OUP user menu

Predicting future shocks in implantable cardioverter defibrillator recipients: the role of biomarkers

Emmanuel M. Kanoupakis, Emmanuel G. Manios, Panos E. Vardas
DOI: http://dx.doi.org/10.1093/europace/eup235 1434-1439 First published online: 3 October 2009


Simple clinical and laboratory indexes have been identified as predictors of arrhythmic events in implantable cardioverter defibrillator patients. Biomarkers, which are playing a growing role in the prognosis and treatment of patients with heart failure, could provide an auxiliary tool in this context, given that their measurement is now easy and widely available.

  • Implantable cardioverter defibrillator
  • Predictors of shocks
  • Biomarkers
  • Natriuretic peptides

The efficacy of implantable cardioverter defibrillators (ICDs) for terminating ventricular arrhythmias is very well established. They represent the treatment of choice in survivors of cardiac arrest or unstable ventricular tachycardias (VTs),13 while a number of recently published randomized controlled trials confirm their superiority over a treatment strategy with anti-arrhythmic drugs alone for the primary prevention of sudden cardiac death (SCD) in certain high-risk groups with left ventricular dysfunction, due to either coronary artery disease or non-ischaemic dilated cardiomyopathy.48 These trials focused on primary prevention, selecting patients for ICD implantation on the basis of a reduced left ventricular ejection fraction (LVEF); although a significant improvement in survival was achieved, it was with relatively low therapeutic efficiency (15–17 ICDs/life saved). Nonetheless, even in secondary prevention trials, approximately one-third of patients had not experienced spontaneous sustained ventricular arrhythmias at 2 years.1 This means that not all patients, even in these high-risk groups, have the same risk of malignant arrhythmias during the limited duration of trials.

In fact, a sizeable percentage of patients who meet the current clinical criteria for ICD implantation will obtain no advantages from these devices, and some of them will experience some potential side effects, including inappropriate shocks,9 device recalls,10,11 worsening heart failure,12,13 infections,14 and even psychosocial problems related to the patients' acceptance of their device.15,16 In light of all these drawbacks, the better identification of subgroups that would benefit most and least from ICD implantation remains a crucial challenge. Moreover, given the high economic cost of these devices, additional risk stratification might improve patient selection for ICD therapy and thereby enhance its cost-effectiveness.1719 According to the most recent ACC/AHA/HRS Task Force on Practice Guidelines, this issue remains as one of the main areas suggested as in need of further research.20

The heterogeneity of risk and the need for better stratification

The publication of randomized clinical trials led to a dramatic explosion in implantable defibrillator devices. However, in inverse relation to the number of implantable devices there was a proportionate decrease in the percentage of patients who received appropriate therapy from their defibrillator. While in the initial studies of primary prevention this percentage was of the order of 50–55%,4,5 in the most recent studies it had fallen to 25–30%.6,8 This caused a good deal of concern, and various sub-analyses were organized in an attempt to detect factors, apart from the degree of left ventricular dysfunction, that could predict the appearance of malignant arrhythmias in these patients and would thus help in the best selection of patients for ICD implantation.2124

In our continuing efforts to identify these factors we can make use of the information provided by those patients who have already received an ICD. From this huge reservoir of patients we may distinguish various clinical or other parameters that are well correlated from the statistical point of view with the occurrence of shock or VT/ventricular fibrillation (VT/VF), namely the ICD discharge endpoint data. The quest to identify such risk factors for VT/VF occurrence or appropriate firings of the ICD could significantly improve post-ICD implantation care, and could aid in the prevention, reduction, or avoidance of arrhythmic events and hence painful shocks. In addition, such information could influence the clinical decision as to whether or not to replace the device after battery exhaustion. Finally, it could contribute considerably to the progress of risk stratification and to the best choice of candidates for ICDs, since our information would be drawn from a clearly longer follow-up period than that used in randomized controlled trials. Besides, it is well known that in patients with ICDs appropriate therapies and complications may occur even years after device implantation.25 This ongoing risk of ventricular arrhythmias, for several years post-implant, may to some degree reflect the random nature of their occurrence, or could be the result of a dynamic arrhythmic substrate, affected by clinical events or progressive conditions such as myocardial infarction or worsening heart failure or medication variation. Once these parameters are recognized, they can be evaluated prospectively for their predictive accuracy and can later be used in clinical practice for the selection of patients who are at high risk.21

Factors affecting appropriate implantable cardioverter defibrillator discharge

Several clinical, laboratory, echocardiographic, and electrophysiological tests have been developed that, unlike LVEF, more directly reflect arrhythmia substrates and might better identify patients who benefit from ICD insertion.

A typical finding emerged from the MADIT-II trial, where although patients with a previous myocardial infarction and LVEF ≤ 30% benefited from ICD implantation, only 24% of patients with an ICD received appropriate therapy during the course of the trial. The most powerful predictor of appropriate ICD therapy was interim hospitalization for heart failure or for coronary events. Other factors that have been found to be well correlated with appropriate ICD therapy for VT, VF, or death were the use or not of drugs (non-use of β-blockers, use of digitalis), obesity (body mass index ≥30 kg/m2), simple laboratory indexes (blood urea nitrogen ((BUN) >25 mg/dl), and of course the patient's symptoms (New York Heart Association (NYHA) functional class >II).22

In contrast, the role of inducibility as an independent predictor of arrhythmic risk was shown to be limited in MADIT-II, because non-inducible subjects in the study had a considerable VT event rate and a higher VF event rate than inducible patients.26 Moreover, non-invasive electrophysiological measures were not noted to be determinants of ICD therapy in the same population.27

In the TOVA study, among 502 patients during a median follow-up of 212 days, 92 patients experienced one or more appropriate ICD discharges. Class III congestive heart failure (CHF) was associated in a statistically significant manner with ICD discharge for VT/VF, even with adjustment for LVEF.28

The PROFIT trial was designed prospectively with exactly the same aim, namely to uncover predictors of VT/VF occurrence in ICD patients in a mixed population, but predominantly secondary prophylactic ICD patients. This study showed that LVEF < 40% is an independent predictor for VT/VF therapy, while its combination with permanent atrial fibrillation and QRS-duration ≥ 150 ms could identify a subgroup of patients with a 100% 2-year risk of shock.29

In another study of heart failure patients who were treated with CRT-D, 73 of 501 patients received appropriate ICD therapy and NYHA class IV symptoms were an independent predictor of shock occurrence.30 It would therefore seem that in severely ill patients, although NYHA class IV is to date not a preferred indication for ICD but rather for CRT, it may be that this benefit could progressively expand the indications for CRT-D implantation in class IV patients. Certainly, further studies are needed to establish this.

The above findings underline the very good relationship that exists between easily accessible clinical and simple laboratory indexes, which mainly reflect the clinical status of patients with heart failure, and malignant arrhythmias or SCD. They also call for a progressive upgrading of the non-invasive and invasive electrophysiological tests used until the present day. Despite the fact that each of these tests may have some predictive accuracy, none has adequate sensitivity and specificity for predicting fatal arrhythmias at the level of the individual.

Biomarkers in heart failure

In recent years, apart from the simple clinical, laboratory, and echocardiographic indexes of heart failure, a group of substances, known as biomarkers, that appear to reflect the complex pathophysiological mechanisms of heart failure has attracted particular clinical interest. Biomarkers are molecules that can be measured in the patient's peripheral blood, with a reasonable cost, and are used in clinical practice for the early detection and diagnosis of sub-clinical disease, for evaluation of the prognosis, and for guiding therapy with a concomitant cost reduction.31,32 The discovery and use of reliable biological markers in heart failure is now viewed as essential, since this syndrome is associated with high mortality and morbidity, and requires prompt diagnosis and the immediate commencement of therapeutic interventions that can improve its prognosis. Apart from heart failure, biomarkers are nowadays also finding clinical application in ischaemic heart disease and are risk factors for the detection of patients who are at high risk of cardiovascular events.33

Given that SCD is responsible for at least a third to one-half of deaths in patients with heart failure, in recent years various studies have appeared in the literature that attempt to answer the question whether these biomarkers could be potential indexes for the prognostic classification of heart failure patients with respect to SCD. This article will review those studies, considering the most commonly used biomarkers and their capacity to predict sudden death or future arrhythmic events in ICD recipients.

Natriuretic peptides

Natriuretic peptides are released primarily from the heart and have hormonal actions in various tissues of the body. In the setting of haemodynamic stress—that is, when the ventricles are dilated, hypertrophied, or subject to myocardial damage—brain natriuretic peptide (BNP) is released as pre-BNP prohormone and then enzymatically cleaved to the biologically active BNP and to the inactive amino-terminal fragment (NT-proBNP).31 Brain natriuretic peptide causes arterial vasodilation, diuresis, and natriuresis, and opposes the activities of the renin–angiotensin–aldosterone system and the sympathetic nervous system.32

Assays for BNP and NT-proBNP are commercially available, but the definition of ‘normal’ values varies, depending upon the clinical feature being detected and the prevalence of that condition in the population being tested. Although their levels are age- and gender-specific, and a number of factors have been associated with higher or lower natriuretic peptide levels, for practical reasons there is a general acceptance of cut-off values.31 Comparing BNP vs. NT-proBNP, the longer half life of NT-proBNP makes it superior to BNP for predicting mortality or morbidity in heart failure patients.34

The clinical use of these substances is mainly in the evaluation of patients with dyspnoea presenting to the emergency department.35,36 They help in the diagnosis and risk stratification of patients with chronic heart failure37 and are predictive of death and of functional status deterioration.33,38,39 They also appear to be useful in screening asymptomatic subjects at risk of developing heart failure, such as the elderly and those with hypertension, diabetes, or asymptomatic coronary artery disease.31 Moreover, natriuretic peptide levels have important prognostic value in patients with stable coronary artery disease and with acute coronary syndromes.40

Apart from the well-documented use of natriuretic peptides in the above indications, there are also interesting studies that showed their use in the evaluation of the risk of SCD, both in patients with heart failure and in those with coronary artery disease. Berger et al.,41 in 452 ambulatory patients with CHF and LVEF < 35%, found that serum BNP level was a strong independent predictor of SCD in patients with ischaemic or non-ischaemic severe heart failure. When a cut-off point was used only 1% of patients with BNP < 130 pg/ml died suddenly, compared with 19% of patients with BNP ≥ 130 pg/ml.

Two years later Tapanainen et al.42 showed that elevated serum BNP provided information about the risk of arrhythmic death in survivors of acute MI. Even after adjusting for clinical variables, including LVEF, BNP was the most specific predictor of SCD among other biomarkers in 521 acute MI patients. Most importantly, the accuracy of a low BNP plasma concentration as a predictor of survival with low risk of SCD was excellent, suggesting that this simple measurement could be used for screening purposes to rule out the risk of future SCD and the need for further risk stratification.

There is no clear causal relationship between malignant ventricular arrhythmogenesis and serum BNP levels. However, there are various hypotheses that could relate the pathophysiological mechanisms of heart failure with the increased production of natriuretic peptides and an increased risk of SCD. Both acute and chronic mechanical stretch of myocardial tissue, which are the stimuli for BNP production, cause immediate and chronic responses that lead to significant changes in cardiac electrophysiological properties via the activation of mechanosensitive or stretch-activated ion channels.4346 Ventricular stretch enhances refractoriness, slows conduction, and triggers ventricular ectopic beats and afterdepolarizations. This phenomenon, called mechanoelectric feedback, is thought to contribute to arrhythmogenesis and SCD in patients with impaired ventricular function and elevated filling pressures.47

The monitoring function of ICD devices may help to clarify in a more objective and reliable way whether natriuretic peptide levels are arrhythmia risk predictors in various groups of patients with ICDs.

We published the first study to show that, in primary ICD recipients according to the MADIT I criteria, the measurement of the NT-proBNP concentration in blood appears to be a sensitive and specific test for the identification of patients who are most likely to experience sustained ventricular tachyarrhythmias during a 1-year follow-up period.48

Verma et al.49 described in a larger prospective cohort study an association between higher BNP levels and occurrence of appropriate shocks or anti-tachycardia pacing in a mixed population of primary and secondary ICD recipients with ischaemic or non-ischaemic cardiomyopathies. They found that pre-implantation BNP above median was a significant predictor of ICD therapy in 345 consecutive patients undergoing first-time ICD implantation. Patients with BNP levels higher than median had a 2.2-fold higher risk for VT/VF events. Another recent study of 121 ICD recipients for secondary prevention tested several serum biomarkers and showed that an elevated BNP was a significant and independent predictor associated with a higher VT incidence.50 Even when pre-implantation serum assay of NT-proBNP is compared with invasive electrophysiological study, it seems to be superior for the prediction of future outcomes in predominantly secondary prophylactic ICD recipients after MI.51 When compared with LVEF, BNP plasma levels in ICD patients with chronic heart failure were better able to predict patients with event-free survival compared with patients reaching the combined endpoint of the study, which was a first tachyarrhythmic event or death.52

The study by Simon et al.53 (no ICD trial) identified elevated NT-proBNP levels as a predictor of arrhythmic events in patients with non-ischaemic cardiomyopathy and EF < 40%. In contrast, reproducible non-sustained VTs are less reliable risk factors for predicting the occurrence of sustained ventricular arrhythmias.

The most recent study validated the findings of other comparable studies and further reported that high BNP levels predict episodes of VT, VF, or sudden death among patients with left ventricular dysfunction who have received an ICD, and that low BNP levels help identify low risk patients who experienced a non-adverse outcome.54

Nevertheless, there is one negative study by Konstantino et al.,55 who showed that BNP was not associated with ventricular tachyarrhythmias in stable patients with moderate to severe left ventricular dysfunction and ICD. However, this was the only study with an intermediate-term follow-up of only 5 months.

Markers of extracellular matrix remodelling

Left ventricular remodelling, one of the main contributors to the pathophysiology of heart failure, is characterized by progressive structural and geometric changes that increase myocardial wall stress and result in left ventricle chamber dilation.56,57 The extracellular matrix (ECM) of the myocardium plays an important role in the remodelling process in the post-myocardial infarction heart, or during the development of non-ischaemic dilated cardiomyopathy. Normally, there is a balance between matrix metalloproteinases (MMPs: proteolytic enzymes responsible for ECM degradation) and tissue inhibitors of metalloproteinases (TIMPs).58 An imbalance, with dominance of MMPs over TIMPS, increases degradation of the myocardial fibrillar collagen network and leads to ventricular dilatation and remodelling.59 On the other hand, an increased deposition of collagen may also be deleterious to cardiac function because the resultant excessive fibrosis can impair ventricular function.32

Experimental and clinical data suggest that biochemical markers of collagen turnover correlate significantly with fibrosis in endomyocardial biopsies in hypertensive patients.60 Serological markers of these processes appear to be important targets of therapy and have been used to provide information about the prognosis in patients with symptomatic heart failure,61 idiopathic or ischaemic dilated cardiomyopathies,62 and after acute MI.63

Extracellular matrix turnover in CHF would be more related to a higher synthesis rate, as assessed by the concentrations of circulating pro-collagen peptides, than to altered collagen degradation, as observed in the RALES population.61 This excessive cardiac fibrosis may provide electrical heterogeneity, which is the substrate for arrhythmogenesis, thus potentially contributing to the occurrence of ventricular arrhythmia and subsequent SCD.

However, as serum markers of collagen turnover are not heart specific, and most of the studies published have a small patient sample size, we cannot draw certain conclusions about the prognostic role of these indexes for SCD or malignant ventricular arrhythmias. To our knowledge only one study by Blangy et al.50 examined this issue in secondary ICD recipients after MI and showed that the combination of decreased pro-collagen type III and increased pro-collagen type I was a significant marker of the risk of VT in the multivariate analysis.

Markers of inflammation

Apart from its clear evidence in atherothrombosis,6466 inflammation plays an important role in the development of CHF by adversely influencing heart contractility and inducing hypertrophy, apoptosis, and fibrosis.67 Biomarkers of inflammation, with main representative C-reactive protein (CRP), have been developed and have a low cost. Despite the fact that it lacks specificity, CRP is considered an independent predictor of adverse outcomes in patients with acute or chronic heart failure.67

Proinflammatory cytokines make up another more recent category of biomarkers. The cytokine hypothesis of heart failure proposes that a precipitating event, such as ischaemic cardiac injury, triggers innate stress responses, including the elaboration of proinflammatory cytokines, and that the expression of these cytokines is associated with deleterious effects on left ventricular function and accelerates the progression of heart failure.68,69

In an earlier study by Albert et al.70 of apparently healthy male physicians, baseline plasma CRP levels were positively associated with risk of SCD over the ensuing 17 years of follow-up. In some cases where the exact mechanism seems not to be acute ischaemia, there were microscopic isolated areas of myocardial fibrosis, which forms an ideal substrate for re-entry. In the MADIT-II study, statins were associated with a reduced risk of cardiac death or VT/VF and with reduced numbers of VT/VF episodes.71 These findings suggest that statins have anti-arrhythmic properties possibly attributable to their anti-inflammatory properties.

To date, only two studies have examined the prognostic role of CRP as a predictor of ventricular tachyarrhythmias in ICD patients. C-reactive protein and other inflammatory biomarkers (interleukin-6 and tumour necrosis factor α) were not predictive of intermediate-term risk of ventricular tachyarrhythmias in a small study of 50 stable CHF patients,55 whereas high-sensitivity CRP was correlated with VT episodes in 121 ICD recipients for secondary prevention of SCD after MI, over a 1-year period.50

The CAMI GUIDE, a prospective observational study, has been designed to evaluate the hypothesis that CRP levels could be included among the risk-predictive tools for the management of patients considered for ICD implantation, by measuring pre-implantation CRP levels in patients undergoing ICD implantation after MI according to MADIT II criteria and correlating these levels with the occurrence of SCD or aborted SCD.72


At this time the only certain fact is that biomarkers are predictors of more ICD interventions, but most of the information at our disposal comes from small studies, or studies limited by the indications for ICD and/or CRT. In order to define their value, in terms of shock reduction (whether or not the patients have heart failure), samples should be obtained from unselected populations, since a large pool of patients who suffer SCD are outside the present ICD/CRT indications. Naturally, many of the questions that arise, concerning the consequences of these observations, remain to be answered. Do these observations have clinical consequences at the present time? Should we program the devices more aggressively (shorter burst intervals, longer bursts and ramps, earlier delivered shocks, preventive pacing, biventricular pacing devices)? Should we treat the haemodynamic and myocardial instability with anti-arrhythmics, statins, or diuretics, based on the criterion of elevated biomarker levels?

Much remains to be elucidated and certainly some of the questions raised above will be answered by new trials.


The quest for factors that have a very good correlation with appropriate ICD discharges in patients with heart failure is logical, since we can make use of them in order to reduce or avoid arrhythmic events, thus improving the quality of life, morbidity, or even mortality of these patients. Simple clinical and laboratory indexes that reflect the haemodynamic status or the clinical profile of patients with heart failure have been identified, both by sub-analyses of large trials and by smaller prospective studies, as predictors of arrhythmic events in ICD patients. Biomarkers, which are playing a growing role in the risk stratification, prognosis, and treatment of patients with heart failure, could provide an auxiliary tool in this context, given that their measurement is now easily reproducible and widely available. Based on data available so far, it appears that natriuretic peptides are attracting the most interest, both from the pathophysiological point of view and in terms of the number of studies. However, it is still too early to tell, since the main characteristic of all biomarkers is their low specificity. This means that it is essential to carry out more and larger prospective studies that will determine the positive and negative predictive accuracy of biomarkers. Only then will we be in a position to judge whether biomarkers, in conjunction with other known risk factors, can play a useful role in the identification of high-risk patients, in the decision as to whether to implant an ICD and how to program it, and in defining the need or not for concomitant drug therapy.

Conflict of interest: none declared.


View Abstract