OUP user menu

Defining potential to benefit from implantable cardioverter defibrillator therapy: the role of biomarkers

Paul A. Scott , Paul A. Townsend , Leong L. Ng , Mehmood Zeb , Scott Harris , Paul J. Roderick , Nick P. Curzen , John M. Morgan
DOI: http://dx.doi.org/10.1093/europace/eur147 1419-1427 First published online: 22 July 2011

Abstract

Aims Implantable cardioverter defibrillator (ICD) therapy improves survival in patients at high sudden cardiac death (SCD) risk. However, some patient groups fulfilling indications for ICD therapy may not gain significant benefit: patients whose absolute risk of SCD is low and patients whose risk of death even with an ICD is high. The value of biomarkers in identifying patients’ potential for survival benefit from ICD therapy is unknown. We performed a pilot study to investigate this.

Methods and results Five established cardiovascular biomarkers were measured in patients with ICDs on the background of left ventricular dysfunction: N-terminal pro-brain natriuretic peptide [NT-proBNP], soluble ST2 [sST2], growth differentiation factor-15, C-reactive protein, and interleukin-6. The endpoints were all-cause mortality and survival with appropriate ICD therapy. One hundred and fifty-six patients were enrolled (age 69 years [Q1–Q3 62–77], 85% male, 76% ischaemic aetiology). During a follow-up of 15 ± 3 months, 12 patients died and 43 survived with appropriate ICD therapy. In a Cox proportional hazards model, the strongest predictors of death were Log sST2 (P< 0.001), serum creatinine (P< 0.001), and Log NT-proBNP (P= 0.002). The strongest predictor of survival with appropriate ICD therapy was Log NT-proBNP (P= 0.01).

Conclusion The biomarkers NT-proBNP and sST2 are promising biomarkers for identifying patients with little potential to gain significant survival benefit from ICD therapy. However, their incremental benefit, in addition to currently available clinical risk prediction models, remains unclear. These results demand a confirmatory prospective cohort study, designed and powered to derive and validate prediction algorithms incorporating these markers.

  • Implantable cardioverter defibrillators
  • Heart failure
  • Biomarkers
  • Mortality
  • Arrhythmias

Introduction

Implantable cardioverter-defibrillator (ICD) therapy improves survival in patients at high sudden cardiac death (SCD) risk.1 However, subpopulations of implanted patients do not derive benefit. Specifically, in patients enrolled in randomized controlled trials (RCTs) who never receive device therapy, it seems that over the time horizon of the study follow-up, their SCD risk was low.2 In contrast, other patients, who have either advanced heart failure or other co-morbidities, will have high mortality despite ICD therapy, as modification of their SCD risk does not offer significant survival benefit.3,4 Given its expense and associated morbidity, identification of patients’ potential to benefit is an important component of refining the application of ICD therapy.

Biomarkers can reflect systemic vascular inflammation, myocardial stretch, neurohormonal activation and myocyte injury, and as such are powerful predictors of mortality in patients with myocardial damage, independent of the assessment of left ventricular ejection fraction (LVEF).5 Furthermore, biomarkers reflecting inflammation, myocardial stretch and collagen turnover, may also predict appropriate anti-tachycardic therapy in patients with ICDs.68

However, the value of biomarkers in identifying patients most likely to gain benefit from ICD therapy is not known. We hypothesized that a range of biomarkers, chosen based on biological plausibility and reported associations with death or appropriate anti-tachycardic ICD therapy, may help to identify patients’ potential for survival benefit from ICD therapy. We tested this hypothesis in a prospective pilot study of ICD recipients.

Methods

Patient population

We recruited consecutive patients attending Southampton University Hospital ICD follow-up service, with left ventricular systolic dysfunction (LVSD) and an ICD or cardiac resynchronization defibrillator (CRT-D). All patients were on stable optimal medical therapy. None had heart failure admissions or therapy changes in the 6 weeks prior to enrolment. Other exclusion criteria were pregnancy, or an acute coronary syndrome or surgery of any type within the preceding 6 weeks.

At study entry, baseline demographic data and clinical characteristics were recorded, a 12-lead resting electrocardiogram performed, and New York Heart Association (NYHA) functional class assessed. All patients had a transthoracic echocardiogram prior to study entry. Blood was drawn from a forearm vein and collected in an ethylenediaminetetraacetic acid tube. Samples were centrifuged and plasma frozen within 1 h of sampling at −80°C pending analysis. The study complied with the Declaration of Helsinki and was approved by the local research ethics committee. Written informed consent was obtained from all patients.

Study endpoints and follow-up

Device programming was at the discretion of the treating physician. Following enrolment patients were followed up 3–6 monthly at a single centre, with a hospital visit or via a remote patient management system. Patients under remote follow-up also attended the hospital every 6 months. At each follow-up the device was interrogated. The occurrence of any ICD therapy was recorded. Appropriate ICD therapy was defined as:

  1. Antitachycardia pacing therapy (ATP) for ventricular tachycardia (VT).

  2. Shock therapy for VT or ventricular fibrillation (VF).

Correct arrhythmia detection/discrimination was confirmed by analysis of stored electrograms by two electrophysiologists blinded to the biomarker analysis.

Three study endpoints were chosen to enable exploration of the utility of biomarkers in defining patients’ potential to benefit from ICD therapy. These were:

  1. All-cause mortality.

  2. All-cause mortality or appropriate ICD therapy (reflecting event-free survival).

  3. Survival with appropriate ICD therapy.

Biomarker analysis

We analysed five plasma biomarkers that reflect a range of pathophysiological processes in LVSD:

  1. N-terminal pro-brain natriuretic peptide (NT-proBNP), an established marker of myocardial stretch.6

  2. Growth differentiation factor-15 (GDF-15), a marker of multiple stress pathways in the heart.9

  3. Serum ST2, the soluble form of ST2 (sST2), a novel marker of myocardial stretch.10

  4. C-reactive protein (CRP), a marker of systemic vascular inflammation.8,11

  5. Interleukin-6 (IL-6), a marker of systemic vascular inflammation.8

The biomarkers were chosen based on previous studies in patients with LVSD, that had demonstrated an independent association with mortality (all biomarkers),6,811 SCD (NT-proBNP and sST2),6,10 and the occurrence of spontaneous ventricular arrhythmias (NT-proBNP and IL-6).6,8

Commercially available antibodies (R and D systems, Abindgon, Oxfordshire, UK) were used for determination of GDF-15, IL-6, and sST2 as detailed below. N-terminal pro-brain natriuretic peptide and CRP were assayed using in-house antibodies. The assays for NT-proBNP and CRP have been demonstrated in a previous study to detect heart failure patients in a community screening programme.12 Moreover, the in-house NT-proBNP assay shows excellent correlation with the NT-proBNP Elecsys assay (Roche diagnostics) (n= 86, r= 0.90, P< 0.0001).

All assays were based on a two-site non-competitive assay format.12 Sheep antibodies were raised to the N-terminal of human NT-proBNP, and monoclonal mouse antibodies were raised to the C-terminal. Samples or NT-proBNP standards were incubated in C-terminal immunoglobulin-coated wells with the biotinylated N-terminal antibody for 24 h at 4°C. Detection was with methyl-acridinium ester-labelled streptavidin (MAE-streptavidin) on an MLX plate luminometer (Dynex Technologies Ltd, Worthing, UK), as described previously.9,12

The CRP monoclonal antibody used for capture of analyte was coated onto ELISA plates (100 ng per well).12 Plates were blocked using 10% foetal calf serum (FCS). Plasma (1 μL per well) or standards were incubated in coated wells for 24 h at 4°C. Following washes, 50 ng of a different biotinylated monoclonal CRP antibody was pipette into wells and incubated for 2 h at room temperature. Plates were developed with MAE-streptavidin as above.

For the GDF-15, sST2, and IL-6 assays, specific mouse monoclonal antibodies for these peptides were coated onto ELISA plates (200 ng/100 μL).9 After incubation for 24 h, all plates were washed and blocked using 10% FCS. Plasma samples were pipetted into the wells (10, 20, and 100 μL per well for the GDF-15, sST2, and IL-6, respectively), together with appropriate standards. After another 24 h of incubation, plates were washed and biotinylated goat antibodies pipetted into the wells (5, 10, or 20 ng/100 μL for GDF-15, sST2, and IL-6, respectively). After another period of incubation of 2 h, plates were washed and developed with MAE-streptavidin as above.

Statistics

Categorical variables were expressed as percentages (numbers). Normally distributed continuous variables are expressed as mean ± standard deviation and compared using the independent-samples t-test. Variables not normally distributed are expressed as median (lower quartile to upper quartile) and compared using the Mann–Whitney U test.

As NT-proBNP, sST2, GDF-15, Hs-CRP, and IL-6 were not normally distributed, log-transformed values were used for analysis. Univariate predictors of the three endpoints were investigated in Cox proportional hazards models. For the endpoints of all-cause mortality or appropriate ICD therapy, and survival with appropriate ICD therapy, multivariable analysis was also performed. Variables demonstrating a significant association (P< 0.10) with all-cause mortality or appropriate ICD therapy (age, history of AF, Log NT-proBNP, Log GDF-15, Log IL-6) and survival with appropriate ICD therapy (age, Log NT-proBNP) in univariate analyses were included in the multivariable Cox proportional hazards models. The multivariable models were built using a backward stepwise approach with a significance of P< 0.05 to remain in the model. For the endpoint of all-cause mortality, in view of the small number of patients reaching the endpoint (n= 12) multivariable analysis was not performed. The proportional hazards assumption was checked using Schoenfeld residuals.13

For the endpoints of all-cause mortality, and all-cause mortality or appropriate ICD therapy, biomarker cut-off points were chosen to identify patients with a high risk of death (NT-proBNP and sST2) and a high chance of event-free survival (NT-proBNP). Kaplan–Meier survival analysis was used for comparison between patient groups stratified according to these biomarker cut-off points, and survival curves were compared using the log-rank test.

Statistical analyses were performed on SPSS Version 17 (SPSS Inc., Chicago, IL, USA). In all analyses P < 0.05 was considered significant, except for inclusion in the multivariable models.

Results

Patient characteristics and clinical outcomes

Two hundred consecutive patients were screened for study enrolment. Five patients were excluded due to recent admissions with heart failure, 23 patients were excluded as they did not fit the study inclusion criteria (9 with arrhythmogenic right ventricular cardiomyopathy, 8 with hypertrophic cardiomyopathy, 3 with Brugada syndrome, 1 with long-QT syndrome, and 2 with primary VF), and 16 patients declined to take part in the study. This left 156 patients who were enrolled in the study, at a mean of 48 ± 45 months following initial ICD implant. Demographics at study entry are shown in Table 1. The ICD VT treatment zone lower setting was similar in patients who did, and did not, receive appropriate ICD therapy (152 ± 11 vs. 154 ± 24 beats per minute (b.p.m.), respectively; P= 0.55).

View this table:
Table 1

Patient characteristics at study entry

Overall (n= 156)
Age (years)71 (62–77)
Male sex % (no.)85 (132)
Heart disease type % (no.)
 Ischaemic76 (119)
 NICM18 (28)
 Other6 (9)
Diabetes % (no.)24 (37)
History of AF % (no.)36 (56)
NYHA class % (no.)
 I35 (55)
 II42 (65)
 III22 (34)
 IV1 (2)
Device type % (no.)
 ICD66 (103)
 CRT-D34 (53)
ICD indication % (no.)
 Primary prevention37 (58)
 Secondary prevention63 (98)
QRS width (ms)125 (100–160)
LVEF
 <30%63 (98)
 30–35%12 (19)
 35–40%14 (22)
 >40%11 (17)
Beta-blocker % (no.)79 (124)
ACE-I/ARB % (no.)92 (143)
Amiodarone % (no)28 (44)
Creatinine (μmol/L)114 (92–141)
Haemoglobin (g/dL)134 ± 18
  • NICM, non-ischaemic cardiomyopathy; AF, atrial fibrillation; NYHA, New York Heart Association functional class; LVEF, left ventricular ejection fraction; CRT-D, cardiac resynchronisation therapy defibrillator; ACE-I, angiotensin-converting enzyme inhibitor; ARBs, angiotensin II receptor blockers.

During a mean follow-up of 15 ± 3 months from study entry 12 (8%) patients died and 47 (30%) experienced appropriate ICD therapy. Four patients who experienced appropriate ICD therapy subsequently died, leaving 43 (28%) patients who experienced appropriate ICD therapy and survived the duration of the study. The distribution of appropriate ICD therapies in these 43 patients was:

  • survival with appropriate therapy for VT (rate < 182 b.p.m.) only—17 patients;

  • survival with appropriate therapy for fast VT (rate ≥ 182 b.p.m.)14—25 patients;

  • survival with appropriate therapy for VF—1 patient

Of these 43 patients, 10 experienced 1 treated VT/VF episode, while the median number of episodes per patient was 3. Twenty-one patients experienced appropriate shock therapy, while the remainder experienced only ATP.

Serum biomarkers

Baseline serum biomarker levels are shown in Table 2. Patients who died had significantly higher levels of NT-proBNP (P< 0.001), sST2 (P< 0.001), GDF-15 (P= 0.04), and IL-6 (P= 0.04), than patients who survived without appropriate ICD therapy (Figure 1). Patients who survived with appropriate ICD therapy had a significantly higher level of NT-proBNP than patients who survived without ICD therapy (P= 0.01).

View this table:
Table 2

Baseline biomarker levels in relation to outcome

BiomarkerAll-cause mortality (n= 12)Event-free survival (n= 101)Survival with appropriate ICD therapy (n= 43)P value death vs. event-free survivalP value survival with appropriate ICD therapy vs. event-free survival
NT-proBNP (pmol/L)2025 (701–2818)412 (135–1173)832 (399–1167)<0.0010.01
GDF-15 (μg/L)4.71 (1.25–9.00)1.72 (0.84–3.44)1.90 (0.87–4.56)0.040.48
sST2 (ng/mL)0.48 (0.30–0.54)0.29 (0.22–0.37)0.28 (0.21–0.36)<0.0010.97
CRP (mg/L)3.41 (1.57–3.81)1.58 (1.16–2.84)1.44 (1.08–2.62)0.090.35
IL-6 (ng/L)10.28 (0.58–100.67)0.46 (0.14–9.07)1.35 (0.14–16.90)0.040.32
Figure 1

Box plots showing the baseline concentrations of N-terminal pro-brain natriuretic peptide (A) and soluble ST2 (B) in patients who died and patients with event-free survival. Biomarker levels are presented as box (25th percentile, median, 75th percentile) and whisker (10th and 90th percentiles) plots.

Predictors of all-cause mortality

In univariate analyses, four of the five biomarkers were significant predictors of all-cause mortality: Log sST2 [hazard ratio (HR) 265; 95% confidence intervals (CI) 16.47–4268; P< 0.001), Log NT-proBNP (HR 25.30; 95% CI 3.16–202; P= 0.002), Log IL-6 (HR 1.67; 95% CI 1.03–2.74; P= 0.04), and Log GDF-15 (HR 3.17; 95% CI 1.03–9.76; P= 0.04) (Table 3). Additional significant clinical and biochemical predictors were serum creatinine (HR per 10 μmol/l 1.12; 95% CI 1.06–1.19; P< 0.001), haemoglobin (HR per g/dl 0.95; 95% CI 0.92–0.99; P= 0.01), and NYHA class (HR Class III/IV vs. I 10.91; 95% CI 1.34–88.9; P= 0.03).

View this table:
Table 3

Predictors of all-cause mortality, all-cause mortality or appropriate Implantable cardioverter defibrillator therapy (event-free survival), and survival with appropriate Implantable cardioverter defibrillator therapy

All-cause mortalityAll-cause mortality or appropriate ICD therapySurvival with appropriate ICD therapy
Univariate analysisUnivariate analysisMultivariable analysisUnivariate analysisMultivariable analysis
P valueHR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P valueHR (95% CI)
Age (per 10 years)0.251.44 (0.77–2.69)0.021.42 (1.06–1.89)0.340.041.39 (1.01–1.93)0.29
Diabetes0.142.36 (0.75–7.46)0.751.10 (0.60–2.02)0.670.86 (0.41–1.78)
History of AF0.092.74 (0.87–8.64)0.081.62 (0.95–2.76)0.520.281.40 (0.76–2.57)
NYHA class (vs. I)
 II0.253.67 (0.41–32.83)0.341.35 (0.73–2.52)0.591.20 (0.62–2.31)
 III/IV0.0310.91 (1.34–88.9)0.281.48 (0.73–3.00)0.590.78 (0.32–1.90)
QRS (per 10 ms increase)0.161.12 (0.96–1.30)0.571.02 (0.95–1.10)0.931.0 (0.91–1.01)
LVEF (per 5% decrease)0.251.28 (0.85–1.93)0.181.11 (0.95–1.30)0.391.08 (0.91–1.27)
Beta-blocker0.771.26 (0.28–5.73)0.321.44 (0.70–2.93)0.371.44 (0.64–3.24)
Amiodarone0.261.94 (0.62–6.130.700.89 (0.48–1.63)0.320.69 (0.33–1.44)
Cr (per 10 μmol/L increase)<0.0011.12 (1.06–1.19)0.171.03 (0.99–1.08)0.630.98 (0.92–1.05)
Hb (per g/dL increase)0.010.95 (0.92–0.99)1.01.00 (0.98–1.02)0.221.01 (0.99–1.03)
Log NT-proBNP0.00225.30 (3.16–202)<0.0013.16 (1.73–5.78)<0.0013.16 (1.73–5.78)0.012.26 (1.21–4.21)0.012.26 (1.21–4.21)
Log GDF-150.043.17 (1.03–9.76)0.061.67 (0.98–2.85)0.820.271.41 (0.76–2.61)
Log sST2<0.001265 (16.47–4268)0.113.35 (0.76–14.83)0.810.82 (0.16–4.25)
Log CRP0.134.82 (0.62–37.20)0.540.80 (0.40–1.62)0.120.55 (0.26–1.16)
Log IL-60.041.67 (1.03–2.74)0.081.22 (0.97–1.54)0.220.381.13 (0.87–1.46)
  • HR, Hazard Ratio; AF, atrial fibrillation; NYHA, New York Heart Association Functional Class; Cr, creatinine; Hb, haemoglobin.

In the subgroup of patients with a primary prevention ICD (n= 58) both Log sST2 (HR 97.20; 95% CI 4.43–2131; P= 0.004) and Log NT-proBNP (HR 30.31; 95% CI 2.27–404; P= 0.01) remained predictive. In the subgroup of patients with a secondary prevention ICD (n= 98) Log sST2 (HR 1091; 95% CI 2.49–479063; P= 0.02) predicted all-cause mortality but Log NT-proBNP (HR 16.17; 95% CI 0.70–373; P= 0.08) did not.

Predictors of event-free survival

Using the combined endpoint of all-cause mortality or appropriate ICD therapy, we evaluated predictors of event-free survival (Table 3). Significant univariate predictors were Log NT-proBNP (HR 3.16; 95% CI 1.73–5.78; P< 0.001) and advancing age (HR per 10 years 1.42; 95% CI 1.06–1.89; P= 0.02). However, in multivariable analysis only Log NT-proBNP remained significantly predictive (HR 3.16; 95% CI 1.73–5.78; P< 0.001).

Log NT-proBNP remained significantly predictive in the subgroups of both primary prevention (HR 3.53; 95% CI 1.36–9.18; P= 0.01) and secondary prevention patients (HR 2.92; 95% CI 1.33–6.39; P= 0.007).

Predictors of survival with appropriate implantable cardioverter-defibrillator therapy

Significant univariate predictors of survival with appropriate ICD therapy were Log NT-proBNP (HR 2.26; 95% CI 1.21–4.21; P= 0.01) and advancing age (HR per 10 years 1.39; 95% CI 1.01–1.93; P= 0.04) (Table 3). However, in multivariable analysis only Log NT-proBNP remained significantly predictive (HR 2.26; 95% CI 1.21–4.21; P= 0.01).

N-terminal pro-brain natriuretic peptide remained significantly predictive in the subgroup of secondary prevention (HR 2.43; 95%CI 1.09–5.42; P= 0.03) but not primary prevention patients (HR 1.93; 95% CI 0.72–5.20; P= 0.19).

Combining biomarkers to identify patients unlikely to benefit from implantable cardioverter defibrillator therapy

Using best cut-off values of sST2 and NT-proBNP (see Methods section), we developed models that divided patients into three groups: (i) patients with a low risk of death or appropriate ICD therapy; (ii) patients with a high risk of appropriate ICD therapy but low risk of death; and (iii) patients with a high risk of death (Table 4).

View this table:
Table 4

Combining biomarkers to identify groups of patients unlikely to gain significant benefit from Implantable cardioverter-defibrillator therapy

Group 1Group 2Group 3
Low risk of ICD therapy and deathHigh risk of ICD therapy and low risk of deathHigh risk of death
Model 1NT-proBNP < 173 pmol/LNT-proBNP ≥ 173 pmol/L and <2350 pmol/LNT-proBNP ≥ 2350 pmol/L
 Total Patients in Group3111411
 Death075
 Survival with any appropriate ICD therapy0412
Model 2NT-proBNP < 173 pmol/LNT-proBNP ≥ 173 pmol/L and sST2 < 0.43 ng/mLsST2 ≥ 0.43 ng/mL
 Total Patients in Group3110718
 Death057
 Survival with any appropriate ICD therapy0385

N-terminal pro-brain natriuretic peptide was used to identify a group of patients at low risk of death or appropriate ICD therapy. None of the 31 patients with NT-proBNP below 173 pmol/L had an event, whereas 55 of the 125 above this level did (P< 0.001) (Figure 2A). Of these 31 patients, 13 (42%) had originally received a prophylactic device, 19 (61%) had CAD, 17 (55%) had an LVEF<35%, and 17 (55%) were in NYHA Class II–IV.

Figure 2

Kaplan–Meier survival curve analysis. (A) All-cause mortality or appropriate implantable cardioverter defibrillator therapy (reflecting event-free survival) in groups stratified by N-terminal pro-brain natriuretic peptide level (cut-off 173 pmol/L); (B) death in groups stratified N-terminal pro-brain natriuretic peptide level (cut-off 2350 pmol/L); and (C) death in groups stratified by soluble ST2 (cut-off 0.43 ng/mL).

The ability of NT-proBNP and sST2 to identify a group of patients at high risk of death was evaluated separately. For NT-proBNP, 5 of the 12 patients with a level >2350 pmol/L died, whereas only 7 of the 145 below this level died (P< 0.001) (Figure 2B). For sST2, 7 of the 18 patients with a level >0.43 ng/mL died, whereas only 5 of the 138 below this level died (P< 0.001) (Figure 2C).

Discussion

The main findings of this study are that the biomarkers sST2 and NT-proBNP are promising candidates for identifying patients with low potential to benefit from ICD therapy. These biomarkers identified a group of patients with advanced heart failure, whose short-term risk of death despite ICD therapy was high. N-terminal pro-brain natriuretic peptide also identified a group of apparently low-risk patients (∼20% of the cohort) who experienced no episodes of ICD therapy over the study's time horizon. However, the study population was patients with pre-existing rather than new ICD implants, and the follow-up was too short for this to be a clear endpoint.

Although RCTs have demonstrated mortality benefit with ICD therapy, translation into clinical practice has been challenging.1,15 Most patients implanted with an ICD based on current guidelines never receive a lifesaving device therapy.2 Furthermore, many patients with advanced heart failure will not have their life meaningfully prolonged by ICD therapy.15 This is either because they will die of another cause without receiving device therapy, or die soon after receiving appropriate ICD therapy from a non-arrhythmic (usually pump failure) death. In these patients the ICD serves only to alter the mode but not time of death. Set against these observations is the fact that ICD therapy is associated with significant morbidity, and is a high-cost therapy with questionable cost-effectiveness.

In order to enrich ICD therapy clinical and cost-effectiveness, effort has been made to identify patients at high SCD risk. To date, there is no consensus on how to do this. Importantly, patients at highest SCD risk do not necessarily have the greatest potential to benefit from ICD therapy, as they are also at highest non-sudden death risk.16 An alternative approach is to identify patients meeting current guidelines, who do not have the potential to gain significant survival benefit from ICD therapy.

A number of studies have evaluated the use of individual clinical characteristics to achieve this. Both serum creatinine and advancing age identify ICD recipients with a high short-term risk of death following ICD implantation.3,17 In our study, creatinine was a significant predictor of death (P< 0.001) and age predicted event-free survival (P= 0.02). Both variables may prove useful combined in a model with biomarkers. However, in view of the small number of patients reaching the endpoint of all-cause mortality in our study (n= 12) it was not possible to evaluate whether combining clinical risk markers and biomarkers could refine the identification of patients with very high short-term mortality despite ICD therapy.

Two studies have evaluated more complex risk scores. Goldenberg et al.4 examined the relationship between a risk score derived from 5 clinical variables, and benefit from ICD therapy, in the 1232 patients enrolled in the MADIT-II trial.4 The 345 patients with no risk markers, whose 2-year mortality was 9%, derived no benefit from an ICD (HR 0.96; P > 0.91).The investigators also described a group of 60 patients with advanced renal dysfunction (BUN > 50 mg/dL and/or serum creatinine >2.5 mg/dL), whose 2-year mortality was 50%, who also did not benefit from their device (HR 1.0, P> 0.99). In our study, although creatinine was a predictor of death, only 2 of the 12 patients who died had a creatinine that would have merited inclusion in this group. Furthermore, of the seven patients with an sST2 above the ROC-derived cut-off (0.43 ng/mL) who died, only one would have been in the group, and none of the five patients who died with an NT-proBNP level above the ROC-derived cut-off (2350 pmol/L) would have been in the group. Levy et al.18 used a modified version of the Seattle Heart Failure Model to examine baseline predicted mortality risk, and the relative and absolute benefit from ICD therapy, in 2487 patients in the SCD-HeFT trial. Patients with the highest quintile of baseline predicted risk of death gained no significant benefit from device therapy. However, neither study has yet had an impact on international guidelines.1

Using their data Levy et al.18 suggested that the benefit of ICD therapy approached null when annual mortality reached 20–25%. Unfortunately, using our data it was not possible to assess the annual mortality rate at which ICD therapy would cease to be clinically effective. However, using the data from Levy et al. both biomarker-defined high-risk groups in our study, where the annual mortality was 36.4% in patients with NT-proBNP ≥ 2350 pmol/L and 33.3% in patients with sST2 ≥ 0.43 ng/mL, may not benefit from ICD therapy.

A number of studies have evaluated the use of biomarkers to predict SCD and appropriate ICD therapy. The most studied of these is BNP/NT-proBNP, which predicts both SCD and appropriate ICD therapy.6 These are consistent with our results. The association of inflammatory markers with ICD therapy has been inconsistent. Although smaller studies have shown an association, other larger-prospective studies have not confirmed these findings.19 In our study neither CRP nor IL-6 predicted appropriate ICD therapy.

A recent nested case-control study by Pascual-Figal et al.10 investigated the association of sST2 and SCD, in patients with heart failure in a multicentre registry. Soluble ST2 levels were significantly higher in 36 patients who died suddenly than 63 matched controls (0.23 vs. 0.12 ng mL, P= 0.001). In contrast, we found that sST2 did not predict survival with appropriate ICD therapy, but did predict overall mortality.

A number of studies have observed that patients who receive appropriate ICD shock therapy have a high risk of subsequent death, and the possibility that shock therapy may be detrimental has been raised. In our cohort four patients experienced appropriate ICD therapy, of which three were shock therapy, and subsequently died. In comparison with the survivors, the baseline sST2 level was significantly higher (0.30 ± 0.12 vs. 0.42 ± 0.09 ng/mL, P= 0.04) and the NT-proBNP non-significantly higher (808 ± 740 vs. 2009 ± 1820 pg/mL, P= 0.28) in these four patients. Furthermore, the mean biomarker levels were similar in patients who died with (n= 4) and without (n= 8) (sST2 0.53 ± 0.30 ng/mL, NT-proBNP 1949 ± 1056 pg/mL) experiencing appropriate device therapy prior to death. These findings are consistent with appropriate shock therapy being a marker of advanced heart failure and high mortality risk in some patients, rather than a causal factor, although no conclusions can be drawn in view of the small numbers and post-hoc nature of the analysis.

Our observations that biomarkers can predict mortality in patients with advanced LVSD are consistent with those of other studies.5 The suggestion that these biomarkers may be able to accurately characterize heart failure severity, and identify patients whose disease is too advanced to gain significant benefit from ICD therapy, is novel.

The use of sST2 and NT-proBNP in a model for refining patient selection for ICDs could have potential benefits in comparison with other tools. Their measurement is cheap, reproducible, straightforward, and can be utilized as a continuous variable or dichotomized using a cut-off point, depending on the model of application. Furthermore, in patients with acquired heart disease, risk is a dynamic rather than static quantity, and the non-invasive nature of biomarker assessment lends itself to repeated measurement over time. However, it is unclear as to how the biomarkers compare with the risk scores described in the studies by Goldenberg et al. and Levy et al., or whether they would add incremental accuracy to these proposed models or to the assessment of renal function alone.3,4,17,18

Study limitations

The study sample is small and the number of deaths low. Our study is essentially hypothesis generating and our findings need repeating in a larger cohort with validation of the optimum biomarker cut-off values.

We included patients with both primary and secondary prevention indications and therefore the study population may not fully reflect the patient population (primary prevention patients) in whom risk stratification tests are most needed. This is most relevant with respect to the low-risk patient group identified by a low NT-proBNP, who experienced no episodes of appropriate ICD therapy during follow-up. However, consistent with international guidelines the identification of patients with advanced heart failure or other comorbidity, whose short-term risk of death despite ICD therapy is high, is relevant to both primary- and secondary-prevention patients, and for this reason secondary-prevention patients were included in the study.1,20

All patients enrolled in our study had their devices already implanted. This is an important limitation that may reduce the generalizability of our results. However, at study entry 75% (n= 117) of patients had an LVEF ≤ 35% and may have had an indication for prophylactic ICD therapy based on current guidelines without the need for further testing.1 Furthermore, the death rate in our study (8%) was similar to the 1-year mortality rate in the ICD arm of MADIT-II (9%), and our rate of appropriate ICD therapy (30%) was slightly higher than in contemporary device trials, likely in part reflecting the high proportion (63%) of secondary prevention patients.2,21 This suggests that, despite being a cohort of prevalent rather than new ICD patients, our study cohort may not differ significantly in terms of arrhythmic and non-arrhythmic risk, from a contemporary population of new ICD implants. In addition, none of the patients who died were in NYHA Class IV heart failure on study entry, and therefore they would not have been excluded from ICD therapy by current guidelines.1

Study follow-up was short. However, consistent with current guidelines, the aim of our study was to identify patients whose chance of survival beyond 12 months was limited.1 Our finding that NT-proBNP may identify patients with a low risk of appropriate ICD therapy, needs repeating in a study with longer follow-up and patients enrolled at initial implant.

Delivery of appropriate ICD therapy is not always a surrogate for preventable SCD. However, with current guidelines widening the recipient population for ICDs, the investigation of predictors of SCD in higher-risk patients is difficult, as most such patients are indicated for an ICD.1

Conclusion

This pilot study suggests that NT-proBNP and sST2 are promising biomarkers for identifying patients with little potential to gain survival benefit from ICD therapy. They may provide a simple strategy for refining patient selection for ICD therapy. The incremental benefit of these biomarkers in addition to currently available clinical risk prediction models remains unclear. Our study indicates the need for a prospective cohort study, from time of ICD implantation, powered to derive and validate algorithms to predict potential to benefit from ICD therapy.

Conflict of interest: P.A.S. and M.Z. are supported by educational grants from Medtronic. P.A.T. has received Honoraria and research grants from Karus Therapeutics, Abcam and Biocompatibles. N.P.C. has received honoraria and research grants from Boston Scientific, Medtronic, Cordis, Abbott Vascular, St Jude Medical, Sorin, GSK, Haemonetics, AstraZeneca, Lilly, Schering-Plough, The Medicines company, and Pfizer. J.M.M. has received Honoraria and research grants from Medtronic, St Jude, Sorin and Boston Scientific. L.L.N., S.H., and P.J.R. report no conflicts of interest.

Funding

This work was supported by an unrestricted educational grant from Medtronic.

References

View Abstract