Selecting Patients for ICD Implantation: Title and subTitle BreakAre Clinicians Choosing Appropriately?

Large-scale registries provide an important window into how medicine is actually practiced. In this issue of JAMA, Al-Khatib et al¹ examined the National Cardiovascular Data Registry's (NCDR’s) Implantable Cardioverter-Defibrillator (ICD) Registry to determine how often ICD implantations were performed in patients who did not meet criteria suggested in current guidelines for ICD use² ; specifically, ICD implantation within 40 days of an acute myocardial infarction (MI) or within 3 months of coronary artery bypass graft (CABG) surgery, recent diagnosis of congestive heart failure (CHF), and presence of New York Heart Association (NYHA) class IV symptoms. Al-Khatib et al¹ found that almost one-quarter of ICD implantations were performed in patients with these characteristics, mortality and complication rates were higher among these patients than for patients who received an ICD based on evidence-based guidelines, and nonelectrophysiologists performed non–guideline-based ICD implantations with a higher frequency than electrophysiologists. These findings should be used to inform public health policies toward the appropriate use of this life-saving but expensive technology.

The first question that needs to be addressed involves the reliability of the data. The ICD Registry is a well-audited tool that is robust and provides important information. Its strengths lie in the large number of patients included in the registry and in the quality of the data monitoring process. Nonetheless, some variables in the registry may not be accurate. For example, more physicians self-reported being board-certified electrophysiologists in the ICD Registry than have actually been board certified.³ In addition, the largest group of patients who received non–evidence-based ICDs in the report by Al-Khatib et al¹ were those with a recent diagnosis of CHF. In most cases, ICD Registry data are retrospectively abstracted from the medical record by nurses or other health care professionals. The precise onset of CHF may not be well documented in the medical record because unlike the date of MI, evaluating the date of diagnosis of CHF may require a history that is targeted to this variable. The data abstractor may find evidence of CHF based on a recent hospitalization, but a thorough history could identify a much earlier onset. Even though these issues may affect some of the quantitative findings, the qualitative findings from this report are not likely to be affected significantly, particularly given the large size of the registry.

Is it appropriate to implant ICDs for patients who do not meet specific criteria recommended in the guidelines? The authors¹ note that there are circumstances in which deviations from guidelines are appropriate based on the physician's judgment. Accordingly, while there will be some background rate of ICD implantation outside of guideline recommendations, it is unclear what this background rate should be. As shown in Figure 1 of the article by Al-Khatib et al,¹ only a very small number of sites had a non–guideline-based ICD implantation rate of less than 6%, suggesting this might be a reasonable lower bound. The upper bound is more difficult to determine, but should be less than the 21% rate of non–guideline-based ICD implantation noted among cardiac electrophysiologists.

Moreover, for 3 of the 4 indications Al-Khatib et al¹ classified as non–evidence-based ICD implantation, the issue is timing of the ICD implantation. For instance, a valid medical concern would be that a patient who recently experienced an MI, underwent CABG surgery, or was diagnosed with CHF is at high-risk for sudden death during the waiting period that is indicated in the guidelines. However, no data are available in the NCDR's ICD Registry regarding this assessment or whether alternative approaches such as the use of the wearable cardioverter-defibrillator were considered. In addition, while there are convincing data that early implantation of an ICD after an MI⁴ does not improve outcomes, the data may be less convincing for other conditions. One post hoc analysis,⁵ although not definitive, has shown that some patients with nonischemic cardiomyopathy may benefit from early ICD implantation. However, the use of non–cardiac resynchronization therapy ICDs in patients with NYHA class IV symptoms is unwarranted.

The other critical findings of this study were the higher rates of non–guideline-based ICD implantation performed by nonelectrophysiologists (cardiologists, thoracic surgeons, and others [ie, internists and surgeons]), as well as the increased complication rate for non–guideline-based ICD implantation. In particular, there was an excess of 4 deaths per 1000 ICD implants when the device was implanted outside the guidelines. A logical conclusion from these data would be to promote more guideline-based use of ICDs. Of note, another study from the ICD Registry has suggested better outcomes when ICDs are implanted by physicians who have had extensive training in electrophysiology.⁶ In this study,⁶ board-certified electrophysiologists or those who had undergone electrophysiology training had a lower complication rate from ICD implantation than other groups of physicians. Taken together, these 2 studies suggest that intensive training may improve both the preoperative evaluation of patients, as well as the operative and immediate postoperative care of patients undergoing ICD implantation.

It is important to consider the potential reasons for improved outcomes when ICDs are implanted by electrophysiologists. First, these outcomes may be related to increased volume of procedures, a factor that has been related to improved outcomes in the ICD Registry⁷ and in other settings.⁸ It is also possible that electrophysiologists have more detailed training regarding ICDs, including preoperative alternatives, operative techniques, and postoperative care. In fact, the addendum to the Heart Rhythm Society's clinical competency statement⁹ notes that after 3 years from its publication date physicians who wish to implant ICDs need to receive formal training in an approved fellowship training program. This indicates the committee's appreciation of the need for a broad level of training that is generally not available to nonelectrophysiologists. In the absence of formal training as advocated by the Heart Rhythm Society's clinical competency statement, there are no widely applied national standards for competency in ICD implantation for nonelectrophysiologists, although the Heart Rhythm Society document⁹ does provide guidelines. The NCDR data support the concept that the absence of formal training results in poorer identification of appropriate indications for ICD implantation and higher complication rates.

Another important implication of this study involves the economic consequences of performing ICD implantation outside of the guideline-based recommendations. Avoiding non–cardiac resynchronization therapy ICD implantation in patients with NYHA class IV symptoms most likely will result in cost savings and improved cost-effectiveness. However, many of the other indications relate to timing of ICD implantation. The question arises whether patients who undergo ICD implantation during the waiting periods would ultimately receive an ICD. It is well documented that following an MI^{10 - 11} or after initiation of medical therapy for recent-onset CHF,¹² left ventricular ejection fraction may improve over time and increase beyond the guideline range for ICD implantation. If a substantial number of these early (within the waiting period) ICD implantations do not meet the guidelines beyond the waiting period, substantial savings could be realized by considering alternative approaches for prevention of sudden cardiac death during the waiting period, such as the wearable cardioverter defibrillator.

To improve public health, the cardiovascular care community must act on these data. There are several important considerations. Further information and specific data are needed to characterize some of the issues, such as how well the NCDR captures some of the subtleties of ICD indications and whether reasons for deviations from the guidelines can be captured accurately. Once this is accomplished, it is possible that prospective data entry in an online system can be developed to provide immediate feedback regarding the presence or absence of an evidence-based indication for an individual patient prior to ICD implantation. It is likely that all physicians require further education to understand the rationale for the guidelines and potential alternative approaches when a patient does not meet guidelines for ICD implantation. In addition, as a matter of public policy, health care organizations must assess whether quality of care and cost-effectiveness can be improved by mandating the Heart Rhythm Society's guideline for formal training in an approved fellowship training program. If properly applied, the findings of the study by Al-Khatib et al¹ may improve practice patterns and outcomes, with the unique opportunity to do so while lowering health care costs.