1.22–1.79, P < 0.001), but not for MI or stroke, after adjusting for clinical risk factors and prognostic biomarkers. Adding GDF‐15 to the model with risk factors and biomarkers improved the net reclassification for CV death, A/H HF, cardiogenic shock, and CSB.

Conclusions

In patients with CAD, admission levels of GDF‐15 were associated with an increased 1 year risk of CV death, HF, and bleeding outcomes, but not with thrombotic events. GDF‐15 may be a prognostic biomarker for CV death, HF, and bleeding outcomes and could be used to refine the risk assessment of these specific clinical outcomes.

Trial Registration: ClinicalTrials.gov Identifier: NCT04044066

Introduction

Coronary artery disease (CAD) is a major cause of morbidity and mortality worldwide. However, newer drug‐eluting stents and novel antiplatelet medications have been commonly used for its treatment. ¹ Thrombotic events such as spontaneous myocardial infarction (MI), ischaemic stroke, and heart failure (HF) events following myocardial ischaemia or infarction are the primary clinical intermediates linking CAD with cardiovascular (CV) death. Specifically, the incidence of HF after CAD has increased steadily in recent decades owing to the current therapy‐enhanced acute survival rate. ² , ³ In addition, antiplatelet medications have increased the risk of bleeding complications, which could adversely affect mortality. ⁴ Therefore, recognizing the risk and predictors of specific outcomes that link CAD may be useful in optimizing the precise management of patients.

Circulating biomarkers reflecting different aspects of the orchestral pathways involved in the progression of CAD have been sought, with the concept that these markers have prognostic value. ⁵ Growth differentiation factor 15 (GDF‐15) is a transforming growth factor β cytokine superfamily member, which provides prognostic information in patients with CV diseases. ⁶ Although circulating levels of GDF‐15 are substantially elevated under acute and severe milieus, such as ischaemia/reperfusion and myocardial stretch, they are also elevated under low‐level, constitutive stresses associated with tissue ageing and frailty. ⁷ , ⁸ GDF‐15 is not condition specific and might be a prognostic marker for multiple outcomes.

Several studies have assessed the associations between GDF‐15 at hospital admission and clinical outcomes with inconsistent results in the setting of CAD. ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ A positive association between GDF‐15 and MI was observed in some studies but not in others. ¹² , ¹³ The association between GDF‐15 and bleeding outcomes was evaluated in the PLATelet inhibition and patient Outcomes (PLATO) trial, which enrolled patients with acute coronary syndrome (ACS), a serious subcategory of CAD. ¹³ However, the findings from PLATO have not been confirmed in a wider spectrum of CAD patients. The prognostic value of circulating GDF‐15 levels for individual clinical outcomes in patients with CAD remains poorly defined. This study aimed to prospectively investigate the independent associations between admission GDF‐15 levels and 1 year CV death, thrombotic events, HF events, and bleeding outcomes in a large Chinese population of patients with CAD.

Methods

Study population

This study was part of a BIomarker‐based Prognostic Assessment for patients with Stable angina and acute coronary Syndrome (BIPass) project (https://clinicaltrials.gov/, NCT04044066). The BIPass study consecutively recruited patients with CAD admitted to five tertiary teaching and comprehensive hospitals in northern China (Supporting Information, Table S1 and Figure S1 ). The inclusion and exclusion criteria are listed in Supporting Information, Table S2 . The BIPass study complied with the Declaration of Helsinki and was approved by the institutional review boards of all participating sites. Written informed consent was obtained from all patients.

Clinical outcome events

Thrombotic events were defined as new/recurrent MI or ischaemic stroke. HF events included acute HF during hospitalization and hospitalization for HF post‐discharge (A/H HF) and cardiogenic shock. Bleeding events included non‐coronary artery bypass grafting (CABG)‐related major bleeding (major bleeding) and non‐CABG‐related clinically significant bleeding (CSB). Definitions of the endpoints are provided in Supporting Information, Table S3 . Clinical outcome events were recorded by trained research assistants at each participating site at discharge and at 1, 6, and 12 months via follow‐up telephone calls. Events were adjudicated by an independent clinical event committee consisting of experienced cardiologists based on source medical documents.

Measurement of plasma growth differentiation factor 15 concentrations

Fasting blood samples were collected in ethylene diamine tetraacetic acid tubes upon admission and immediately centrifuged. Plasma samples were stored at −80°C until biomarker analyses were performed centrally at the Shandong Provincial Clinical Research Center for Emergency and Critical Care Medicine (Jinan, China). GDF‐15 concentrations were determined using an R&D system for detection with a monoclonal antibody sandwich assay on a Luminex 200 platform (Luminex, Austin, TX, USA). The assay characteristics have been described previously. ¹⁴

Statistical analysis

Continuous variables were compared using the Mann–Whitney U tests, and categorical variables were compared using χ ² or Fisher's exact tests, as appropriate. Natural logarithmic (log) transformations were performed for continuous variables with skewed distributions [GDF‐15, high‐sensitivity cardiac troponin T (hs‐cTnT), N‐terminal pro‐B‐type natriuretic peptide (NT‐proBNP), and cystatin C levels]. Associations between serum GDF‐15 levels and baseline characteristics, CV risk factors, and other biomarkers were assessed using univariate and multivariate linear regression models. The functional relationship between GDF‐15 levels and clinical events was explored using restricted cubic splines with four knots at the 5th, 35th, 65th, and 95th percentiles of GDF‐15 levels. The spline graph model indicated an overall linear increase in individual clinical outcomes with increasing GDF‐15 levels.

The association with each clinical outcome was analysed using Cox proportional hazards models. Hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated per unit increase in log‐transformed GDF‐15 as a continuous variable. Two models were used, and the adjusted covariates for each model are shown in Supporting Information, Table S4 . Subgroup analyses were performed by age (<65 and ≥65 years), sex (male and female), history of diabetes (with and without diabetes), and CAD subtype [ACS and stable angina (SA)] to assess the associations between GDF‐15 and outcomes. Interactions between GDF‐15 levels and subgroup variables on different outcomes were tested by including their interaction terms in the Cox model. The association between GDF‐15 levels and clinical outcomes at different time intervals (in‐hospital and post‐discharge follow‐ups) was also assessed.

A prediction model was established using stepwise multivariable Cox regression analysis of CV risk factors and prognostic biomarkers. We compared the conventional model (including the CV risk factors, hs‐cTnT, NT‐proBNP, and cystatin C, Model 1) and the new model (including the above variables and GDF‐15, Model 2) by net reclassification improvement (NRI) and integrated discrimination improvement. ¹⁵ , ¹⁶ A two‐tailed P‐value < 0.05 was considered statistically significant. All analyses were performed using R statistical software (Version 4.0.2; R Foundation for Statistical Computing, Vienna, Austria).

The expanded methods are provided in the supporting information.

Results

Baseline characteristics

The BIPass study enrolled 8152 patients with CAD between November 2017 and September 2019. After excluding 1830 patients without GDF‐15 data, 6322 patients were included in the study. The median age of the patients was 63.7 [inter‐quartile range (IQR) 56.0–70.1] years, and 4134 (65.4%) were male. Among the included patients, 5215 (82.5%) were admitted with ACS, 1107 (17.5%) with SA, 4007 (63.4%) with hypertension, 1958 (31.0%) with diabetes mellitus, and 768 (12.2%) with a previous MI. The concentrations of plasma GDF‐15 at admission ranged from 361 to 6769 ng/L, with a median of 1091 (IQR 790.5–1635.0) ng/L (Table  1 ). In the multivariate linear analyses, the major independent drivers of the clinical variables for elevated GDF‐15 levels were age, male sex, current smoking, diabetes mellitus, renal dysfunction, and previous percutaneous coronary intervention or CABG. Plasma GDF‐15 levels were associated with most biochemical indicators, including white blood cell counts and haemoglobin, hs‐cTnT, cystatin C, and NT‐proBNP levels (Supporting Information, Table S5 ).

Table 1.

Baseline characteristics

Variable	Total patients	GDF‐15 tertile			P‐value
		Q1 (≤887.7 ng/L)	Q2 (887.7–1395 ng/L)	Q3 (>1395 ng/L)
N	6322	2108	2108	2106
Demographics
Age, years	63.7 [56.0, 70.1]	57.7 [51.9, 64.1]	64.4 [58.0, 69.6]	68 [61.6, 74.9]	<0.001
Sex, male, n (%)	4134 (65.4)	1307 (62.0)	1406 (66.7)	1421 (67.5)	<0.001
BMI, kg/m2	25.7 [23.8, 27.4]	25.7 [24.1, 27.7]	25.7 [23.7, 27.4]	25.7 [23.6, 27.3]	0.002
Risk factors and medical history
Current smoking, n (%)	3190 (50.5)	968 (45.9)	1097 (52.0)	1125 (53.4)	<0.001
Hypertension, n (%)	4007 (63.4)	1203 (57.1)	1316 (62.4)	1488 (70.7)	<0.001
Diabetes mellitus, n (%)	1958 (31.0)	380 (18.0)	622 (29.5)	956 (45.4)	<0.001
Previous MI, n (%)	768 (12.2)	197 (9.3)	231 (11.0)	340 (16.2)	<0.001
Congestive heart failure, n (%)	45 (0.7)	4 (0.2)	4 (0.2)	37 (1.8)	<0.001
Renal dysfunction, n (%)	89 (1.4)	3 (0.1)	7 (0.3)	79 (3.8)	<0.001
Peripheral arterial disease, n (%)	204 (3.2)	38 (1.8)	65 (3.1)	101 (4.8)	<0.001
Previous PCI or CABG, n (%)	1438 (22.8)	373 (17.7)	465 (22.1)	600 (28.5)	<0.001
Previous stroke, n (%)	829 (13.1)	183 (8.7)	290 (13.8)	356 (16.9)	<0.001
Previous haemorrhage, n (%)	39 (0.6)	13 (0.6)	13 (0.6)	13 (0.6)	0.946
Anticoagulants, n (%)	73 (1.2)	17 (0.8)	27 (1.3)	29 (1.4)	0.005
eGFR, mL/min/1.73 m2	98.9 [83.7, 114.6]	106.0 [93.4, 121.0]	99.5 [85.7, 115.2]	88.5 [71.0, 105.6]	<0.001
eGFR category					<0.001
eGFR < 60 mL/min/1.73 m2	323 (5.3)	12 (0.6)	28 (1.4)	283 (13.9)
eGFR 60–89 mL/min/1.73 m2	1778 (29.2)	376 (18.5)	630 (31.2)	772 (38.0)
eGFR ≥ 90 mL/min/1.73 m2	3981 (65.5)	1649 (81.0)	1358 (67.4)	974 (48.0)
Admission characteristics
Heart rate, beats/min	72 [65, 80]	71 [64, 78]	72 [65, 80]	73 [66, 81]	<0.001
Systolic blood pressure, mmHg	134 [121, 146]	132 [121.0, 144.2]	134 [122, 147]	134 [121, 148]	0.007
ECG ST‐T changes, n (%)	3165 (51.3)	977 (47.0)	1052 (51.2)	1136 (55.7)	<0.001
Disease classification
ACS, n (%)	5215 (82.5)	1712 (81.2)	1721 (81.6)	1782 (84.6)	0.007
Stable angina, n (%)	1107 (17.5)	396 (18.8)	387 (18.4)	324 (15.4)
In‐hospital medications
Aspirin	6185 (97.9)	2064 (98.0)	2068 (98.2)	2053 (97.6)	0.366
Clopidogrel	4722 (74.8)	1558 (73.9)	1572 (74.6)	1592 (75.7)	0.433
Ticagrelor	1298 (20.6)	437 (20.8)	461 (21.9)	400 (19.0)	0.061
DAPT	5925 (93.8)	1964 (93.2)	2005 (95.3)	1956 (93.0)	0.003
Biochemical analyses
White blood cell counts, 109/L	6.3 [5.2, 7.4]	6.2 [5.1, 7.1]	6.3 [5.2, 7.3]	6.6 [5.4, 7.9]	<0.001
Haemoglobin, g/L	137 [128, 147]	139 [130, 149]	137 [130, 148]	134 [123, 144]	<0.001
Platelet, 109/L	216 [180, 249]	219 [190, 255]	214 [179, 248]	208 [172, 245]	<0.001
LDL‐C, mmol/L	2.3 [1.8, 2.8]	2.3 [1.8, 2.9]	2.3 [1.8, 2.8]	2.2 [1.7, 2.7]	<0.001
hs‐cTnT, ng/L	11.2 [5.9, 41.7]	7.0 [4.2, 14.2]	11.0 [6.2, 34.5]	20.8 [9.6, 131.6]	<0.001
hs‐cTnT > 14 ng/L, n (%)	2711 (42.9)	533 (25.3)	882 (41.8)	1296 (61.5)	<0.001
Cystatin C, mg/L	1.0 [0.9, 1.2]	0.9 [0.8, 1.0]	1.0 [0.9, 1.1]	1.2 [1.0, 1.4]	<0.001
Creatinine, μmol/L	73.2 [63.0, 85.0]	69.0 [59.1, 79.0]	72.0 [63.0, 83.0]	79.0 [68.0, 94.0]	<0.001
BUN, mmol/L	5.1 [4.2, 6.2]	4.8 [4.0, 5.7]	4.9 [4.2, 5.9]	5.6 [4.5, 7.1]	<0.001
NT‐proBNP, ng/L	334.1 [78.9, 767.7]	139.0 [52.0, 767.7]	281.8 [83.2, 767.7]	767.7 [165.2, 767.7]	<0.001
GDF‐15, ng/L	1091.0 [790.5, 1635.0]	696.3 [574.6, 790.5]	1091.0 [987.4, 1225.0]	2020.0 [1636.2, 2759.0]	<0.001

ACS, acute coronary syndrome; BMI, body mass index; BUN, blood urea nitrogen; CABG, coronary artery bypass grafting; DAPT, dual antiplatelet therapy; ECG, electrocardiogram; eGFR, estimated glomerular filtration rate; GDF‐15, growth differentiation factor 15; hs‐cTnT, high‐sensitivity cardiac troponin T; LDL‐C, low‐density lipoprotein‐cholesterol; MI, myocardial infarction; NT‐proBNP, N‐terminal pro‐B‐type natriuretic peptide; PCI, percutaneous coronary intervention.

Values are median [inter‐quartile range] for continuous variables and n (%) for categorical variables.

Clinical outcomes

Cardiovascular death

During the 1 year follow‐up, 163 (2.58%) patients experienced all‐cause death, and 121 patients (1.91%) had CV death. The rates of all‐cause and CV deaths were 2.63%/year and 2.02%/year, respectively. Spline graphs showed a slight increase in CV death event rates at GDF‐15 concentrations up to 2080 ng/L, after which the risk increased more prominently up to and beyond 6000 ng/L (Figure  1 ). The estimates of CV death rates gradually increased across GDF‐15 tertiles (Supporting Information, Figure S2a ).

Figure 1.

Twelve months of event rate of the different outcomes. A/H HF, acute heart failure during hospitalization and hospitalization for heart failure post‐discharge; CV death, cardiovascular death; GDF‐15, growth differentiation factor 15; MI, myocardial infarction.

Higher concentrations of GDF‐15 were strongly associated with an increased risk of CV death after adjusting for clinical risk factors (HR 2.47, 95% CI 1.75–3.48, P < 0.001) and further adjusting for hs‐cTnT, NT‐proBNP, and cystatin C (HR 1.98, 95% CI 1.35–2.88, P < 0.001) (Table  2 ). Positive associations were observed in the ≥65 years of age, female sex, and without diabetes subgroups (Supporting Information, Table S7 ). A marginal interaction was observed between plasma GDF‐15 levels and age.

Table 2.

Association between growth differentiation factor 15 levels and clinical outcomes

	Event rate, N (%)	Model 1		Model 2
		HR (95% CI)	P‐value	HR (95% CI)	P‐value
CV death	121 (1.91)	2.47 (1.75, 3.48)	<0.001	1.98 (1.35, 2.88)	<0.001
MI	129 (2.04)	2.01 (1.43, 2.82)	<0.001	1.36 (0.93, 1.98)	0.110
Ischaemic stroke	39 (0.62)	1.60 (0.84, 3.04)	0.151	1.06 (0.52, 2.18)	0.869
A/H HF	140 (2.21)	3.72 (2.74, 5.05)	<0.001	2.69 (1.92, 3.77)	<0.001
Cardiogenic shock	137 (2.17)	1.96 (1.43, 2.67)	<0.001	1.46 (1.04, 2.05)	0.029
Major bleeding	73 (1.15)	1.82 (1.16, 2.85)	0.009	1.48 (0.91, 2.42)	0.116
Clinically significant bleeding	524 (8.29)	1.48 (1.25, 1.75)	<0.001	1.48 (1.22, 1.79)	<0.001

A/H HF, acute heart failure during hospitalization and hospitalization for heart failure post‐discharge; CI, confidence interval; CV death, cardiovascular death; HR, hazard ratio; MI, myocardial infarction.

The HR with 95% CI were calculated with per unit increase of log‐transformed growth differentiation factor 15 levels. For CV death, thrombotic events, and heart failure outcomes, we adjusted for age, sex, body mass index, hypertension, diabetes, previous MI, congestive heart failure, peripheral arterial disease, current smoking, stroke, previous percutaneous coronary intervention or coronary artery bypass grafting, estimated glomerular filtration rate, admitting diagnosis, low‐density lipoprotein‐cholesterol, platelet, haemoglobin, and white blood cells in Model 1. For bleeding events, we adjusted for covariates of CV death plus dual antiplatelet therapy at hospitalization, history of haemorrhage, and previous anticoagulants in Model 1. For all the outcomes, we adjusted for Model 1 plus high‐sensitivity cardiac troponin T, N‐terminal pro‐B‐type natriuretic peptide, and cystatin C in Model 2.

Thrombotic events

A total of 129 patients (2.04%) had spontaneous MI, and 39 (0.62%) had ischaemic stroke. The annual rates of MI and ischaemic stroke were 2.18% and 0.65%, respectively. Spline graphs showed a small increase in the event rate for MI at GDF‐15 concentrations up to 2080 ng/L, after which the event rates further increased (Figure  1 ). There was no increase in stroke incidence as the GDF‐15 concentration increased. The estimates of event rates by GDF‐15 tertile showed a gradually increasing risk of MI and ischaemic stroke (Supporting Information, Figure S2b,c ).

Higher concentrations of GDF‐15 were associated with a significantly increased risk of MI (HR 2.01, 95% CI 1.43–2.82, P < 0.001) but were not associated with an increased risk of ischaemic stroke (HR 1.60, 95% CI 0.84–3.04, P = 0.151) after adjusting for clinical risk factors. The association between GDF‐15 and MI was attenuated and was not statistically significant after further adjusting for biomarkers (Table  2 ).

Heart failure outcomes

A total of 140 patients (2.21%) had A/H HF, and 137 (2.17%) had cardiogenic shock during the 1 year follow‐up. The A/H HF and cardiogenic rates were 2.37%/year and 2.32%/year, respectively. Spline graphs showed a slight increase in A/H HF and cardiogenic shock rates at GDF‐15 concentrations up to 2080 ng/L, after which the risk increased more prominently for A/H HF up to and beyond 6000 ng/L, whereas event rates for cardiogenic shock increased smoothly (Figure  1 ). The estimates of event rates by GDF‐15 tertiles showed a gradually increasing risk of A/H HF and cardiogenic shock (Supporting Information, Figure S2d,e ).

Higher concentrations of GDF‐15 were associated with a markedly increased risk of both A/H HF (HR 3.72, 95% CI 2.74–5.05, P < 0.001) and cardiogenic shock (HR 1.96, 95% CI 1.43–2.67, P < 0.001) after adjusting for risk factors. The association of GDF‐15 with A/H HF (HR 2.69, 95% CI 1.92–3.77, P < 0.001) and cardiogenic shock (HR 1.46, 95% CI 1.04–2.05, P = 0.029) remained significant after adjusting for prognostic biomarkers (Table  2 ). Significant associations were observed for HF during hospitalization and after discharge, whereas these associations were not found for cardiogenic shock (Supporting Information, Table S6 ). Positive associations were observed for HF when stratified by age, sex, diabetes comorbidities, and admitting diseases (Supporting Information, Table S7 ).

Bleeding outcomes

A total of 637 (10.08%) patients experienced bleeding events, of whom 524 (8.29%) experienced CSB, and 73 (1.15%) experienced major bleeding events. The annual CSB and major bleeding rates were 9.45% and 1.23%, respectively. Spline graphs showed a small increase in the event rate for major bleeding at GDF‐15 concentrations up to 2080 ng/L, after which the event rates further increased. A slight increase in CSB was observed at GDF‐15 concentrations ranging from 365 to beyond 6000 ng/L (Figure  1 ). The estimates of event rates according to GDF‐15 tertile showed a gradually increasing risk of bleeding outcomes (Supporting Information, Figure S2f,g ).

Higher concentrations of GDF‐15 were associated with significantly increased risks of both major bleeding (HR 1.82, 95% CI 1.16–2.85, P = 0.009) and CSB (HR 1.48, 95% CI 1.25–1.75, P < 0.001) after adjusting for clinical risk factors. Higher GDF‐15 concentrations remained associated with CSB (HR 1.48, 95% CI 1.22–1.79, P < 0.001) but were not associated with major bleeding after further adjusting for biomarkers (Table  2 ). No significant interactions existed among age, sex, diabetes, or CAD stratum subtype (Supporting Information, Table S7 ).

Prognostic utility of growth differentiation factor 15

We examined the additive predictive value of GDF‐15 (Model 2) as a clinical risk factor and prognostic biomarker (Model 1) for clinical outcomes. Adding GDF‐15 did not significantly improve the C‐index for any individual outcome. However, NRI showed that GDF‐15 improved prediction for CV death (Model 2 vs. 1, NRI = 0.209, 95% CI 0.040–0.280, P = 0.020), A/H HF (Model 2 vs. 1, NRI = 0.217, 95% CI 0.000–0.372, P = 0.040), cardiogenic shock (Model 2 vs. 1, NRI = 0.130, 95% CI 0.033–0.218, P = 0.020), and CSB (Model 2 vs. 1, NRI = 0.090, 95% CI 0.047–0.137, P < 0.001) but not for MI, ischaemic stroke, or major bleeding (Table  3 ).

Table 3.

Prognostic utility of growth differentiation factor 15 incorporated into the model of clinical risk factors and prognostic biomarkers

	NRI (95% CI)		IDI (95% CI)
	Model 2 vs. 1	P‐value	Model 2 vs. 1	P‐value
CV death	0.209 (0.040, 0.280)	0.020	0.009 (−0.001, 0.032)	0.129
MI	0.117 (−0.084, 0.201)	0.129	0.002 (0, 0.008)	0.119
Ischaemic stroke	0.088 (−0.164, 0.267)	0.577	0 (−0.001, 0.016)	1
A/H HF	0.217 (0.000, 0.372)	0.040	0.009 (−0.019, 0.058)	0.567
Cardiogenic shock	0.130 (0.033, 0.218)	0.020	0.003 (0, 0.011)	0.040
Major bleeding	0.073 (−0.080, 0.215)	0.299	0.001 (−0.001, 0.010)	0.219
Clinically significant bleeding	0.090 (0.047, 0.137)	<0.001	0.004 (0.001, 0.009)	<0.001

A/H HF, acute heart failure during hospitalization and hospitalization for heart failure post‐discharge; CI, confidence interval; CV death, cardiovascular death; IDI, integrated discrimination improvement; MI, myocardial infarction; NRI, net reclassification improvement.

For CV death, thrombotic events, and heart failure outcomes, we did penalized least absolute shrinkage and selection operator (LASSO) regression analysis including age, sex, body mass index, current smoking, hypertension, diabetes, previous MI, congestive heart failure, peripheral arterial disease, previous stroke, previous percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG), estimated glomerular filtration rate (eGFR), admitting diagnosis, low‐density lipoprotein‐cholesterol (LDL‐C), platelet, haemoglobin, and white blood cells as covariates for Model 1. For bleeding outcomes, we did penalized LASSO regression analysis including age, sex, current smoking, body mass index, hypertension, diabetes, previous MI, congestive heart failure, peripheral arterial disease, previous stroke, previous PCI or CABG, eGFR, admitting diagnosis, LDL‐C, platelet, haemoglobin, white blood cells, dual antiplatelet therapy at hospitalization, history of haemorrhage, and previous anticoagulants as covariates for Model 1. The covariates screened by LASSO and high‐sensitivity cardiac troponin T, N‐terminal pro‐B‐type natriuretic peptide, and cystatin C were included in Model 1 for analysis, and the Cox regression model was used for modelling. For all outcomes, covariates from Model 1 plus growth differentiation factor 15 were included in Model 2.

Discussion

In this study on patients with CAD, we observed a strong association between higher admission levels of GDF‐15 and CV death after adjusting for clinical risk factors and prognostic biomarkers, mainly driven by the relationship between GDF‐15, A/H HF, and bleeding complications. The association of GDF‐15 with thrombotic events (MI events) was evident after adjusting for clinical predictors; however, these associations were attenuated and were not significant after further adjustment for hs‐cTnT, NT‐proBNP, and cystatin C. Our results verified the independent association between GDF‐15 and CV mortality and added information linking increased GDF‐15 levels and specific intermediates of CV death. Adding GDF‐15 may improve the NRI for CV death, HF events, and bleeding outcomes within models incorporating risk factors and prognostic biomarkers, indicating that including plasma GDF‐15 measurements in established CV risk factors may improve risk stratification. The main findings of this study are summarized in Figure 2 .

Figure 2.

The associations and prognostic utility of growth differentiation factor 15 (GDF‐15) with clinical outcomes. A/H HF, acute heart failure during hospitalization and hospitalization for heart failure post‐discharge; CV death, cardiovascular death; HR, hazard ratio; MI, myocardial infarction; NRI, net reclassification improvement.

Our results showed independent associations between admission levels of GDF‐15 and CV death, in line with previous studies that included patients with ACS, ⁹ , ¹³ patients with stable coronary heart disease, ¹² and community‐dwelling individuals with or without CV diseases. ¹⁷ , ¹⁸ Our results suggest that GDF‐15 provides incremental information beyond prognostic biomarkers. Accordingly, GDF‐15 showed a strong and independent association with HF outcomes, even after adjusting for clinical characteristics and conventional prognostic biomarkers. GDF‐15 was also modestly correlated with major bleeding and CSB events, even after controlling for age, sex, disease comorbidities, dual antiplatelet therapy usage, and organ dysfunction biomarkers. There were no significant associations between GDF‐15 levels and thrombotic events after adjusting for other prognostic biomarkers, consistent with previous studies. ⁹ , ¹¹ , ¹² , ¹⁹ These findings suggest that the relationship between GDF‐15 and CV death is not related to an increased risk of new coronary/cerebrovascular ischaemic events but rather may be due to the progression of underlying myocardial dysfunction with an increased risk of developing HF and to the progression of organ dysfunction with an increased risk of bleeding complications.

Identifying a biomarker with independent and differing associations with individual outcome events provides an opportunity for the improved assessment of the risk of specific events. For example, the balance between the risk of thrombotic events and bleeding outcomes in patients receiving antiplatelet treatment is currently difficult to achieve based on the assessment of clinical characteristics alone. Most clinical characteristics are related to an increased risk of thrombotic and bleeding events. ²⁰ , ²¹ The modest associations between GDF‐15 and bleeding outcomes, but not thrombotic events, indicate that elevated levels of GDF‐15 at hospital admission may be applied to differentiate CAD patients into high‐risk bleeding subgroups but not high‐risk thrombotic event subgroups. The close and independent association between GDF‐15 and the risk of bleeding was consistent across patients with ACS receiving dual antiplatelet therapy in PLATO ¹³ and patients with atrial fibrillation treated with oral anticoagulation therapy in the Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation (ARISTOTLE) trial. ²² , ²³ These findings extend the current knowledge of predictors of bleeding outcomes.

The association between serum GDF‐15 levels and HF events has clinical implications. NT‐proBNP is a major biomarker with prognostic value for HF. However, the NT‐proBNP levels vary widely with age. A systematic review by Santaguida et al. found 79 studies with follow‐up intervals from 14 days to 7 years, with cut points associated with mortality ranging from 3001 to 17 860 ng/L. ²⁴ It is known that other confounders for NT‐proBNP and prognosis, such as age, renal filtration function, obesity, and race, play a role in this large expanse of values that could be important for patients. The association of GDF‐15 with HF could complement the limitations of NT‐proBNP levels.

Our results showed independent associations between higher GDF‐15 levels and various risk factors, including older age, diabetes mellitus, and renal dysfunction, implying that CV risk factors could be the major drivers of increased circulating GDF‐15 levels. These findings are consistent with those of previous studies in community‐dwelling elderly individuals and patients with CAD, in which GDF‐15 was related to clinical risk predictors for CV outcomes. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ²⁵ Although the production and secretion of GDF‐15 in multiple tissues (i.e. coronary atherosclerotic plaques, cardiomyocytes, and adipose tissues) in response to acute and chronic stressors have been recognized, ⁶ , ²⁶ little is known about the biological effects of GDF‐15 on local and systemic tissues. Limited evidence has shown that GDF‐15 has anti‐apoptotic, anti‐inflammatory, cytoprotective, and pro‐survival effects after MI. ²⁷ , ²⁸ There is a divergence between the clinical indications and biological effects of GDF‐15. GDF‐15 should be conceptualized as a predictive marker that reflects the combined information of multiple clinical characteristics rather than as an effector that directly influences CAD prognosis.

Remarkable associations between GDF‐15 levels and other biomarkers, such as hs‐cTnT, NT‐proBNP, and cystatin C, were also observed in this study. In contrast to these biomarkers, which carry specific information on disease category and severity (i.e. hs‐cTnT reflects myocardial damage, NT‐proBNP reflects myocardial stretch and increased wall tension, and cystatin C reflects renal dysfunction), GDF‐15 lacks disease and tissue specificity. ⁶ These findings suggest that GDF‐15 mediates distinct disease pathways compared with these biomarkers.

The additive predictive value of GDF‐15 compared with CV risk factors and prognostic biomarkers for clinical outcomes was observed in our study. This evidence indicates that GDF‐15 has potential clinical utility in predicting the 1 year risk of CV death, HF events, and bleeding outcomes after hospitalization. GDF‐15 assays for clinical use are under development or in pilot tests across several European regions. ⁶ , ¹² The findings of our study add new knowledge to the prognostic utility of GDF‐15 in a large Chinese CAD population.

Circulating levels of GDF‐15 also have potential in CV risk management, owing to the relationship of this marker with various clinical conditions. Several risk factors, such as diabetes and hypertension, are modifiable and can be controlled if patients adhere to healthy behaviours. ²⁹ Focusing on increased GDF‐15 concentrations may motivate individuals to follow a healthier lifestyle and provide reasons for stricter risk factor control and monitoring.

Limitations

This study has several limitations. First, although significant associations and prognostic value of GDF‐15 beyond conventional risk factors and biomarkers were observed, further validation studies and prospective trials are warranted to evaluate risk prediction tools incorporating GDF‐15. Second, only a single measurement of the GDF‐15 level was available at admission, and the dynamic change trend of the GDF‐15 level was not observed in this study. However, other studies have found that the concentration of GDF‐15 decreases by 4% over 4–6 months. ³⁰ Therefore, changes in the concentrations of GDF‐15 probably have little influence on the results. Third, no corrections for multiple analyses were performed owing to the exploratory nature of this study. Fourth, although we evaluated the associations between GDF‐15 levels and outcomes in various subgroups, we did not investigate their interactions with treatment effects.

Conclusions

GDF‐15 is associated with an increased 1 year risk of CV death, HF events, and bleeding outcomes but not thrombotic events in Chinese patients with CAD. The inclusion of GDF‐15 in a clinical risk model with or without prognostic biomarkers could improve the NRI for CV death, HF events, and bleeding outcomes. These results indicate that GDF‐15 may be a prognostic biomarker for CV death, HF, and bleeding outcomes, thus refining the risk assessment of these specific outcomes.

Conflict of interest

The authors declare that they have no conflicts of interest.

Funding

This study was supported by the National Key Research and Development Program of China (2020YFC1512700, 2020YFC1512705, 2020YFC1512703), the National Science and Technology Fundamental Resources Investigation Project (2018FY100600, 2018FY100602), the Key Research and Development Program of Shandong Province (2021ZLGX02, 2021SFGC0503), the Taishan Pandeng Scholar Program of Shandong Province (tspd20181220), the Taishan Young Scholar Program of Shandong Province (tsqn20161065, tsqn201812129), the Youth Top‐Talent Project of National Ten Thousand Talents Plan and Qilu Young Scholar Program, and the ECCM Program of Clinical Research Center of Shandong University (2021SDUCRCA005).

Supporting information

Table S1. Characteristics of participating hospitals in the BIPass study.

Table S2. Patient inclusion and exclusion criteria.

Table S3. The definitions of the end points.

Table S4. The adjusted covariates in the association analysis.

Table S5. Associations of GDF15 levels with baseline characteristics.

Table S6. Associations of GDF‐15 levels with clinical outcomes at different time intervals.

Table S7. Associations of GDF‐15 levels with clinical outcomes in subgroups.

Table S8. Prognostic utility of GDF‐15 incorporated into model of CV risk factors.

Figure S1. Geographic setting of the participating hospitals.

Figure S2. Kaplan–Meier estimated event rates of clinical outcomes.

Wang, J. , Zhang, T. , Xu, F. , Gao, W. , Chen, M. , Zhu, H. , Xu, J. , Yin, X. , Pang, J. , Zhang, S. , Wei, M. , Chen, J. , Liu, Y. , Yu, X. , Chew, D. P. , and Chen, Y. (2023) GDF‐15 at admission predicts cardiovascular death, heart failure, and bleeding outcomes in patients with CAD. ESC Heart Failure, 10: 3123–3132. 10.1002/ehf2.14484.