Impact of Chronic Kidney Disease on the Associations of Cardiovascular Biomarkers With Adverse Outcomes in Patients With Suspected or Known Coronary Artery Disease: The EXCEED-J Study.

Background The impact of chronic kidney disease (CKD) on the prognostic utility of cardiovascular biomarkers in high-risk patients remains unclear. Methods and Results We performed a multicenter, prospective cohort study of 3255 patients with suspected or known coronary artery disease (CAD) to investigate whether CKD modifies the prognostic utility of cardiovascular biomarkers. Serum levels of cardiovascular and renal biomarkers, including soluble fms-like tyrosine kinase-1 (sFlt-1), N-terminal pro-brain natriuretic peptide (NT-proBNP), high-sensitivity cardiac troponin-I (hs-cTnI), cystatin C, and placental growth factor, were measured in 1301 CKD and 1954 patients without CKD. The urine albumin to creatinine ratio (UACR) was measured in patients with CKD. The primary outcome was 3-point MACE (3P-MACE) defined as a composite of cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke. The secondary outcomes were all-cause death, cardiovascular death, and 5P-MACE defined as a composite of 3P-MACE, heart failure hospitalization, and coronary/peripheral artery revascularization. After adjustment for clinical confounders, sFlt-1, NT-proBNP, and hs-cTnI, but not other biomarkers, were significantly associated with 3P-MACE, all-cause death, and cardiovascular death in the entire cohort and in patients without CKD. These associations were still significant in CKD only for NT-proBNP and hs-cTnI. NT-proBNP and hs-cTnI were also significantly associated with 5P-MACE in CKD. The UACR was not significantly associated with any outcomes in CKD. NT-proBNP and hs-cTnI added incremental prognostic information for all outcomes to the model with potential clinical confounders in CKD. Conclusions NT-proBNP and hs-cTnI were the most powerful prognostic biomarkers in patients with suspected or known CAD and concomitant CKD.

3‐point major adverse cardiovascular events

5‐point major adverse cardiovascular events

chronic kidney disease

Establishment of the method to extract a high‐risk population employing novel biomarkers to predict cardiovascular events in Japan

fms‐like tyrosine kinase 1

high‐sensitivity cardiac troponin‐I

integrated discrimination improvement

major adverse cardiovascular events

neutrophil gelatinase‐associated lipocalin

National Hospital Organization

net reclassification improvement

placental growth factor

soluble fms‐like tyrosine kinase 1

urine albumin to creatinine ratio

vascular endothelial growth factor

vascular endothelial growth factor receptor

Clinical Perspective

What Is New?

This is the first dedicated and large‐scale prospective cohort study to demonstrate that higher levels of NT‐proBNP (N‐terminal pro‐brain natriuretic peptide) and hs‐cTnI (high‐sensitivity cardiac troponin‐I), but not those of sFlt‐1 (soluble fms‐like tyrosine kinase 1) or UACR (urine albumin to creatinine ratio), independently predicted cardiovascular events and mortality in patients with suspected or known coronary artery disease and concomitant chronic kidney disease.

What Are the Clinical Implications?

Despite the possible chronic elevation of serum levels by renal insufficiency, NT‐proBNP and hs‐cTnI serve as powerful prognostic biomarkers beyond the other biomarkers, including sFlt‐1 and UACR, in patients with suspected or known coronary artery disease and concomitant chronic kidney disease.

Chronic kidney disease (CKD) is a global public health problem due to its rising prevalence, poor outcomes, and high treatment cost. , , , It increases the risk of all‐cause mortality, cardiovascular disease, and progression to kidney failure, independent of known coronary artery disease (CAD) risk factors such as hypertension, diabetes, and dyslipidemia. , , , Among subjects with CKD, cardiovascular disease is the leading cause of morbidity and mortality.

Circulating levels of established cardiovascular biomarkers, ie, N‐terminal pro‐brain natriuretic peptide (NT‐proBNP), high‐sensitivity cardiac troponin‐I (hs‐cTnI), and high‐sensitivity C‐reactive protein (hs‐CRP), can be chronically elevated by decreased renal clearance, and they are thus equally or less predictive for cardiovascular events in CKD than in non‐CKD subjects in the general population. , , However, the impact of CKD on the prognostic utility of established cardiovascular biomarkers in high‐risk patients with suspected or known CAD remains unclear. In addition, there may be better predictors of cardiovascular events and mortality than established cardiovascular biomarkers in patients with suspected or known CAD and concomitant CKD.

The vascular endothelial growth factor (VEGF) family members, including VEGF, VEGF‐B, VEGF‐C, VEGF‐D, and placental growth factor (PlGF), exhibit different patterns of binding to VEGF receptors (VEGFRs) on endothelial cells, and they differentially regulate blood and lymphatic vessel development and growth. VEGF binds to VEGFR‐1 (also called fms‐like tyrosine kinase 1 [Flt‐1]) and VEGFR‐2. While VEGF‐VEGFR‐2 signaling is essential for vascular development and maintenance, Flt‐1 acts as an anti‐angiogenic decoy receptor for VEGF and is required for proper vascular development. , , A soluble truncated form of Flt‐1 (sFlt‐1) is secreted by endothelial cells by alternative splicing of the Flt‐1 mRNA. sFlt‐1 has been shown to cause endothelial dysfunction, decrease angiogenesis, impair capillary repair, and increase proteinuria. ,

A previous study found that increased sFlt‐1 levels are associated with endothelial dysfunction in patients with CKD. Since endothelial dysfunction, which is one of the initial pathological processes of atherosclerosis, is associated with an increased cardiovascular risk, , an increase in circulating sFlt‐1 may be associated with cardiovascular risk in CKD. A relatively small‐scale observational study showed that circulating sFlt‐1 levels were associated with adverse outcomes in patients with CKD (stages 2–4). However, whether sFlt‐1 can predict cardiovascular events and mortality in patients with CKD should be confirmed in a larger cohort study.

Cystatin C and neutrophil gelatinase‐associated lipocalin (NGAL) are renal biomarkers for acute kidney injury and CKD progression. , Cystatin C is a serum measure of renal function that appears to be independent of age, sex, and lean muscle mass. Cystatin C has been shown to be associated with all‐cause death, cardiovascular death, and cardiovascular events in elderly persons living in the community. In another study, the associations of cystatin C with all‐cause death and cardiovascular death were independent of the glomerular filtration rate. NGAL is a glycoprotein released by damaged renal tubular cells and is a sensitive marker of acute kidney injury. , Circulating levels of NGAL have been independently associated with all‐cause death, cardiovascular death, and cardiovascular events in community‐dwelling older adults.

In the present study, therefore, we investigated whether possible novel biomarkers, including sFlt‐1, and established cardiovascular biomarkers, ie, NT‐proBNP, hs‐cTnI, and hs‐CRP, as well as renal biomarkers can predict cardiovascular events and mortality and whether CKD modifies the prognostic utility of these biomarkers in a large‐scale, multicenter prospective cohort study of patients with suspected or known CAD.

METHODS

Study Population

The data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure. Patients with suspected or known CAD (ie, stable angina, ischemic heart disease, chest pain, positive cardiac stress test) undergoing elective coronary angiography were recruited in the EXCEED‐J (Establishment of the method to extract a high risk population employing novel biomarkers to predict cardiovascular events in Japan) study (UMIN000018807): a nationwide, multicenter, prospective cohort study to determine whether sFlt‐1 or other biomarkers can predict cardiovascular events in patients with CKD or other risk factors, and to establish the methods to efficiently extract high‐risk patients. The EXCEED‐J study group consists of 17 National Hospital Organization (NHO) institutions across Japan, and the present study was conducted by nationally certified cardiologists. The exclusion criteria included malignancy, inflammatory disease, heparin use, steroid or other hormone replacement therapy, inability to consent, scheduled follow‐up angiography after coronary revascularization, and patients determined as ineligible by the attending physician. Between November 2013 and May 2017, a total of 3311 patients were consecutively enrolled. After excluding 47 patients who did not provide blood samples and 9 patients who withdrew consent, a total of 3255 (1301 CKD and 1954 non‐CKD) patients were eligible. The estimated glomerular filtration rate (eGFR) was calculated with the new Japanese coefficient for the abbreviated Modification of Diet in the Renal Disease Study equation, including a correction factor of 0.739 for women. CKD is defined as a creatinine‐based eGFR <60 mL/min per 1.73 m2. The prevalence of risk factors for cardiovascular disease was determined by the examining physician (as described in Data S1). Data on demographic characteristics, smoking status, medical history, and medication use were collected from medical records. Submitted data were examined for completeness and accuracy by the coordinating center (Clinical Research Institute, Kyoto Medical Center, Kyoto, Japan), and data queries were sent to study sites. The study was approved by the central ethics committee of the NHO headquarters and each institution’s ethical committee. All patients provided written informed consent.

Outcomes and Follow‐Up

The primary outcome was 3‐point major adverse cardiovascular events (3P‐MACE) defined as a composite of cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke. The prespecified secondary outcomes were all‐cause death, cardiovascular death, and 5P‐MACE defined as a composite of 3P‐MACE, heart failure hospitalization, and coronary/peripheral artery revascularization. The patients were monitored over 3 years (1080 days) for the occurrence of 3P‐MACE, all‐cause death, cardiovascular death, and/or 5P‐MACE. The follow‐up was performed by personnel blinded to the biomarker data through medical record/chart reviews, a survey letter, and/or telephone interviews.

Sudden death resulting from an unknown but presumed cardiovascular cause in high‐risk patients was included in cardiovascular death. All deaths and MACE were recorded in the official medical chart of the hospitals where the patients received care. The reported deaths, myocardial infarctions, and strokes were reviewed and adjudicated by the expert committee (three independent and blinded cardiologists). The follow‐up continued even after nonfatal myocardial infarction and/or nonfatal stroke had occurred. At the end of the follow‐up (day 1080), the survival status and detailed information about MACE were available in 3220 patients (98.9%), and 35 patients (1.1%) were lost to follow‐up.

Exposures, Sample Collection, and Biomarker Measurement

Heparin‐free fasting blood samples for serum were collected from the peripheral vein before each patient's coronary angiography. The serum was stored at −80 °C for a mean of 4 months until it was assayed for sFlt‐1, hs‐CRP, cystatin C, neutrophil gelatinase‐associated lipocalin (NGAL), VEGF, and PlGF after one freeze‐thaw cycle. The serum levels were measured with specific, commercially available, kits according to the manufacturers’ instructions (Quantikine, R&D Systems, Minneapolis, MN, for sFlt‐1, cystatin C, VEGF, and PlGF; CycLex, Medical & Biological Laboratories Co., Ltd. [MBL], Nagano, Japan for hs‐CRP; BioPorto A/S, Hellerup, Denmark for NGAL). The sensitivity of the assay for sFlt‐1 was 3.5 pg/mL. The inter‐/intra‐assay coefficients of variation of the ELISA for sFlt‐1 were <10%/<4%. The sensitivities of the assays for hs‐CRP, cystatin C, NGAL, VEGF, and PlGF were 0.0286 mg/L, 0.102 ng/mL, 4 pg/mL, 5 pg/mL, and 7 pg/mL, respectively. The inter‐/intra‐assay coefficients of variation of ELISAs for hs‐CRP, cystatin C, NGAL, VEGF, and PlGF were <6%/<4%, ≤7%/<7%, 8.2%/3.0%, ≤7%/<5%, and <12%/≤7%, respectively. The assays were performed by an investigator blinded to the sources of the samples.

The details of the assay for NT‐proBNP are described elsewhere. , , Briefly, the serum levels of NT‐proBNP were measured using a validated, sandwich electrochemiluminescence immunoassay (Elecsys; Roche Diagnostics, Indianapolis, IN). The sensitivity of the assay for NT‐proBNP was 5 pg/mL, and the assay coefficient of variation at values of the measuring range (5–35 000 pg/mL) was <10%. The hs‐cTnI values were measured using a cardiac troponin assay (Architect Stat High‐Sensitive Troponin I; Abbott Laboratories, Abbot Park, IL, USA). The limit of detection in this assay is 1.9 pg/mL (range, 0–50 000 pg/mL) and the 99th percentile cut‐off is 26.2 pg/mL. The urine albumin to creatinine ratio (UACR) was measured by the routine method. Additional details are described elsewhere. , ,

Statistical Analysis

We divided the patients into 2 groups according to the presence or absence of CKD. The baseline data were compared between CKD and non‐CKD groups and significant differences were determined using the Wilcoxon and χ2 tests. The relationships between sFlt‐1 and other variables were assessed in simple and stepwise multiple linear regression analyses. Stepwise variable selection was performed in a forward direction with the Bayesian information criterion. Because sFlt‐1, the Gensini score, NT‐proBNP, hs‐cTnI, hs‐CRP, cystatin C, NGAL, VEGF, and PlGF were normally distributed after logarithmic transformation, the logarithms of these parameters were used in the linear regression analyses. The cumulative incidences of clinical outcomes were estimated by the Kaplan‐Meier method. The relationships between the baseline biomarkers levels (as continuous variables, tertiles, and the top tertile [ie, tertile 3 versus tertiles 1 and 2]) and the outcomes were investigated with the use of Cox proportional hazard regression in models adjusted for potential clinical confounders (ie, age, sex, body mass index [BMI], hypertension, dyslipidemia, diabetes, current smoking, eGFR, the Gensini score, previous myocardial infarction, previous stroke, previous heart failure hospitalization, atrial fibrillation, anemia [defined as a hemoglobin level below 13 g/L in men and 12 g/L in women], antihypertensive drug use, statin use, and aspirin use). The biomarkers were log‐transformed for use as continuous variables.

We evaluated the incremental predictive performance of selected biomarkers by calculating changes in the C‐statistic, continuous net reclassification improvement (NRI), and integrated discrimination improvement (IDI) metrics. We assessed the model calibration by comparing predicted probabilities with observed probabilities. A residual analysis was used to assess the model fit. Additional details are described in Data S1.

All statistical tests were two‐sided, and P<0.05 was considered significant. Since all analyses were considered exploratory, the P‐values were not adjusted for multiple comparisons. The analyses were performed using JMP13 (SAS, Cary, NC) and R, ver. 3.5.2 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Baseline Characteristics

The baseline characteristics of the entire cohort and those divided according to the presence or absence of CKD are shown in Table 1 and Table S1. The proportions of CKD stages were as follows: stage 3a, 64.1%; stage 3b, 24.8%; stage 4, 5.6%; and stage 5, 5.6% (Table S1). Patients with CKD had older age, higher rates of hypertension, diabetes, former smoking, obstructive CAD, previous stroke, previous heart failure hospitalization, atrial fibrillation, anemia and antihypertensive drug use, lower rate of current smoking, lower eGFR, and higher Gensini score. There were no significant differences in the rate of male sex, the BMI, or the rates of obesity, dyslipidemia, previous myocardial infarction, statin use, or aspirin use. Serum levels of sFlt‐1, NT‐proBNP, hs‐cTnI, hs‐CRP, cystatin C, and NGAL were significantly higher in patients with CKD than in those without CKD. Those of VEGF and PlGF were similar between the two groups. The sFlt‐1/PlGF ratio was significantly higher in patients ith CKD than in patients without CKD. The baseline characteristics according to the tertiles of sFlt‐1 levels in the entire cohort, patients with CKD, and patients without CKD are shown in Tables S2 through S4, respectively.

Table 1

Baseline Characteristics in the Entire Cohort, Patients With CKD*, and Patients Without CKD

Baseline characteristics and incidence of events	Entire cohort (n=3255)	CKD (n=1301)	Non‐CKD (n=1954)	P value†
Age, mean (SD), y	70.2 (10.4)	73.5 (8.5)	68.0 (11.0)	<0.001
Male	2272 (69.8)	892 (68.6)	1380 (70.6)	0.210
Body mass index, mean (SD)	24.5 (4.0)	24.5 (4.0)	24.5 (4.0)	0.673
Obesity ‡	1311 (40.3)	538 (41.4)	773 (39.6)	0.307
Hypertension	2483 (76.3)	1112 (85.5)	1371 (70.2)	<0.001
Dyslipidemia	2480 (76.2)	1003 (77.1)	1477 (75.6)	0.323
Diabetes	1281 (39.4)	566 (43.5)	715 (36.6)	<0.001
Current smoker	591 (18.2)	189 (14.5)	402 (20.6)	<0.001
Former smoker	1390 (42.7)	594 (45.7)	796 (40.7)	0.005
eGFR, mean (SD), mL/min per 1.73 m2	63 (20)	45 (13)	76 (14)	<0.001
Gensini score, median (IQR) §	10.5 (2.0–31.5)	13.0 (3.0–34.8)	9.5 (2.0–28.5)	<0.001
Obstructive coronary artery disease	1988 (61.1)	828 (63.6)	1160 (59.4)	0.014
Previous myocardial infarction	446 (13.7)	181 (13.9)	265 (13.6)	0.776
Previous stroke	385 (11.8)	185 (14.2)	200 (10.2)	<0.001
Previous heart failure hospitalization	285 (8.8)	173 (13.3)	112 (5.7)	<0.001
Atrial fibrillation	324 (10.0)	172 (13.2)	152 (7.8)	<0.001
Anemia ‖	928 (28.5)	516 (39.7)	412 (21.1)	<0.001
Antihypertensive drug use	2684 (82.5)	1154 (88.7)	1530 (78.3)	<0.001
Statin use	1922 (59.1)	742 (57.0)	1180 (60.4)	0.057
Aspirin use	1714 (52.7)	674 (51.8)	1040 (53.2)	0.427
sFlt‐1, median (IQR), pg/mL	108 (91–131)	112 (94–134)	105 (89–129)	<0.001
NT‐proBNP, median (IQR), pg/mL	165 (65–598)	301 (106–1268)	115 (51–339)	<0.001
hs‐cTnI, median (IQR), pg/mL	8 (4–16)	10 (6–23)	6 (4–13)	<0.001
hs‐CRP, median (IQR), mg/L	0.6 (0.2–1.8)	0.7 (0.3–2.1)	0.5 (0.2–1.6)	<0.001
Cystatin C, median (IQR), mg/L	0.8 (0.7–1.0)	1.0 (0.8–1.2)	0.7 (0.6–0.9)	<0.001
NGAL, median (IQR), ng/mL	97 (68–139)	122 (85–178)	85 (62–117)	<0.001
VEGF, median (IQR), pg/mL	300 (184–468)	306 (190–477)	294 (180–462)	0.117
PlGF, median (IQR), pg/mL	14 (11–16)	14 (11–17)	14 (11–16)	0.371
sFlt‐1/PlGF ratio, median (IQR)	7.9 (6.1–10.6)	8.1 (6.2–10.8)	7.8 (6.1–10.5)	0.036
UACR, median (IQR), mg/g ¶	…	20 (8–83)	…	…

Values are expressed as number (percentage) unless otherwise indicated. CKD indicates chronic kidney disease; eGFR, estimated glomerular filtration rate; hs‐CRP, high‐sensitivity C‐reactive protein; hs‐cTnI, high‐sensitivity cardiac troponin I; IQR, interquartile range; NGAL, neutrophil gelatinase‐associated lipocalin; NT‐proBNP, N‐terminal pro‐brain natriuretic peptide; PlGF, placental growth factor; sFlt‐1, soluble fms‐like tyrosine kinase 1; UACR, urine albumin to creatinine ratio; and VEGF, vascular endothelial growth factor.

CKD is defined as an estimated glomerular filtration rate of <60 mL/min per 1.73 m2 of body surface area.

The P‐value represents a comparison of the differences between CKD and Non‐CKD, and is based on the χ2 test of independence for categorical variables, and the Wilcoxon test for continuous variables.

Obesity is defined as a body mass index of 25 or more.

The Gensini score represents the angiographic severity of coronary artery disease employing a nonlinear points system for degree of luminal narrowing.

Anemia is defined as a hemoglobin level of <13 g/dL in men and <12 g/dL in women.

There are missing data for 223 patients.

Figure S1 shows the comparison of sFlt‐1 levels among patients without CKD and those with CKD stages 3a, 3b, 4, and 5. The sFlt‐1 level increased in proportion to the severity of CKD. The correlations of sFlt‐1 with other variables are shown in Table S5. Stepwise regression analysis revealed that higher sFlt‐1 levels were independently associated with previous heart failure hospitalization, atrial fibrillation, absence of anemia, no use of statins, higher levels of NT‐proBNP, hs‐CRP and NGAL, and lower levels of cystatin C and VEGF.

Incidence of Outcomes

Incidences of prespecified outcomes in the entire cohort, patients with CKD, and patients without CKD are shown in Table 2. During the 3‐year follow‐up, 156 patients developed 3P‐MACE (12 myocardial infarctions, 77 strokes, and 67 cardiovascular deaths), 215 died from any cause (82 cardiovascular and 133 non‐cardiovascular deaths), and 1361 developed 5P‐MACE (156 3P‐MACEs, 132 heart failure hospitalizations, and 1141 coronary/peripheral artery revascularizations). Figure 1 shows the cumulative incidence of 3P‐MACE according to the tertiles of sFlt‐1 levels in the entire cohort (Figure 1A), patients with CKD (Figure 1B), and patients without CKD (Figure 1C). Patients in the top tertile of sFlt‐1 had the greatest risk of 3P‐MACE within the entire cohort, patients with CKD, and patients without CKD. Figures S2 through S4 show the cumulative incidence of all‐cause death, cardiovascular death, and 5P‐MACE according to the tertiles of sFlt‐1 levels in the entire cohort, patients with CKD, and patients without CKD, respectively. The top tertile of sFlt‐1 also had the greatest risks of all‐cause death and cardiovascular death irrespective of the presence or absence of CKD. In contrast, there was no difference in the cumulative incidence of 5P‐MACE among tertiles of sFlt‐1. The incidences of prespecified outcomes according to tertiles of sFlt‐1 levels in the entire cohort, patients with CKD, and patients without CKD are shown in Tables S2 through S4.

Table 2

Incidence of Outcomes in the Entire Cohort, Patients With CKD, and Patients Without CKD

Type of outcomes	Entire cohort (n=3255)	CKD (n=1301)	Non‐CKD (n=1954)
3‐point MACE*	156 (16.8)	88 (24.2)	68 (12.0)
All‐cause death	215 (22.9)	128 (34.6)	87 (15.2)
Cardiovascular death	82 (8.7)	50 (13.5)	32 (5.6)
5‐point MACE †	1361 (226.8)	595 (261.3)	766 (205.7)
Myocardial infarction	12 (1.3)	5 (1.4)	7 (1.2)
Stroke	77 (8.3)	42 (11.5)	35 (6.2)
Heart failure hospitalization	179 (19.5)	107 (30.1)	72 (12.8)
Revascularization for coronary/peripheral artery disease	1151 (183.2)	477 (196.4)	674 (174.9)
PCI	936 (137.1)	365 (135.3)	571 (138.2)
CABG	137 (15.1)	74 (21.1)	63 (11.4)
Peripheral artery disease	134 (14.7)	62 (17.4)	72 (13.0)

Values are expressed as number (/1000 person‐years). CABG indicates coronary artery bypass grafting; MACE, major adverse cardiovascular events; and PCI, percutaneous coronary intervention.

3‐point MACE is defined as a composite of cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke.

5‐point MACE is defined as a composite of 3‐point MACE, heart failure hospitalization, and revascularization for coronary/peripheral artery disease.

Figure 1

Cumulative incidence of 3P‐MACE in the entire cohort (A), patients with CKD (B), and patients without CKD (C) according to the serum sFlt‐1 level at baseline.

Follow‐up results are truncated after 3 years. 3P‐MACE is defined as a composite of cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke. CKD is defined as an estimated glomerular filtration rate of <60 mL/min per 1.73 m2 of body surface area. The tertiles of sFlt‐1 levels were as follows: (A) tertile 1, ≤96.59; tertile 2, 96.59<, ≤121.17; tertile 3, >121.17 pg/mL; (B) tertile 1, ≤100.00; tertile 2, 100.00<, ≤124.91; tertile 3, >124.91 pg/mL; (C) tertile 1, ≤94.45; tertile 2, 94.45<, ≤119.69; tertile 3, >119.69 pg/mL. 3P‐MACE indicates 3‐point major adverse cardiovascular events; CKD, chronic kidney disease; and sFlt‐1, soluble fms‐like tyrosine kinase 1.

Multivariate Cox Regression Analyses

Figure 2 shows adjusted hazard ratios (HRs) of each biomarker level as (1) a natural log‐transformed continuous variable (per 1‐SD increase), (2) tertiles, and (3) the top tertile (ie, tertile 3 [versus tertiles 1 and 2]) for 3P‐MACE in the entire cohort, patients with CKD, and patients without CKD. The tertiles of biomarker levels and numbers of patients are summarized in Table S6. After adjusting for potential clinical confounders (ie, age, sex, BMI, hypertension, dyslipidemia, diabetes, current smoking, eGFR, the Gensini score, previous myocardial infarction, previous stroke, previous heart failure hospitalization, atrial fibrillation, anemia, antihypertensive drug use, statin use, and aspirin use), the sFlt‐1 level (as a continuous variable) was significantly associated with 3P‐MACE in the entire cohort (HR, 1.26; 95% CI [95% CI], 1.11–1.43) and in patients without CKD (HR, 1.40; 95% CI, 1.18–1.65), but not in patients with CKD (HR, 1.13; 95% CI, 0.92–1.35). The tertile analysis of sFlt‐1 levels revealed that there was an apparent threshold effect between tertile 2 and tertile 3 in the incidence of 3P‐MACE. Thus, sFlt‐1 was also modeled as a dichotomous variable by applying a threshold of the tertiles 1 and 2 versus tertile 3. The top tertile (ie, tertile 3 [versus tertiles 1 and 2]) of sFlt‐1 was significantly associated with 3P‐MACE in the entire cohort (HR, 1.63; 95% CI, 1.17–2.26) and in patients with CKD (HR, 1.78; 95% CI, 1.14–2.76), but not in patients without CKD (HR, 1.63; 95% CI, 0.99–2.68).

Figure 2

Adjusted hazard ratios of the biomarker levels for 3P‐MACE in the entire cohort, patients with CKD, and patients without CKD.

The data were adjusted for age, sex, body mass index, hypertension, dyslipidemia, diabetes, current smoking, estimated glomerular filtration rate, the Gensini score, previous myocardial infarction, previous stroke, previous heart failure hospitalization, atrial fibrillation, anemia, antihypertensive drug use, statin use and aspirin use. CKD is defined as an estimated glomerular filtration rate of <60 mL/min per 1.73 m2 of body surface area. The biomarkers are modeled as (1) continuous variables, (2) tertiles, and (3) the top tertile (ie, tertile 3 vs tertiles 1 and 2), and are natural log‐transformed for use as continuous variables. NT‐proBNP indicates N‐terminal pro‐brain natriuretic peptide; hs‐cTnI, high‐sensitivity cardiac troponin I; hs‐CRP, high‐sensitivity C‐reactive protein; NGAL, neutrophil gelatinase‐associated lipocalin; VEGF, vascular endothelial growth factor; PlGF, placental growth factor; and UACR: urine albumin to creatinine ratio. Other abbreviations used in this figure are the same as in Figure 1. The tertiles of biomarker levels and number of patients are summarized in Table S6.

Serum levels of NT‐proBNP and hs‐cTnI (as continuous variables) were significantly associated with 3P‐MACE in the entire cohort (NT‐proBNP: HR, 1.90; 95% CI, 1.58–2.28; hs‐cTnI: HR, 1.44; 95% CI, 1.28–1.63), patients with CKD (NT‐proBNP: HR, 1.98; 95% CI, 1.50–2.61; hs‐cTnI: HR, 1.51; 95% CI, 1.25–1.79), and patients without CKD (NT‐proBNP: HR, 1.81; 95% CI, 1.39–2.36; hs‐cTnI: HR, 1.38; 95% CI, 1.17–1.64). Among other biomarkers, only cystatin C levels (as a continuous variable) were significantly associated with 3P‐MACE in the entire cohort, and no biomarkers including UACR were significantly associated with 3P‐MACE in patients with CKD.

Adjusted HRs of each biomarker level for all‐cause death, cardiovascular death, and 5P‐MACE are shown in Figures S5 through S7, respectively. After adjusting for potential clinical confounders, the sFlt‐1 level (as a continuous variable) was significantly associated with all‐cause death in the entire cohort (HR, 1.23; 95% CI, 1.10–1.37), patients with CKD (HR, 1.19; 95% CI, 1.03–1.38), and patients without CKD (HR, 1.29; 95% CI, 1.09–1.53) (Figure S5), whereas it was significantly associated with cardiovascular death in the entire cohort (HR, 1.34; 95% CI, 1.14–1.56) and in patients without CKD (HR, 1.55; 95% CI, 1.24–1.93), but not in patients with CKD (HR, 1.20; 95% CI, 0.96–1.50) (Figure S6). In contrast, the sFlt‐1 level was not significantly associated with 5P‐MACE either as a continuous variable or the top tertile in the entire cohort, patients with CKD or patients without CKD (Figure S7).

Serum levels of NT‐proBNP and hs‐cTnI (as continuous variables) were significantly associated with all‐cause death in the entire cohort (NT‐proBNP: HR, 1.68; 95% CI, 1.43–1.98; hs‐cTnI: HR, 1.25; 95% CI, 1.12–1.41), patients with CKD (NT‐proBNP: HR, 1.75; 95% CI, 1.38–2.22; hs‐cTnI: HR, 1.32; 95% CI, 1.12–1.56), and patients without CKD (NT‐proBNP: HR, 1.53; 95% CI, 1.20–1.96; hs‐cTnI: HR, 1.20; 95% CI, 1.01–1.43) (Figure S5); and with cardiovascular death in the entire cohort (NT‐proBNP: HR, 2.42; 95% CI, 1.87–3.12; hs‐cTnI: HR, 1.53; 95% CI, 1.30–1.79), patients with CKD (NT‐proBNP: HR, 2.76; 95% CI, 1.88–4.07; hs‐cTnI: HR, 1.70; 95% CI, 1.34–2.15), and patients without CKD (NT‐proBNP: HR, 2.31; 95% CI, 1.56–3.43; hs‐cTnI: HR, 1.42; 95% CI, 1.12–1.80) (Figure S6). Those of hs‐cTnI were also significantly associated with 5P‐MACE in the entire cohort, patients with CKD and patients without CKD, while those of NT‐proBNP were significantly associated with 5P‐MACE in the entire cohort and in patients with CKD, but not in patients without CKD (Figure S7).

Among other biomarkers, the serum levels of hs‐CRP, cystatin C, and PlGF (as continuous variables) were significantly associated with all‐cause death and 5P‐MACE in the entire cohort and in patients without CKD. However, the serum levels of hs‐CRP, but not those of cystatin C or PlGF, were significantly associated with all‐cause death, and those of hs‐CRP and cystatin C, but not those of PlGF, were significantly associated with 5P‐MACE in patients with CKD (Figures S5 and S7). Only the serum levels of cystatin C were significantly associated with cardiovascular death in the entire cohort and in patients without CKD, but not in patients with CKD (Figure S6). UACR, either as a continuous variable or at values of 30 mg/g or more, was not significantly associated with all‐cause death, cardiovascular death, or 5P‐MACE in patients with CKD (Figures S5 through S7).

Discrimination, Reclassification, and Calibration

Table 3 shows the incremental predictive performance of selected biomarkers for 3P‐MACE in the entire cohort, patients with CKD and patients without CKD. The C statistics for 3P‐MACE by the model with potential clinical confounders (base model) were 0.712 in the entire cohort, 0.673 in patients with CKD and 0.735 in patients without CKD. The addition of sFlt‐1 (as a continuous variable) to the base model significantly improved the prediction of 3P‐MACE in the entire cohort (P=0.006 for NRI, P=0.027 for IDI) and in patients without CKD (P=0.032 for NRI, P<0.050 for IDI), but not in patients with CKD (P=0.093 for NRI, P=0.169 for IDI). On the other hand, the addition of the top tertile (versus tertiles 1 and 2) of sFlt‐1 to the base model significantly improved the prediction of 3P‐MACE in the entire cohort (P<0.001 for NRI, P=0.014 for IDI) and in patients with CKD (P=0.002 for NRI, P=0.013 for IDI), but not in patients without CKD (P=0.022 for NRI, P=0.163 for IDI).

Table 3

Incremental Predictive Performance of Selected Biomarkers for 3‐Point MACE in the Entire Cohort, Patients With CKD, and Patients Without CKD

Subgroups and prediction models	C statistics	∆C statistics	Continuous NRI, 95% CI	P value	IDI, 95% CI	P value
Entire cohort
Base model*	0.712	…	…		…
Base+sFlt‐1 †	0.724	0.012	0.227 (0.067 to 0.388)	0.006	0.005 (0.001 to 0.009)	0.027
Base+sFlt‐1 (top tertile) †	0.721	0.009	0.310 (0.150 to 0.470)	<0.001	0.004 (0.001 to 0.006)	0.014
Base+NT‐proBNP †	0.748	0.037	0.384 (0.225 to 0.543)	<0.001	0.021 (0.011 to 0.031)	<0.001
Base+hs‐cTnI †	0.751	0.039	0.393 (0.233 to 0.554)	<0.001	0.011 (0.005 to 0.017)	<0.001
CKD
Base model*	0.673	…	…		…
Base+sFlt‐1 †	0.673	0.000	0.186 (−0.031 to 0.402)	0.093	0.002 (−0.001 to 0.005)	0.169
Base+sFlt‐1 (top tertile) †	0.686	0.014	0.333 (0.117 to 0.548)	0.002	0.007 (0.001 to 0.012)	0.013
Base+NT‐proBNP †	0.719	0.046	0.484 (0273 to 0.696)	<0.001	0.023 (0.010 to 0.036)	<0.001
Base+hs‐cTnI †	0.714	0.041	0.538 (0.325 to 0.751)	<0.001	0.016 (0.006 to 0.025)	0.001
Non‐CKD
Base model*	0.735	…	…		…
Base+sFlt‐1 †	0.758	0.023	0.264 (0.023 to 0.504)	0.032	0.014 (0.000 to 0.028)	0.050
Base+sFlt‐1 (top tertile) †	0.743	0.008	0.282 (0.041 to 0.523)	0.022	0.003 (−0.001 to 0.007)	0.163
Base+NT‐proBNP †	0.776	0.041	0.410 (0.171 to 0.649)	<0.001	0.015 (0.002 to 0.029)	0.027
Base+hs‐cTnI †	0.777	0.042	0.371 (0.130 to 0.611)	0.003	0.006 (−0.002 to 0.014)	0.146

Follow‐up results are truncated after 3 years. The biomarkers are natural log‐transformed and are modeled as continuous variables unless otherwise indicated. The ΔC statistic, continuous NRI and IDI show the change in model performance from the base model. hs‐cTnI indicates high‐sensitivity cardiac troponin I; IDI, integrated discrimination improvement; MACE, major adverse cardiovascular events; NRI, net reclassification improvement; NT‐proBNP, N‐terminal pro‐brain natriuretic peptide; and sFlt‐1, soluble fms‐like tyrosine kinase 1.

The base model is based on age, sex, body mass index, hypertension, dyslipidemia, diabetes, current smoking, estimated glomerular filtration rate, the Gensini score, previous myocardial infarction, previous stroke, previous heart failure hospitalization, atrial fibrillation, anemia, antihypertensive drug use, statin use, and aspirin use.

The change of model performance was evaluated against the base model.

Table S7 shows the incremental predictive performance of selected biomarkers for all‐cause death in the entire cohort, patients with CKD, and patients without CKD. The addition of sFlt‐1 (as a continuous variable) to the base model significantly improved the prediction of all‐cause death in the entire cohort (P<0.001 for NRI, P=0.024 for IDI), but not in patients with CKD (P=0.004 for NRI, P=0.064 for IDI) or patients without CKD (P<0.001 for NRI, P=0.132 for IDI). However, the addition of the top tertile (versus tertiles 1 and 2) of sFlt‐1 to the base model significantly improved the prediction of all‐cause death in the entire cohort (P<0.001 for NRI, P=0.023 for IDI) and in patients with CKD (P=0.001 for NRI, P=0.019 for IDI), but not in patients without CKD (P<0.001 for NRI, P=0.133 for IDI).

Table S8 shows the incremental predictive performance of selected biomarkers for cardiovascular death in the entire cohort, patients with CKD, and patients without CKD. The addition of sFlt‐1 (as a continuous variable) to the base model significantly improved the prediction of cardiovascular death in the entire cohort (P=0.002 for NRI, P=0.042 for IDI), but not in patients with CKD (P=0.050 for NRI, P=0.306 for IDI) or patients without CKD (P=0.023 for NRI, P=0.070 for IDI). The addition of the top tertile (versus tertiles 1 and 2) of sFlt‐1 to the base model significantly improved the prediction of cardiovascular death in the entire cohort (P<0.001 for NRI, P=0.013 for IDI), but not in patients with CKD (P=0.003 for NRI, P=0.066 for IDI) or patients without CKD (P=0.008 for NRI, P=0.090 for IDI).

Notably, the addition of either NT‐proBNP or hs‐cTnI significantly improved the prediction of 3P‐MACE, all‐cause death, and cardiovascular death not only in the entire cohort, but also in patients with CKD (Table 3 and Tables S7, S8). Moreover, the addition of either NT‐proBNP or hs‐cTnI significantly improved the prediction of 5P‐MACE in patients with CKD (Table S9). Calibration of the models with or without each biomarker showed no evidence of lack of fit.

DISCUSSION

This is the first dedicated and large‐scale prospective cohort study to demonstrate that higher levels of NT‐proBNP and hs‐cTnI, but not those of sFlt‐1 or UCAR, independently predicted cardiovascular events and mortality in patients with suspected or known CAD and concomitant CKD. The strengths of our investigation include the large sample size, multi‐center prospective design, inclusion of both CKD and non‐CKD patient data, and high follow‐up rate (98.9%).

Many clinical studies have evaluated novel biomarkers due to the importance of improving risk stratification and supporting clinical decision making in patients with CKD. To date, however, only a few biomarkers, including eGFR and proteinuria, have been approved for large‐scale clinical application.

NT‐proBNP has previously been shown to be predictive of cardiovascular morbidity and mortality in the general population, and among patients with acute coronary syndrome, those with heart failure, and those with stable CAD. , , , , Pre‐proBNP is synthesized within the cardiac myocytes in response to ventricular wall stress and stretch. After removal of a signaling peptide within the cytosol, proBNP is further cleaved into an inactive form (NT‐proBNP) and an active form (brain natriuretic peptide [BNP]) at the time of release from the myocytes or in the circulation. NT‐proBNP is more stable, with a longer half‐life, and may be a better biomarker for chronic volume expansion or stress than BNP. The clearance of NT‐proBNP is predominantly renal, and NT‐proBNP levels are inversely correlated with eGFR, and are often elevated in asymptomatic patients with CKD. , , Since circulating NT‐proBNP levels are mostly determined by the cardiac myocyte production and renal clearance, the coexistence of ventricular wall stress/stretch and renal insufficiency obscures the implications of elevated NT‐proBNP: the largest determinant of NT‐proBNP elevation depends on which is more severe, the ventricular wall stress/stretch or renal insufficiency. In any case, higher NT‐proBNP levels can be a cardiorenal comprehensive prognostic biomarker, because both abnormal cardiac ventricular stress/stretch and renal insufficiency are associated with the risks of cardiovascular events and mortality.

hs‐cTnI has also been shown to be a predictor of cardiovascular morbidity and mortality in the general population, and among patients with acute coronary syndrome, those with heart failure, and those with stable CAD. , , Cardiac troponin I (cTnI) and T (cTnT) are components of the contractile apparatus of myocardial cells and are expressed almost exclusively in the heart. , An increase in cTnI values has not been reported to occur following injury to non‐cardiac tissues, whereas injured skeletal muscle expresses proteins that are detected by the cTnT assay. cTnI and cTnT are the preferred biomarkers for the evaluation of myocardial injury, and hs‐cTnI and hs‐cTnT assays are recommended for routine clinical use. , hs‐cTnI was significantly and inversely associated with eGFR. Increased hs‐cTnI levels were common in CKD without acute coronary syndrome, and are influenced by both underlying cardiac and renal disease : troponin elevation does not necessarily indicate acute ischemia from coronary atherosclerosis but may be due to decreased renal clearance or chronic myocardial injury. Our findings that NT‐proBNP and hs‐cTnI serve as prognostic biomarkers even in the presence of CKD in high‐risk patients with suspected or known CAD have extended the findings of the previous study in the general population.

The Kidney Disease: Improving Global Outcomes (KDIGO) Controversies Conference reported that albuminuria (defined as an UACR of 30 mg/g or more) was independently associated with all‐cause death and cardiovascular death in the general population, among high‐risk patients (ie, those with hypertension, diabetes, clinical cardiovascular disease, or a history of kidney disease), and among patients with CKD. Albuminuria was also independently associated with incident atherosclerotic vascular disease events and death in patients with CKD without a history of cardiovascular disease. By contrast, UACR was not independently associated with cardiovascular events or mortality in the present study. Among the very high‐risk patients with overlapping risks of CAD and CKD, the impact of UACR on poor prognosis may be relatively reduced. Further investigation is necessary to confirm these findings.

sFlt‐1 has been shown to be independently associated with adverse outcomes among patients with chronic heart failure , and among patients with CKD stages 2–4. PlGF, a specific ligand for Flt‐1, has been suggested to be independently associated with all‐cause death in the general population ; with cardiovascular events among patients with suspected and definite acute coronary syndrome , ; with all‐cause death and cardiovascular death among patients with acute heart failure ; and with all‐cause death and cardiovascular events among patients with CKD. The sFlt1/PlGF ratio has been shown to predict adverse outcomes among women with suspected preeclampsia , and among pregnant women with hypertension. In the present study, sFlt‐1 was independently associated with hard end points (3P‐MACE, all‐cause death, and cardiovascular, and cardiovascular death), but not with soft end points (5P‐MACE), among patients with suspected or known CAD (ie, in the entire cohort). However, these associations were attenuated in the subgroup of CKD. PlGF was independently associated with all‐cause death and 5P‐MACE, but not with 3P‐MACE or cardiovascular death, in the entire cohort. These associations were attenuated in the subgroup of CKD. The sFlt‐1/PlGF ratio was independently associated with cardiovascular death, but not with 3P‐MACE, all‐cause death or 5P‐MACE, in the entire cohort. This association was attenuated in the subgroup of CKD as well. These findings suggest that sFlt‐1, PlGF, and the sFlt‐1/PlGF ratio were less predictive for poor prognosis in patients with CKD than in patients without CKD among patients with suspected or known CAD. However, the addition of sFlt‐1 to the base model with potential clinical confounders significantly improved the prediction of 3P‐MACE, all‐cause death, and cardiovascular death, but not that of 5P‐MACE, in the entire cohort. Further investigation will clarify whether there are subgroups in which sFlt‐1 shows better prognostic utility than established biomarkers such as NT‐proBNP and hs‐cTnI.

We recently demonstrated that serum levels of VEGF were not independently associated with all‐cause death, cardiovascular death, or 3P‐MACE in patients with suspected or known CAD. In the present study, we observed similar results in the entire cohort, patients with CKD, and patients without CKD.

hs‐CRP has been shown to be predictive of cardiovascular events and mortality in the general population, among patients with acute coronary syndrome, and among patients with stable CAD. In the present study, hs‐CRP was significantly associated with: 3P‐MACE in patients without CKD, but not in the entire cohort or in patients with CKD; cardiovascular death in the entire cohort and in patients without CKD, but not in patients with CKD; and all‐cause death and 5P‐MACE in the entire cohort, patients with CKD, and patients without CKD. The addition of hs‐CRP to the base model with potential clinical confounders significantly improved the prediction of all‐cause death, but not that of cardiovascular death or 5P‐MACE, in the entire cohort, patients with CKD, and patients without CKD. Thus, hs‐CRP is a very powerful predictor of all‐cause death irrespective of the presence or absence of CKD, but seems to be less predictive of cardiovascular events and cardiovascular death than NT‐proBNP and hs‐cTnI in patients with suspected or known CAD, especially in the presence of CKD.

In the present study, cystatin C was significantly associated with 3P‐MACE and all‐cause death even after adjustment for potential clinical confounders, including creatinine‐based eGFR, in the entire cohort and in patients without CKD, but not in patients with CKD. Cystatin C was independently associated with 5P‐MACE in the entire cohort, patients with CKD, and patients without CKD. The addition of cystatin C to the model with potential clinical confounders including creatinine‐based eGFR further improved the prediction of 5P‐MACE in the entire cohort and in patients without CKD, but not in patients with CKD. In contrast, the addition of cystatin C did not further improve the prediction of any hard end points (3P‐MACE, all‐cause death, or cardiovascular death) in patients with suspected or known CAD, regardless of the presence or absence of CKD. These findings may suggest that cystatin C is a powerful predictor of atherosclerotic cardiovascular events, but is less predictive of hard cardiovascular events and mortality than NT‐proBNP and hs‐cTnI in patients with suspected or known CAD, especially in the presence of CKD.

The sFlt‐1 levels in the top tertile of the present study were >121.17 pg/mL. A previous study reported that the sFlt‐1 levels in the top tertile among patients with CAD, including both patients with stable angina pectoris and those with acute coronary syndrome, were >160.0 pg/mL. In the same study, the sFlt‐1 levels were higher in patients with acute coronary syndrome than in patients with stable angina pectoris. The present study included patients with suspected or known CAD undergoing elective coronary angiography, but not those with acute coronary syndrome requiring urgent coronary intervention. Thus, the difference in the sFlt‐1 levels in the top tertile could be explained by the inclusion rate of patients with acute coronary syndrome. The NT‐proBNP and hs‐cTnI levels in the top tertiles in the present study were >352 pg/mL and >11.8 pg/mL, respectively. A previous study reported that the NT‐proBNP levels in the third quartile among patients with stable CAD were 170 to 455 pg/mL. These values of NT‐proBNP are similar to those in the present study. Another study showed that the hs‐cTnI levels in the third quartile among patients with stable CAD were 4.6 to 7.3 pg/mL in men and 4.0 to 6.3 pg/mL in women. Although these values of hs‐cTnI are lower than those in the present study, the difference could be explained by the inclusion of a small number of unstable patients with CAD not requiring urgent coronary intervention in the present study.

Limitations

First, we did not include patients with stages 1 to 2 CKD as defined by albuminuria with preserved glomerular filtration rate in the CKD subgroup, because we had no collected data on UACR in patients with eGFR ≥60 mL/min per 1.73 m2. We also did not include patients with severe CKD who had not been introduced to dialysis and would be discouraged from using contrast media. Second, we had no collected cardiovascular imaging data, such as echocardiography (especially, left ventricular ejection fraction and valvular disease), cardiovascular magnetic resonance, computed tomography, intravascular ultrasound/optical coherence tomography, or nuclear imaging data. Third, we had no collected data of a history of COPD. Fourth, this was an observational study, and other unmeasured confounding factors may have existed. Finally, because the EXCEED‐J study cohort consists exclusively of Asian individuals with suspected or known CAD, our results may not be generalizable to general Asian populations, or to other ethnic groups.

CONCLUSIONS

Nevertheless, our results clearly demonstrate that higher serum levels of NT‐proBNP and hs‐cTnI, but not those of sFlt‐1 or UACR, independently augmented the prediction of both cardiovascular events and mortality achieved by potential clinical confounders in patients with suspected or known CAD and concomitant CKD undergoing elective coronary angiography.

Sources of Funding

The EXCEED‐J study is supported by a Health Labour Sciences Research Grant (2013–2014), AMED (2015–2017, Grant Number JP17ek0210008) and a Grant‐in‐Aid for Clinical Research from the National Hospital Organization (2018–2020).

Disclosures

None.

Appendix S1

Data S1

Tables S1–S9

Figures S1–S7

References 64, 65

Click here for additional data file.