Biomarkers as predictors of sudden cardiac death in coronary artery disease patients with preserved left ventricular function (ARTEMIS study)

Aims Biomarkers have shown promising results in risk assessment of cardiovascular events. Their role in predicting the risk of sudden cardiac death (SCD) is not well established. We tested the performance of several biomarkers in risk assessment for SCD in patients with coronary artery disease (CAD) and preserved left ventricular function. Methods and results The study population consisted of 1,946 CAD patients (68% male; mean age 66.9±8.6 yrs; type 2 diabetes (T2D) 43%) enrolled in the ARTEMIS study. The study subjects underwent examinations with echocardiography and measurement of several biomarkers. The primary endpoint of the study was SCD. During the mean follow up of 76±20 months 50 patients experienced SCD. Elevated high sensitive CRP (hs-CRP, p = 0.001), soluble ST2 (sST2, p<0.001), B-type natriuretic peptide (BNP, p<0.001), and highly sensitive TroponinT (hs-TnT, p<0.001) predicted the occurrence of SCD in univariate analysis. Using the optimal cutoff points, elevated sST2 (≥27.45ng/mL; hazard ratio [HR] 2.7; 95%CI 1.4–5.1, p = 0.003) and hs-TnT (≥15 ng/mL; HR 2.9; 95% CI 1.5–5.6, p = 0.002) were the strongest predictors of SCD followed by hs-CRP (HR 2.4; 95% CI 1.3–4.4, p = 0.004) and BNP (HR 1.9; 95% CI 1.0–3.7, p = 0.046) in adjusted analysis. Combination of elevated hs-TnT and sST2 resulted in adjusted HR of 6.4 (95% CI 2.6–15.5, p<0.001). Conclusion Elevated sST2 and hs-TnT predict the occurrence of SCD among patients with CAD and preserved left ventricular function. The association between sST2, hs-TnT and SCD may be explained by an ongoing myocardial apoptosis followed by fibrosis leading to vulnerability to malignant arrhythmias.


Conclusion
Elevated sST2 and hs-TnT predict the occurrence of SCD among patients with CAD and preserved left ventricular function. The association between sST2, hs-TnT and SCD may be explained by an ongoing myocardial apoptosis followed by fibrosis leading to vulnerability to malignant arrhythmias. PLOS

Introduction
Coronary artery disease (CAD) is still the most common cause of death in Western society. Even though the prevention and treatment of CAD has improved during the last three decades, the impact of sudden cardiac death (SCD) remains prominent [1][2][3]. The majority of SCDs are due to CAD, but to date only few clinical or biological markers have succeeded in identifying individuals at risk in an otherwise intermediate risk population of stable CAD patients [4]. The risk assessment of SCD in CAD populations has recently concentrated on markers of heart failure and impaired autonomic nervous system. In addition to left ventricular ejection fraction (LVEF), numerous non-invasive ECG-derived methodologies have been proposed to risk stratify patients for SCD. Among them, Holter analysis including heart rate variability and turbulence are commonly recommended for non-invasive risk stratification [5]. Although these markers have been shown to predict SCD, the majority of SCD victims are not identified by these measures. Feasible markers for SCD risk assessment in intermediate risk populations such as CAD patients with preserved left ventricular function are lacking. B-type natriuretic peptide (BNP) and high-sensitive C-reactive protein (hs-CRP) have provided some prognostic information in various populations, and recently, highly sensitive troponin (hs-TnT) and markers of cardiac fibrosis, such as sST2 and galectin-3, have shown promising results in risk assessment of cardiovascular events in patients with CAD and congestive heart failure (CHF) [6][7][8][9][10][11][12]. The occurrence of ventricular arrhythmias is dependent on re-entry formation, and the key factors in this phenomenon are areas of slow conduction. The current hypothesis assumes that these areas are a result of accumulation of fibrosis caused by a myocardial infarction or, more commonly, by repeated episodes of transient ischemia. In many cases, this accumulation is heterogeneous in nature and does not form a stable local scar which does not conduct electricity, but an area where conduction is slow. This fibrosis accumulation is ultimately assumed to lead to reduction of systolic function, but the majority of patients develop concomitant muscle hypertrophy which delays the occurrence of pump failure. On the other hand, there is a risk for arrhythmias. The aim of the study was to test the performance of several biomarkers in risk assessment for SCD in CAD patients with preserved LV function. Ultimately, the goal was to find biomarkers which would coincide with the pathological pathway of fibrosis accumulation and arrhythmia risk elevation.

Methods
The study population consisted of 1,946 patients with stable CAD from the ARTEMIS (Innovation to Reduce Cardiovascular Complications of Diabetes at the Intersection) study conducted in the Division of Cardiology of Oulu University Hospital (Oulu, Finland), ClinicalTrials.gov (Identifier NCT01426685). The goal of ARTEMIS study is to study the prognostic value of several traditional and novel cardiovascular risk markers in patients with stable CAD during a 5-year follow-up period. The design has been previously described in detail [9,13]. The patients were recruited to a thorough risk assessment from a consecutive series of patients referred to coronary angiography in the Division of Cardiology of Oulu University Hospital (Oulu, Finland) 3 to 6 months after the angiography. Most patients had either percutaneous coronary angioplasty or coronary artery bypass during or right after the index procedure. Almost half of the patients had experienced an acute coronary event >3 months before the enrollment into the risk profile study. Exclusion criteria were: age <18 years or >85 years, New York Heart Association or Canadian Cardiovascular Society class IV, significant valvular disease, permanent pacemaker or permanent or planned implantable cardioverter defibrillator implantation (LV ejection fraction <35%), end-stage renal failure needing dialysis, life expectancy <1 year. Patients who were psychologically or physically (due to any other illness) unfit for participation in the study were also excluded from this study. The study was performed according to the Declaration of Helsinki, and the local committee of research ethics of the Northern Ostrobothnia Hospital District approved the protocol, and all the subjects gave written informed consent. Laboratory samples were obtained after 12-hour overnight fast using standardized methods. Subjects without known diabetes had an oral glucose tolerance test. T2D, impaired glucose tolerance, and impaired fasting glycemia were verified according to current criteria of the World Health Organization [14]. Patients who were on antidiabetic medication based on previous diagnosis of diabetes were also classified as having T2D. Blood samples and urine samples were obtained for the analysis of renal function (albumin-to-creatinine ratio), inflammation markers (hs-CRP), lipids, glycated hemoglobin (HbA1c), and cardiac markers (BNP and hs-TnT), and they were analyzed by the hospital laboratory. The concentrations of BNP and hs-TnT were determined from plasma samples (ADVIA Centaur XP, Siemens Healthcare Diagnostics and MODULAR ANALYTICS, Roche Diagnostics, respectively). The concentrations of hs-CRP, galectin-3, and sST2 were determined from serum samples (BN Prospec System, Siemens Healthcare Diagnostics; Human ST2/IL-1 R4 Quantikine ELISA, R&D Systems Inc., Minneapolis, Minnesota; BG Medicine, Waltham, Massachusetts, respectively). Echocardiography was performed according to the American Society of Echocardiography guidelines. Left ventricular (LV) mass was calculated using the formula recommended by the American Society of Echocardiography (LV mass = 0.8 × [1.04 {(LVIDd + PWTd + SWTd) 3 -(LVIDd) 3 }] + 0.6 g). LV mass index was calculated by dividing LV mass by body surface area. Patients with LVEF <35% who had an indication for prophylactic implantable cardioverter-defibrillator were excluded from the consecutive series. SYNTAX Score was calculated by three experienced interventional cardiologists using the Web-based calculator version 2.11 on the SYNTAX Score Web site (www.syntaxscore.com).
The primary endpoint of the ARTEMIS study is SCD, including aborted cardiac arrests. The occurrence of SCD and aborted cardiac arrests was defined by an endpoint committee (HVH and MJJ) from death certificates including autopsy reports, hospital records, and contacts with next of kin. Only probable or possible sudden deaths were classified as SCDs, including witnessed sudden death within one hour after the onset of symptoms, or within 24 hours of when the victim was last seen alive without any evidence of non-cardiac cause of death at autopsy. The secondary endpoint of the study was defined as all-cause mortality (ACM).

Statistical analyses
Values are expressed as mean (+/-SD), median (interquartile range) or the number of subjects (%). Two-tailed t-test for independent samples, Mann-Whitney U-test (if non-Gaussian distribution of the data) or chi-square were used to compare the survivors with the groups of SCD and ACM. Receiver operating characteristic (ROC) was used to establish the optimal cutoff value for the novel biomarkers using SCD as endpoint. The cutoff was set to the value that provided the maximum sum of sensitivity and specificity, sensitivity being at least 20% but less than 50% of patients having positive result. The prognostic significances of biomarkers were assessed according to the recommendations by Hlatky et al. [15]. These recommendations suggest that when a new risk marker is introduced in a system of multiple risk factors the ability to improve the model (which is generated with standard risk factors) should be assessed. In our study, attempts to assess the additive risk value of biomarkers have been made with C-statistics and integrated discrimination improvement (IDI) calculation. Hazard ratios (HR) with 95% confidence intervals (CI) were calculated by univariate Cox regression analysis for each risk marker. Thereafter, the predictive value of each biomarker was assessed in multivariate Cox regression where age, sex, body mass index, Canadian Cardiovascular Society grading of angina pectoris, left ventricular ejection fraction, albumin-creatinine ratio, estimated glomerulus filtration rate (eGFR), HbA1c, and T2D were entered in the model as continuous variables when applicable. The covariates were selected based on clinical significance and relevant differences found in characteristics between the groups with and without SCD. Multimarker Cox regression models were also established for the most significant biomarkers followed by further adjustments for potential confounders. Competing risks survival analyses were also conducted using approaches of SCD vs.non-sudden cardiac death.
Due to the varying follow-up time, discrimination and reclassification analyses were based on individual time-dependent estimated risk for SCD that was obtained by multivariate Cox regression analyses. The added discrimination of the risk markers was assessed by the C-index and the integrated discrimination index (IDI). Categorical and continuous net reclassification index (NRI; risk levels: 0-<6, 6-<20 and !20%) were also calculated [16]. Kaplan-Meier analysis with log rank analysis was used to describe the SCD-free survival in low-and high-risk groups. The data were analyzed using SPSS software (IBM SPSS Statistics 21, IBM Corp., New York, USA) and R Statistics (3.3.1, The R Foundation for Statistical Computing, Vienna, Austria). A p-value <0.05 was considered statistically significant.

Results
The characteristics of the patients are presented in Table 1. Mean follow-up time was 76±20 months during which 205 deaths occurred of which 99 were determined to be of cardiac cause. Among cardiac deaths 50 SCDs, including 8 aborted cardiac arrests, were observed. Non-sudden cardiac death occurred in 44 subjects, which include pump failure and heart failure deaths. The concentrations of biomarkers, except for galectin-3, were higher in the patients who suffered SCD than those without the primary endpoint (Table 1).
Elevated sST2 and hs-TnT were the most powerful predictors of the primary endpoint in the univariate analyses and also in multivariate analyses with 2.7-and 2.9-fold risk for SCD (p = 0.003 and p = 0.002), respectively ( Table 2). They also were the most significant risk markers in the crude and adjusted multimarker Cox regression, being significantly associated with SCD independently of hs-CRP and BNP (Table 3). When the two most powerful biomarkers, i.e. sST2 and hs-TnT, were combined, elevated concentrations in both biomarkers yielded a hazard ratio (HR) of 6.4 (2.6-15.5, p<0.001) compared to those with both at normal level in multivariate Cox analysis (Table 2). Hs-TnT, hs-CRP, BNP and sST2 were also significant predictors of all-cause mortality in multivariate analysis (p<0.001 for all, Table 2).
Elevated hs-TnT and sST2 were associated with SCD regardless of competing risk for non-sudden cardiac death in univariate analysis (HR

Discussion
This study showed that measurement of biomarkers from blood samples provides information on SCD risk among the patients with CAD and preserved left ventricular function. Soluble ST2 and hs-TnT were the most powerful predictors after adjustments with several clinical and laboratory variables as well as LVEF. A combination of elevated concentrations of both sST2 and hs-TnT yielded high predictive accuracy in SCD prediction in this population. The present study aimed to focus on the risk assessment of SCD among CAD patients with preserved left ventricular function, since the majority of SCD victims have a heart disease with normal or mildly impaired cardiac function [17,18]. Furthermore, the majority of prior risk stratification studies have focused on patients with CHF or on those with a recent acute myocardial infarction. This has raised the question of whether any of the risk stratification methods can be improved in order to develop appropriate preventive strategies for this large number of potential victims of SCD included in the present study.

Biomarkers in risk prediction
All biomarkers included in this study have provided prognostic information in various patient populations in prior studies, but the utility of these variables in prediction of SCD risk has not been thoroughly studied. To our knowledge, BNP is the only biomarker that has been shown to predict SCD after acute myocardial infarction [19]. The other biomarkers, such as sST2, hs-TnT, galectin-3 and hs-CRP, have predicted all-cause and cardiac mortality among patients with heart failure, stable CAD and even in general population [6,8,10,20,21], but only one of these studies focused specifically on the occurrence of SCD in patients with CHF [11] while none focused on the occurrence of SCD in patients with preserved left ventricular function. Multimarker strategy, i.e. combination of two or more risk variables, has been shown to improve the prediction of cardiovascular events in previous studies [10,22]. Combination of elevated sST2 and hs-TnT in the present population also provided the most powerful information on the risk of SCD.

Potential mechanisms of association between sST2, hs-TnT and SCD
Soluble ST2 and hs-TnT were the strongest predictors of SCD in the present study. Soluble ST2 is a member of the interleukin-1 receptor family and circulating sST2 concentrations are believed to reflect cardiovascular stress and cardiac fibrosis [23,24]. Soluble ST2 is elevated in CHF, older age and in diabetes [25]. It has also been shown that genetic factors explain up to 45% of the sST2 concentrations in general population [26]. Hs-TnT is a novel marker reflecting ongoing myocyte death and apoptosis without any evidence of an acute ischemic event. Elevated levels of hs-TnT have been associated with higher risk of incident coronary event, hospitalization due to heart failure, cardiac mortality, cardiovascular mortality, stroke and cancer [6][7][8]27,28]. In our prior pilot study of the ARTEMIS population, we found that hs-TnT is a strong predictor of a composite endpoint of cardiac death and hospitalization for heart failure during a 2-year follow-up [9].
The mechanism why both sST2 and hs-TnT predicted SCD remains speculative. These two biomarkers obviously reflect partly different aspects of cardiovascular stress and tissue damage leading to untoward cardiac events, either progressive heart failure or occurrence of SCD even without prior evidence of left ventricular systolic dysfunction. It can be hypothesized that elevated hs-TnT is a marker of ongoing myocyte loss and elevated sST2 reflects the consequent cardiac replacement fibrosis as a result of cell death, which eventually creates a substrate for fatal arrhythmia. Fibrotic scarring has been shown to correlate strongly with an increased incidence of arrhythmias and SCD [29]. Our results thus suggest that CAD patients who would otherwise be at low to intermediate risk of SCD can be included in a higher risk group with biomarkers of inducted replacement fibrosis accumulation. This paradigm explains partly why some patients with stable CAD and normal LV function enter a downslope in prognosis through continuous cell death and formation of replacement fibrosis. This scenario would eventually end up in decreased LV systolic function, but we do not currently often encounter CAD patients with decreasing LV function. On the other hand, the number of patients with some form of heart failure with preserved EF is increasing. Thus more patients with accumulated fibrosis are at risk of lethal arrhythmias than lethal pump failure.

Potential limitations
The ARTEMIS study design including a relatively large number of patients with T2D is a potential limitation of the present study. The power of the study does not allow risk prediction separately for non-diabetic patients. Therefore, it remains uncertain whether the present results are applicable only to CAD patients with T2D but not to non-diabetic patients. Further studies should focus also on non-diabetic patients with preserved left ventricular function. A larger sample size is needed for this purpose, since the incidence of SCD is much smaller in non-diabetic patients with CAD as compared to their diabetic counterparts [30]. It is clear that these results are not strong enough for interventional therapy with ICD, but the hypothesis of ongoing fibrosis accumulation in CAD patients would hopefully result in more precise examination methods and more sophisticated intervention methods in the future. Additionally, in our analysis we used low sensitivity ELISA assays which are not approved for clinical diagnostics and can be viewed as a potential limitation. Also the number of subjects in this study prevented us from dividing the population to training and test cohorts. Unfortunately we do not have sufficient interim information on EF% development in this study and thus cannot show whether the biomarker levels are specifically associated with SCD or deterioration of LV systolic function.

Conclusions
The present study showed that analysis of serum biomarkers can provide important information on the risk of SCD in CAD patients with preserved left ventricular systolic function. The multimarker strategy appeared very useful in this respect. Future research should focus on preventive strategies among patients with elevated levels of biomarkers, especially both hs-TnT and sST2. Examples include angiotensin-converting enzyme inhibitors, aldosterone antagonists, nesiritide, emerging antifibrotic therapies [31], or perhaps novel pharmacological antidiabetic strategies [32].