If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Implantable cardioverter-defibrillator (ICD) shocks are associated with an increased risk of death. It is unclear whether ICD shocks are detrimental per se or a marker of higher risk patients.
We aimed to assess the association between ICD shocks and time to death after correction for baseline mortality based on the Seattle Heart Failure Model (SHFM).
The primary analysis compared time-to-death between patients receiving no shocks and patients receiving shocks of any type adjusted for SHFM score at time of implantation and other comorbidities. Subgroup analyses were performed to further describe the relationship between shocks and mortality risk.
Over a median follow-up of 41 months (interquartile range 23–64), one or more shock episodes occurred in 59% of 425 patients and 40% of the patients died. Patients receiving shocks of any type had increased risk of death (hazard ratio 1.55; 95% confidence interval 1.07–2.23; P = .02) versus patients receiving no shocks. While patients with 1–5 days with shock (shock days) did not show evidence of increased risk of death (1.30 [0.88–1.94]; P = 0.19), those with 6–10 shock days (2.22 [1.21–4.08]; P <.01) and >10 shock days (3.66 [1.86–7.19]; P <.01) had increasingly higher risk. There was no increased hazard for death (0.73 [0.34–1.57]; P = .41) in patients treated only with antitachycardia pacing (ATP).
ICD shocks were associated with increased mortality risk after adjustment for SHFM-predicted mortality, and the burden of shocks played a role in this association. ATP did not increase mortality risk, suggesting that shocks may themselves be detrimental.
The Antiarrythmics versus Implantable Defibrillator (AVID) Investigators A comparison of antiarrhythmic-drug therapy with implantable defibrillators in patients resuscitated from near-fatal ventricular arrhythmias.
It is unclear whether ICD shocks are detrimental per se or a marker of a sicker patient population with a higher baseline risk of death. To answer this question, we looked at the association among shocks, shock burden, and death, correcting for baseline mortality based on the Seattle Heart Failure Model (SHFM).
Our study population included patients with device implantation selected from a prospective database of patients implanted at the Portland VA Medical Center. The analysis included all 425 patients who received ICDs between January 1994 and January 2008 and who subsequently received their follow-up at the same hospital. The type of ICD implanted and programming parameters were determined at the discretion of the implanting electrophysiologist according to standard clinical practice. During this period, essentially all patients had ATP programmed on, regardless of the indication for implant.
ICD implant data, programming parameters, follow-up dates and results, ICD therapy events, and date of death were entered prospectively into an independent database maintained by the Portland VA Medical Center Electrophysiology Department. ICD events were classified by the attending electrophysiologist based on all available clinical and ICD data (including electrograms) as being for ventricular tachycardia (VT), ventricular fibrillation (VF), atrial fibrillation (AF), sinus tachycardia or other supraventricular tachycardia, or oversensing and entered prospectively into the independent database. However, no distinction was made between appropriate versus inappropriate shocks in the final analysis. To obtain baseline clinical characteristics not recorded in the database, chart reviews were conducted by two investigators (JE and GL) according to a standard protocol.
Calculation of the SHFM score
The SHFM score was calculated from baseline data using the equation described by Levy et al.
To calculate this score, we required that a participant have no missing variables other than lymphocyte percent or uric acid level. Forty-seven (11%) of 425 patients were excluded for this reason. In the case of missing data for lymphocyte percent and/or uric acid levels, the median values for the analytic data set were used.
The primary analysis compared time to death for all patients receiving shocks with that for patients who did not receive shocks (Figure 1). Patients who received only episodes of ATP were included in the no-shock group. The effect of ATP was examined by stratifying the no-shock group into patients receiving ATP only and patients receiving no therapy. In the time-to-event analysis, follow-up times for patients with ATP only before the time of their first shock were included in the ATP-only group. After the first shock episode, they became part of the shock group. Shock burden was examined by looking at both cumulative days with shocks (shock days) and cumulative number of shocks stratified into 1–5 shock days/total shocks, 6–10 shock days/total shocks, and >10 shock days/total shocks. To further delineate the effects of timing of shocks, we eliminated patients with five or more shocks in a 24-hour period (“shock storm”) and stratified this group into those with <5 shock days/total shocks and ≥5 shock days/total shocks. We then compared patients with ≥5 shock days/total shocks stratified into shock storm and no-shock storm groups. For all analyses of shock burden, the reference group was the no-shock group, including both ATP only and no therapy, except in the shock storm versus no-shock storm comparison.
Descriptive statistics on all statistical endpoints and baseline characteristics were performed for the complete study population (n = 425) as well as for the shock and no-shock groups. Baseline characteristics between the study groups were compared using Wilcoxon rank-sum tests for continuous variables, χ2-tests for dichotomous variables, or Fisher's exact test as appropriate for dichotomous variables. Follow-up time was calculated as the interval from time of implant to time of death or last follow-up.
Kaplan-Meier survival curves were constructed for the estimation of unadjusted survival distributions between the shock and no-shock group as well as for subgroup analyses. Log-rank tests were used for the comparison of overall survival between groups. Cox proportional-hazards models were used to examine the relationship between ICD therapies and time to death. Univariate Cox proportional-hazards models were run on baseline characteristics including SHFM score and ejection fraction (EF) as well as presence/absence of chronic kidney disease (CKD), congestive heart failure (CHF), coronary artery disease (CAD), hypertension (HTN), AF, diabetes mellitus (DM), QRS duration (>120 ms), and left bundle branch block (LBBB). Multivariate Cox models were run that included covariates with P-values of ≤.20 in univariate analyses, and stepwise selection was used to determine the most parsimonious model. Shock was modeled as a time-dependent covariate in all analyses with the risk changing after the occurrence of first shock episode as well as subsequent shock episodes in the shock burden analyses. ATP was modeled as a time-dependent covariate in the subgroup analysis comparing shocks, no therapy, and ATP only. All tests were conducted at the two-sided .05 significance level. Analysis was performed using SAS version 9.2 (Cary, NC).
Deaths and ICD event analysis
Of the 425 patients included in the study, 252 (59%) received a shock of any type during the study period. There were 190 (45%) patients with 1–5 shock days, 44 (10%) with 6–10 shock days, and 18 (4%) with >10 shock days, while 121 (28%) received one to five shocks, 51 (12%) received six to 10 shocks, and 80 (19%) received >10 shocks. Ninety-six patients (23%) experienced one or more episodes of shock storm. In patients without episodes of shock storm, 54 (17%) had five or more shocks and 34 (8%) had ≥5 shock days. In the no-shock group, 130 (31%) patients received no therapy and 43 (10%) received ATP only. An additional 77 (18%) patients had ATP episodes before subsequently having shock episodes and in the time-dependent analysis contributed data to the ATP-only group until they had a shock. During the median follow-up period of 41 months (interquartile range [IQR] 23–64), 171 (40%) patients died, with 102 of those receiving shocks. The median time to first shock episode was 8.1 months (IQR 2.2–21.7), and the median number of shocks received was 6 (IQR 2–14).
Baseline clinical characteristics and programming data
Baseline clinical characteristics are shown in Table 1. Overall, the two groups were similar, with the exception of a greater percentage furosemide use, primary prevention, and occurrence of diabetes in the no-shock group. Eighty patients (19%) were excluded from the multivariate analysis owing to missing data on clinical characteristics. Four hundred seven of 425 patients had VT detection and ATP programmed on. The mean VT detection heart rate was 164 ± 11 bpm in those 407 patients. The mean VF detection rate was 210 ± 13 bpm.
Table 1Selected baseline clinical characteristics for primary study groups
Total population (n = 425)
Shock (n = 252)
No shock (n = 173)
Male, % (n)
Current smokers, % (n)
Uric acid, mg/dL
Total cholesterol, mg/dL
QRS duration, ms
QRS >120 ms, % (n)
LBBB, % (n)
Systolic blood pressure, mm Hg
Primary prevention, % (n)
ACE inhibitor, % (n)
Beta-blocker, % (n)
Angiotensin receptor blocker, % (n)
Statin, % (n)
Allopurinol, % (n)
Aldosterone antagonist, % (n)
Diuretics, % (n):
New York Heart Association class:
Congestive heart failure, % (n)
Coronary artery disease, % (n)
Atrial fibrillation, % (n)
HTN, % (n)
DM, % (n)
CKD, % (n)
SHFM 1-year survival, %
SHFM 5-year survival, %
Note: Continuous variables are shown as median (25th, 75th percentiles).
Risk of mortality from all causes was increased in the shock versus the no-shock group, although the difference approached but did not reach statistical significance (hazard ratio [HR] 1.32; 95% confidence interval [CI] 0.97–1.81; P = .08). In unadjusted (univariate) Cox proportional-hazards analysis, several variables were significantly associated with death, including SHFM score (HR 1.99; 95% CI 1.63–2.42; P <.01), CKD (HR 2.04; 95% CI 1.44–2.89; P <.01), and QRS >120 ms (HR 1.48; 95% CI 1.04–2.10; P = .03); however, the SHFM score and CKD were the only covariates that predicted mortality in a multivariate analysis. After adjustment, shocks were significantly associated with an increased risk of death (HR 1.55; 95% CI 1.07–2.23; P = .02). Results for all multivariate Cox hazards models can be found in Table 2. The adjusted survival curves for the shock versus no-shock groups are shown in Figure 2.
Table 2Adjusted hazard ratios for primary and subgroup analyses
Shock vs. no shocks
1–5 shock days vs. no shocks
6–10 shock days vs. no shocks
>10 shock days vs. no shocks
1–5 shocks vs. no shocks
6–10 shocks vs. no shocks
>10 shocks vs. no shocks
Shock vs. no therapy
ATP only vs. no therapy
No shock storm:
<5 shock days vs. no shocks
≥5 shock days vs. no shocks
<5 shocks vs. no shocks
≥5 shocks vs. no shocks
≥5 Shock days
Shock storm vs. no shock storm
Shock storm vs. no shock storm
Note: The no-shock group includes ATP only and no therapy. Shock days represents the cumulative number of days a patient experienced one or more shock. Shock storm is defined at five or more shocks in a 24-hour period. Covariates in the multivariate analyses include the SHFM score and CKD. Other covariates tested included CHF, CAD, QRS duration >120 ms, LBBB, EF, smoking status, DM, HTN, and AF.
Given the decreased survival in the shock group, we conducted further analyses to look at how shocks may be contributing to this finding. To examine whether ATP was associated with increased mortality risk, we stratified the no-shock group into ATP only and no therapy. Patients receiving only ATP did not have a significantly increased risk of death (HR 0.73; 95% CI 0.34–1.57; P = .41), while the effect remained in the shock group (HR 1.55; 95% CI 1.06–2.26; P = .02) in multivariate Cox analysis.
Shock burden was analyzed by looking at the cumulative number of days with shocks (shock days), cumulative number of shocks, and episodes of shock storm. In the adjusted analyses, those patients with 1–5 shock days did not have a significantly increased risk of death (HR 1.30; 95% CI 0.88–1.94; P = .19), while those with 6–10 shock days (HR 2.22; 95% CI 1.21–4.08; P <.01) and >10 shock days (HR 3.66; 95% CI 1.86–7.19; P <.01) had increasingly higher risk (Figure 3). Likewise, patients who received one to five total shocks did not have an increased risk of death (HR 1.08; 95% CI 0.68–1.70; P = .75), while those receiving 6–10 shocks (HR 2.07; 95% CI 1.20–3.57; P <.01), or > 10 shocks (HR 2.31; 95% CI 1.43–3.74; P <.01) had a greater than twofold increased risk of death as compared with patients who received no shocks.
Using time-dependent analysis and with the exclusion of subjects with shock storm, the presence of five or more shocks remained significantly associated with mortality. However, when compared with those without shock storm, the presence of shock storm was not associated with increased mortality in individuals with five or more shocks or ≥5 shock days (Table 2).
Overall, the results of this study raise concerns that shocks themselves may be causally associated with an increased risk of death and are not just a marker of increased risk due to a deteriorating substrate that leads to the arrhythmias and shocks. The relationship between shocks and increased risk of death persisted after adjustment for baseline SHFM predicted mortality and other risk factors. The SHFM is a validated measure for the prediction of survival in heart failure patients that provides an estimate of survival forward in time based on a large number of clinical, pharmacological, device, and laboratory characteristics.
Use of this measure allowed us to account for the changing risk of death in this high-risk population and suggests that differences in predicted survival at baseline do not entirely account for the increased mortality associated with shocks.
Subsequent subgroup analyses were consistent with the conclusion that shocks may be detrimental per se. Our analysis comparing patients receiving no therapy with patients receiving ATP only indicated that ATP is not associated with increased risk of death. This is in agreement with the findings reported in previous studies.
The strongest associations demonstrated in this study involved the dose of ICD shocks. This is a unique finding as the Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT) and the Multicenter Automatic Defibrillator Implantation Trial II (MADIT II) analyses used therapy episodes with one or more shocks but did not have data on the total number of shocks.
Patients in our study who received more shocks or had more days with shocks had a significantly increased risk of death as compared with patients receiving no shocks. These results indicate that shock burden in the forms of cumulative days with shock, and cumulative number of shocks, may play an important role in the relationship between shocks and increased risk of mortality. In our analysis, this association persists when controlling for delivery of multiple shocks in 1 day (shock storm). Previous studies have found that an increased number of shock episodes confers greater risk of death. Sweeney et al
reported that patients with episodes of ventricular arrhythmia (VA) and shocks have higher mortality (with ≈20% increased risk per shocked episode) and that VA occurrence rates, durations, and electrical therapy burden were highest among patients who were shocked and died. Our analysis comparing patients with ≥5 shock days/total shocks who experienced episodes of shock storm with those without shock storm did not indicate an increased risk of death in the shock storm group. Electrical storm, commonly defined as three or more ventricular tachyarrythmia detections in 24 hours treated by ATP or shock or eventually untreated, is associated with higher mortality than isolated VT/VF.
Given previous findings, along with the limitation of the small sample size in our analysis, this is a question that should be addressed in future studies. It is worth noting that the categories chosen for our dose-response analyses were arbitrary and that categories for the subgroup analysis not including shock storm patients were chosen based on the definition of our shock storm variable. Therefore, our results do not imply that there is no increased mortality risk for patients receiving less than five shocks, and no conclusion can be drawn as to an exact number of shocks associated with decreased survival. Instead, the conclusion from the shock burden analysis is that an increasing dose of shocks appears to be detrimental. Whether this is the case when shocks are delivered over different time frames is a question that remains to be answered.
Analysis of inappropriate versus appropriate shocks has had mixed results in prior studies, with MADIT II and SCD-HeFT finding a twofold increased risk of death
reporting no increased mortality risk. The inconsistency among studies may reflect the fact that this is an inherently difficult analysis to perform as the means of stratifying these groups is complicated. Simply stratifying patients into those who only have appropriate shocks or only have inappropriate shocks isolates the type of shock exposure but leaves out a large proportion of patients with both exposures. It may be possible to use a longitudinal design to compare survival in intervals of all appropriate shocks against all inappropriate shocks; however, the present analysis considered the overall effects of shocks to mortality risk and does not address the risk of appropriate versus inappropriate shocks.
The underlying mechanism for why shocks may be detrimental is currently unclear, although there is literature outlining the adverse effects of shocks on myocardial function. The mechanism is likely a composite of alterations in electrophysiological function, hemodynamic function, molecular and neurohumoral changes, and direct myocardial damage interacting with the underlying substrate.
have shown that ICD shocks >9 J delivered during sinus rhythm or VF resulted in a 10%–15% reduction in the cardiac index but that shocks of lesser energy did not cause this reduction. This and similar findings may help to explain why shocks, but not ATP, have been associated with increased risk of death. Additionally, patient discomfort and anxiety associated with ICD shock therapy deserves mention. Patients with ICD shocks have increased levels of psychological distress, anxiety, anger, post-traumatic stress disorder, and depression as compared with patients who do not receive shocks, and these psychological sequelae may be a contributing factor to the increased mortality seen in patients who receive ICD shocks.
It is clear that more research is needed to help further elucidate the underlying mechanisms for how shocks might be contributing to mortality risk. However, regardless of the mechanism, the stakes are high given that the annual insertion of ICDs has increased by 20-fold in the past 15 years, with the National ICD Registry reporting the implantation of nearly 500,000 ICDs between the years of 2006 and 2009 and registry implants accruing at the rate of 10,000 per month.
If, in fact, shocks are detrimental, the observed clinical efficacy of ICD therapy may be the result of competing influences, with shocks terminating potentially fatal arrhythmias but also increasing the risk of death through other mechanisms. Previous studies, such as the Defibrillators in Nonischemic Cardiomyopathy Treatment Evaluation (DEFINITE) trial, found that not all shocks delivered for episodes of VAs are necessary and that many such rhythms would spontaneously convert to normal rhythms without therapy.
It follows that if the same number of life-threatening arrhythmias could be terminated with fewer shocks and shocks could be used less often for self-terminating or non-life-threatening arrhythmias, then perhaps the overall efficacy of ICD therapy could be improved. What is needed are prospective randomized trials looking at different programming options that would reduce the number of shock such as (1) increased use of ATP, (2) more aggressive use of discriminators designed to prevent inappropriate therapy, (3) increasing the heart rate that will trigger therapy, and (4) delaying therapy to give more rhythms a chance to self-terminate so that less therapy of any kind including shocks is required. Each potential fix carries potential risks related to delaying therapy with an increased risk of hemodynamic compromise before definitive therapy is delivered or by preventing therapy all together for potentially life-threatening arrhythmias. Randomized trials are required to determine whether the net effect of these interventions actually reduces shocks and prolongs survival.
Our sample consisted of a diverse ICD population from a single VA medical center including both primary and secondary prevention with a long follow-up period. The heterogeneity of our population may be considered a strength for evaluating ICD shocks in a real-world clinical practice; however, our results may not be generalizable to other populations. The nature of data collection, namely, chart review, is subject to missing data and misclassification that may have affected study results. Calculation of the SHFM score required imputation of certain variables, and all calculations where made assuming implantation of a standard ICD. Survival is difficult to predict in heart failure patients, and the ability to accommodate for a patient's changing risk over time is a challenge in any analysis involving this patient population. The use of the SHFM was an attempt to capture this changing risk. While repeat measurements of baseline risk factors may have added to our ability to control for changing mortality risk, when to remeasure such factors to capture changing risk is a complex question. Given that this is a retrospective analysis, nothing can be said about whether ICD shocks are truly causative of increased risk of death.
The results of our study indicate that shocks may contribute to increased total mortality. Patients receiving cumulatively more shocks or more days with shocks are at increased risk for death. Further prospective research, in the form of a randomized clinical trial, is needed to look at optimizing ICD therapy.
Improved survival with an implanted defibrillator in patients with coronary disease at high risk of ventricular arrhythmias.