Author + information
- Gregory M. Marcus, MD, MAS∗ ( and )
- Thomas A. Dewland, MD
- Department of Medicine, Division of Cardiology, Electrophysiology Section, University of California–San Francisco, San Francisco, California
- ↵∗Reprint requests and correspondence:
Dr. Gregory M. Marcus, Department of Cardiology, Electrophysiology Section, University of California–San Francisco, 505 Parnassus Avenue, M-1180B, Box 0124, San Francisco, California 94143-0124.
Clinical trials have convincingly demonstrated the efficacy of implantable cardioverter-defibrillators (ICD) for the prevention of sudden cardiac death in appropriately selected patients (1–3). As a result, over 140,000 ICD-related procedures are performed each year in the United States (4). Despite evidence that defibrillators save lives on a population level, medical decision making for individual patients can be difficult. Our limited ability to discern which ICD candidates will derive benefit from their device and which are most prone to complications contributes to the complexity of ICD counseling. In addition, the overall cost associated with defibrillator implantation has generated an interest in ICD outcomes that extends far beyond the doctor-patient encounter. Ongoing changes in the healthcare landscape, including provisions fundamental to the Affordable Care Act, promise heightened emphasis on quality of care and increased public scrutiny of procedural complications.
In the context of these clinical uncertainties and growing administrative oversight, multiple studies have attempted to identify risk factors for ICD complications. A variety of patient, procedural, and operator characteristics have been linked to adverse outcomes (5–7), and in 2011, an ICD implantation risk score was published using data from the NCDR (National Cardiovascular Data Registry) ICD Registry (8). This previous model included 10 variables independently associated with inpatient complications and provided the implanting physician with a metric to gauge procedural risk. In this issue of the Journal, Dodson et al. (9) present a second clinical risk score using more contemporary data to predict the likelihood of inpatient complication or death after ICD implantation or generator change. They developed an algorithm using 21 variables included on standard NCDR ICD registry forms to derive a score that can be used to quantify procedural risk. To enhance clinical relevance, a simpler 12 variable model is also presented.
The Dodson algorithm has several strengths. First, the investigators used real-world data from the NCDR ICD Registry to develop their model. This registry contains detailed information from over 1,400 centers for each Medicare-funded primary prevention device procedure (4). It is also noteworthy that the investigators divided their cohort into derivation and validation groups and demonstrated similar model performance in both populations. Whereas the validation dataset is not “external” and therefore does not reveal how well the risk score performs among patients not enrolled in the NCDR, this methodology provides reassurance that model performance is not overly optimistic due to statistical “overfitting.”
This clinical risk algorithm has several potential applications. The most natural use of this score would be to estimate a patient's likelihood of procedural complication. Such an application has clear clinical merit, especially because temporal trends may diminish the accuracy of previously published data. However, we would caution against the use of this score to justify withholding ICD treatment, as it is possible that patients at greatest risk for an acute procedural complication may also derive the most benefit from device implantation. An analogous phenomenon has recently been described for atrial fibrillation stroke prophylaxis, whereby patients at the highest risk of bleeding on anticoagulant medications (as identified by the HAS-BLED [Hypertension, Abnormal liver/renal function, Stroke history or Bleeding predisposition, Labile international normalized ratios, “Elderly” (age ≥65 years), Drugs/alcohol use] score) also have the greatest risk of thromboembolic events (10). As a result, these high-risk patients actually derive the greatest absolute benefit from anticoagulation. Because several of the covariates in the clinical risk algorithm for ICD complications may be associated with a greater likelihood of ICD benefit (such as increased New York Heart Association class or a history of a cardiac arrest), this score should not be used in isolation to dissuade ICD implantation. In addition, it remains unclear how individual patients and physicians conceptualize absolute risk percentages and use these numbers for clinical decision making. For instance, the investigators chose to classify patients with a risk score ≥30 (absolute risk 4.2%) as “high risk.” Is it reasonable to counsel an ICD candidate that they are high risk for device implantation when the likelihood of an aggregate complication is less than 1 in 20, when the complication may be fairly minor, and when the therapy has mortality benefit?
A second application is to standardize complication rates across healthcare settings by adjusting for the risk score. By leveling the playing field, physicians and hospitals could then compare adverse event frequencies in a more equitable fashion. The investigators present risk-standardized complication rates to illustrate how their score could be used to benchmark hospital performance. Whereas these findings may provide reassurance that most centers deliver fairly comparable implantation results, it could be argued that a metric that only identifies 3.8% of hospitals as underperforming may not sufficiently bolster efforts to lower ICD complications. Moreover, whether the current risk score classifies hospital performance more appropriately than crude adverse event rates remains untested. Finally, although many hospitals contribute data to the NCDR for all ICD implantations, the feasibility of collecting and reporting similar data for the remainder of device procedures needs additional exploration. Other possible uses of the risk score, including identifying patients who can be discharged with limited post-operative observation, remain speculative and require further study.
The investigators have appropriately acknowledged the major limitations of their ICD risk score. Because of the data source, the model is limited to inpatient complications and relies on the accuracy of the data supplied by the implanting hospitals. In addition, even the parsimonious model lacks clinical convenience and may prove too unwieldy for immediate, “on the fly” calculations at the bedside or in a busy clinic. However, as the investigators point out, widespread adoption of electronic medical records and mobile technologies may render their risk score more accessible to the clinician.
The current investigation has also identified several specific findings that should influence clinical practice. Dual chamber ICDs, compared with single-chamber devices, are again associated with increased implantation risk (5), and operators should carefully weigh the decision to implant a dual-chamber defibrillator in a primary prevention patient without a pacing indication. It is also noteworthy that hematoma and pneumothorax are among the most common complications. When anticoagulation is required, implanting physicians should consider recent data demonstrating reduced pocket hematomas when certain device-related procedures are performed on uninterrupted warfarin (11). In addition, implanting physicians should be comfortable with techniques that may minimize the risk of lung injury, including extrathoracic axillary vein access or cephalic cutdown. Finally, the investigators have shown that device implantation during a nonelective admission is a strong risk factor for periprocedural complication. It should be remembered that primary prevention ICD implantation is an elective procedure that, in the absence of other urgent indications (such as the need for permanent pacing), should be deferred to the outpatient setting after medical optimization.
In light of their carefully constructed and well-performing risk model, Dodson et al. (9) should be applauded for advancing our understanding of ICD procedural outcomes. Device implantation is a calculated risk taken by a patient who wishes to trade the acute risk of implantation complication for chronic protection against arrhythmic death. The current effort better informs this decision and promises to play an important role in future attempts to evaluate physician and hospital performance.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Marcus has received research support from Medtronic. Dr. Dewland has received educational travel grants from Medtronic and Boston Scientific.
- American College of Cardiology Foundation
- Dewland T.A.,
- Pellegrini C.N.,
- Wang Y.,
- Marcus G.M.,
- Keung E.,
- Varosy P.D.
- Freeman J.V.,
- Wang Y.,
- Curtis J.P.,
- Heidenreich P.A.,
- Hlatky M.A.
- Al-Khatib S.M.,
- Greiner M.A.,
- Peterson E.D.,
- Hernandez A.F.,
- Schulman K.A.,
- Curtis L.H.
- Haines D.E.,
- Wang Y.,
- Curtis J.
- Dodson J.A.,
- Reynolds M.R.,
- Bao H.,
- et al.,
- on behalf of the NCDR
- Friberg L.,
- Rosenqvist M.,
- Lip G.Y.