The benefits of implantable cardioverter-defibrillators (ICDs) have been shown in randomized clinical trials. The factors that affect the risk–benefit ratio in a community setting, however, have not been evaluated.
Sudden cardiac arrest is responsible for a substantial proportion of total mortality in the United States and other developed countries.1 Prevention of a future event among cardiac arrest survivors is a central objective in managing the care of these patients. Implantable cardioverter-defibrillators (ICDs) are a widely accepted management approach for preventing cardiac arrest. These devices have been shown to decrease arrhythmic death and all-cause mortality in survivors of cardiac arrest and to prevent sudden cardiac arrest when used as primary prophylaxis in patients known to be at risk for ventricular tachyarrhythmias.2
The use of ICDs in the community setting has increased substantially in Canada and the United States over the past decade.3 Despite the benefits of ICDs shown in several large randomized clinical trials,4-9 the factors that affect the risk—benefit ratio of these devices in a “real-world” setting have not been fully explored. Knowledge of such factors may help clinicians determine which candidates would derive the most benefit from these devices. Consequently, potentially unnecessary invasive procedures could be averted in those who would not benefit from ICD placement. Because randomized trials are highly selective in the types of patients who are enrolled, we sought to investigate the effects of age, sex, and comorbid conditions on mortality following ICD implantation in a community-based setting.
Subjects and methods
Using the CIHI (Canadian Institute for Health Information) database, which includes the records of all hospital admissions and procedures in Ontario, Canada, we assessed subjects who received ICDs between April 1, 1997, and March 31, 2003. Survival of ICD recipients was compared with matched control subjects who were randomly selected from the Registered Persons Database, which contains information on residents of Ontario and their vital status. Nondefibrillator control subjects were matched to ICD recipients based on age, heart failure status, history of arrhythmia, and presence of comorbidities.
We identified comorbid conditions using all recorded diagnoses during the 3 years before the date of ICD implantation and the secondary diagnosis fields of the CIHI database from the ICD implant admission. The Deyo-Charlson comorbidity classification system was used to categorize the comorbidities. We documented the frequency of cardiac and noncardiac comorbidities in ICD recipients, as well as unadjusted 1- and 2-year mortality rates.
We divided age into 3 categories: 18 to 64 years, 65 to 74 years, and 75 years or older. Variables associated with mortality after ICD insertion were analyzed using Cox proportional hazards regression. Using cubic spline analysis, we determined the effect of age continuously after correcting for sex and all significant comorbid conditions. The Kaplan-Meier method and adjusted survival curves were used to evaluate survival. All patients were followed for at least 2 years after ICD placement, at which time they were censored unless death occurred during the follow-up period. Given the potentially important consequences of heart failure on mortality, we analyzed the effect of the number of previous heart failure events occurring in the 3 years before ICD placement on mortality after device implantation.
A total of 2467 subjects, with a mean age of 62.5 ± 13.4 years, received an ICD during the study period. Men made up 79% of subjects and tended to be older than women. Diabetes, peripheral or cerebrovascular disease, respiratory disease, and renal disease were the most common noncardiac conditions. The frequency of these and other comorbidities are shown in Table 1. At 1 year, overall mortality was 7.8%, and at 2 years, mortality was 14.0%.
When adjusted for sex and Deyo-Charlson comorbidity score, the hazard ratio (HR) for mortality for the older age groups (65-74 years and ≥75 years) were 2.05 (95% confidence interval [CI], 1.70-2.47) and 3.00 (95% CI, 2.43-3.71), respectively, relative to those more than 65 years of age (both <.001). There was no significant difference in mortality risk by sex of the ICD recipient (HR = 1.15; 95% CI, 0.93-1.41) when differences in age and the Deyo-Charlson comorbidity score were accounted for in the analysis.
In multivariable analysis, comorbid conditions associated with death included chronic obstructive pulmonary disease, diabetes with microvascular complications, renal insufficiency, peripheral vascular disease, and heart failure (Table 2). Relative to those without noncardiac comorbidities, the HRs for death increased with a greater number of noncardiac comorbidities. Hazard ratios adjusted for age, sex, and previous heart failure were 1.72 (95% CI, 1.44-2.05; <.001), 2.79 (95% CI, 2.15-3.62; <.001), and 2.98 (95% CI, 1.74-5.10; <.001) for those with 1, 2, and ≥3 noncardiac comorbidities, respectively.
An episode of heart failure before ICD placement was associated with a doubling of the risk of death (Figure). Those with a recent episode of heart failure (occurred within 6 months before ICD implantation) had a near tripling of mortality risk, with an adjusted HR of 2.98 (95% CI, 2.41-3.69). Even heart failure events that occurred more remotely (eg, the most recent prior event occurred more than 6 months before ICD implantation) had an HR for mortality of 2.06 (95% CI, 1.71-2.48). These effects were statistically significant (both <.001), even after adjustment for age and the presence of significant noncardiac comorbid conditions. Similarly, patients who had more frequent heart failure events before ICD placement had a greater risk of death, which increased with the number of previous heart failure events. Previous cardiac arrhythmias, including past occurrence of cardiac arrest, ventricular fibrillation, or ventricular tachycardia, were not associated with death in ICD recipients after multivariable adjustment, which likely reflects the very high efficacy of the defibrillator in treating potentially lethal ventricular tachyarrhythmias.
Compared with nondefibrillator controls, there was a trend toward an overall survival benefit in ICD recipients after multivariable adjustment (HR = 0.86; 95% CI, 0.72-1.02; = .09). There was a correlation between ICD implantation and a significant increase in survival among those with previous ventricular tachycardia (adjusted HR = 0.80; 95% CI, 0.65-0.99;
= .043). A 2% absolute difference in mortality was shown at 2 years among those with previous ventricular tachycardia or cardiac arrest and a 1% absolute difference was shown overall, both favoring the ICD.
In this study, we identified the associations of several factors that influenced survival among ICD recipients in a “real-world” population setting; therefore, these findings have practice implications that extend beyond the confines of controlled clinical trials. Our principal findings related to the effects of age, comorbidities, and heart failure status on survival among ICD recipients.
First, we found that increasing age conferred an increased risk of death after ICD placement, with an accelerated increase in risk after 70 years of age. This is in contrast to an analysis of the CIDS (Canadian Implantable Defibrillator Study) study, in which ICD placement was found to be more beneficial in older patients.10 A plausible explanation for the contrasting results is that in a population-based sample of individuals, older patients will have a greater burden of comorbid conditions, as our data demonstrated. However, even after adjustment for comorbidities, older patients still had an increased mortality risk compared with younger patients.
Second, several common noncardiac comorbidities were associated with a higher rate of death after ICD placement, particularly, renal failure, chronic obstructive pulmonary disease, peripheral vascular disease, and diabetes with microvascular complications. Data reported from other jurisdictions have shown that patients with comorbidities, such as renal failure, liver failure, and cancer, are less likely to receive an ICD than those without these conditions.3 This suggests that some clinicians may empirically be deciding to restrict the use of ICDs in those with potentially life-limiting comorbidities. Our study confirmed that such statistical associations exist, and we quantified the effects of comorbidities on mortality.
Third, heart failure was the only cardiac disease that significantly influenced mortality after ICD placement, indicating the critical importance of characterization of this condition for determining survival. The SCD-HeFT (Sudden Cardiac Death in Heart Failure Trial) trial also indicated the potential importance of heart failure on survival among ICD recipients. Compared with nondefibrillator controls, SCD-HeFT showed a survival benefit of ICD placement in patients with New York Heart Association functional class II heart failure, but not in those with functional class III heart failure.11 Careful characterization of heart failure status may be highly informative in determining the potential benefit of ICDs, but further research is required.
Finally, ICD implantation was associated with significantly improved survival in patients with previous ventricular tachycardia compared with control subjects who had previous ventricular tachycardia but did not receive an ICD. Similarly, prior arrhythmia did not predict mortality among ICD recipients, which likely reflects the uniform efficacy of the defibrillator for treating ventricular tachyarrhythmias. Our study suggests that as ICDs become used more widely, the presence of competing comorbidities and progression toward end-stage heart failure will be of increasingly greater clinical importance.
The potential limitations of our study should be considered. Despite the vast amount of data contained in administrative data sources and the excellent degree of correlation between administrative and clinical data sources,12 clinical characterization of cardiac and noncardiac comorbidities is limited. Future studies with clinical data sources could potentially refine the predictors of adverse outcomes after ICD placement in a population-based sample. Some clinical variables that may have affected mortality outcomes were not available in the administrative databases used in this study. These included left ventricular ejection fraction and electrophysiologic inducibility of ventricular tachycardia. Likewise, continuous variables, such as laboratory values, were not included in the analysis because of the restriction of administrative data to binary variables. Some of these variables, such as serum sodium, have been shown in previous studies to predict mortality in heart failure patients.13
As with any invasive treatment, the risk—benefit ratio of ICD placement should be carefully considered in clinical decision-making. We sought to identify factors that may affect survival of ICD recipients after device implantation. In this evaluation of a community-based population of ICD recipients, we found that increasing age and the presence of noncardiac comorbidities and heart failure, but not a patient’s sex or history of previous arrhythmia, were associated with an increased risk of death after ICD placement. Knowledge of these factors may provide clinicians with important information that will help to determine whether a patient being considered for an ICD would benefit from it.
The authors have no relationship with any commercial entity that might represent a conflict of interest with the content of this article.