Author + information
- Alfred E. Buxton, MD⁎ ()
Reprint requests and correspondence:
Dr. Alfred E. Buxton, Cardiovascular Division, Department of Medicine, Beth Israel Deaconess Medical Center, 185 Pilgrim Road, West, Baker 4, Boston, Massachusetts 02215
Last week, an 83-year-old man was transferred to our service from another hospital for placement of an implantable cardioverter-defibrillator (ICD). He was hospitalized at the other institution because he had developed a retroperitoneal bleed while anticoagulated because of atrial fibrillation. He had a nonischemic dilated cardiomyopathy (absence of significant coronary obstructions had been proven by an earlier catheterization). His ejection fraction (EF) was 20%, and his New York Heart Association functional class was III. He had a history of chronic obstructive lung disease, and a serum creatinine level of 1.8 mg/dl, and no symptomatic ventricular tachyarrhythmia had been documented. Because nonsustained ventricular tachycardia was seen on telemetry, a consulting cardiologist told the patient's family that he “needed” an ICD.
It was unfortunate that the cardiologist said that the patient needed an ICD, because this wording conveyed to the family the idea that the ICD would save his life. It would have been far more appropriate for the physician to explain to the patient and his family what his estimated survival chances over a 2-year to 3-year follow-up period would be with versus without an ICD, as well as the likely impact of the ICD on his quality of life. This was obviously not done, probably because the consultant had no idea of the patient's estimated survival with or without an ICD. Instead, the consultant probably knows that several randomized clinical trials have demonstrated statistically significant reductions in mortality with ICD therapy, and he believes that current guidelines state that any patient with a left ventricular EF ≤30% should receive an ICD.
This case illustrates a number of issues we face today in trying to reduce the risk for sudden unexpected cardiac death in patients with recognized cardiovascular disease. Let us review some of these. First is the fact that clinical trials must have clearly defined entry and exclusion criteria. These criteria are then used to formulate practice guidelines. However, patients enrolled in clinical trials have multiple characteristics that influence their intrinsic risk for endpoints such as sudden cardiac death. The heterogenous nature of patients means that the average risk reduction observed with a treatment in a clinical trial may actually not pertain to many patients enrolled in the trial (1). A consequence of the wide variation in risk for total mortality and sudden death of patients enrolled in a given trial is varying survival benefit of treatment. For example, the relative risk reduction for total mortality with ICD therapy was 54% in the Multicenter Unsustained Tachycardia Trial and MADIT (Multicenter Automatic Defibrillator Implantation Trial), 31% in MADIT II, and 23% in SCD-HeFT (Sudden Cardiac Death in Heart Failure Trial) (2–5). However, multivariate analyses have demonstrated that in MUSTT and MADIT II, 25% to 30% of the study populations had 2-year total mortality risk of 5% and 8%, respectively, and in SCD-HeFT, 20% of the study population had 2-year mortality of 5%, without ICD therapy (6–8). Populations with 2-year total mortality of 5% to 8% can be expected to have a yearly sudden death risk of 1% to 2% (a level of risk at which there is unlikely to be significant ICD benefit). Conversely, in MADIT II and SCD-HeFT, 20% of the study populations had 2-year mortality of 30% to 50%. The wide variation in risk within these trials translates into vastly different treatment effects for a similar relative risk reduction of 20% to 30%. That is, a subject whose baseline risk is 2% to 3% per year will see much less absolute benefit from treatment than a person with 15% to 20% annual risk (1).
However, estimating the potential benefit of ICD treatment is more complex than just knowing predicted total mortality for an individual patient, because there are multiple potential causes of mortality, and the ICD can reduce only those deaths that occur from ventricular tachyarrhythmias. Paradoxically, those patients at highest mortality risk did not experience any survival benefit from the ICDs in the randomized trials because of 2 factors. First, multiple competing conditions lead to nonarrhythmic death, accounting for more deaths as overall risk increases. Second, sudden death accounts for a progressively smaller proportion of mortality as overall mortality risk increases (7,8).
Thus, there is a “sweet spot” range of mortality in which the risk for sudden death is high enough, and competing risk for nonsudden death is low enough, that an ICD can affect a significant reduction in mortality. How can we best identify those patients most likely to benefit? The published models referred to previously are a reasonable starting point. However, although current ICDs have not changed significantly from those used in the randomized trials, other cardiac treatment is very different at present compared to that used in the trials. The use of beta-adrenergic blocking agents and other pharmacotherapy was not as frequent as is appropriate for patients with abnormal left ventricular function. Modes of coronary revascularization were not the same during the trials. In addition, there may be less obvious differences between patients enrolled in clinical trials and those in which the guidelines are applied at present. Thus, there is a need to develop valid contemporary algorithms to guide ICD use for the primary prevention of sudden unexpected cardiac death.
In this issue of the Journal, Bilchick et al. (9) describe a model to predict risk for mortality based on 2 populations of Medicare patients who received ICDs for the primary prevention of sudden death from 2005 to 2007. Their model identified risk factors common to earlier models, as well as some new variables: age ≥75 years, New York Heart Association class III, chronic kidney disease, left ventricular EF ≤ 20%, atrial fibrillation, chronic obstructive pulmonary disease, and diabetes mellitus. In this analysis, 37% of patients in the development cohort of approximately 18,000 ICD recipients and 31% of the 28,000 patients in the validation cohort died over a median follow-up period of 4 years. The 20% of patients in the highest risk group had 2-year mortality of almost 40%, while the 20% in the lowest risk group had 2-year mortality of 7%. If their model is applied to the patient described earlier, he would have a projected 2-year mortality of 65%, despite having an ICD implanted. Would the ICD benefit this patient? I think the chances are small.
This study joins a growing number of similar efforts. This year alone, 3 other reports of models to predict mortality in ICD recipients have been published (10–12). These models have a number of factors in common with Bilchick et al.'s (9) model and share similar limitations. They are based on analyses of patients who received ICDs. Thus, they cannot be used to estimate ICD benefit because none included untreated patients. Nevertheless, all identify substantial subgroups of ICD recipients with very high total mortality. It seems highly unlikely that ICDs were helpful to those patients who died within 2 years after implantation. Second, the variables incorporated into each of these models bear no mechanistic relation to ventricular tachyarrhythmias. Thus, they cannot possibly target those patients with the greatest chance to benefit from the ICD: those with the substrate to support ventricular tachycardia or fibrillation. Third, because the endpoint in each study is total mortality, they do not provide insight into a second subgroup of patients receiving ICDs under current guidelines: those at low risk for ventricular tachyarrhythmias despite low EFs. In previous analyses, these patients have accounted for at least 25% of patients meeting guidelines for primary prevention ICDs.
Two requirements must be met to validate the ability of models to identify patients at low risk for sudden death despite reduced EFs. First, a prospective study format must be used along with uniform programming of ICD detection criteria to accurately understand the meaning of “appropriate” ICD discharges, or a matched untreated control group must be compared with ICD recipients. Second, tests capable of identifying patients having 1 or more of the electrophysiologic substrates to support re-entry, triggered activity, or automaticity in ventricular muscle must be incorporated into the model. One such test has previously been validated as a method to guide ICD therapy: programmed electrical stimulation (electrophysiologic testing) (13). However, electrophysiologic testing must be supplemented by other markers of arrhythmic risk, such as left ventricular hypertrophy, because not all ventricular tachyarrhythmias result from re-entry.
There is a message here. Multiple investigators and clinicians in the United States and other countries realize that current guidelines are inadequate to use ICDs in the most appropriate, cost-effective manner for primary prevention of sudden death. They must be changed. However, change will require the completion of 1 or more randomized controlled trials that demonstrate superiority (or at least noninferiority) of a multivariate model approach to the current standard. In our current climate, clinicians are extremely hesitant to randomize patients to a therapy that goes against current guidelines. The conduct of a successful randomized clinical trial will require the cooperation of multiple governmental agencies and third-party payers, together with informed clinicians. Clinical cardiologists in turn must be willing to take the time to explain to their patients what ICDs can and cannot do. Clinicians must also be able to estimate patients' risk for mortality with and without ICDs. This is the next logical step, and we must take it. We owe it to the practice of medicine. We owe it to our patients!
Dr. Buxton is a member of the Events Committees of Clinical Trials (not related to the subject matter of the current manuscript) for both Boston Scientific and Medtronic; and has received honoraria from Medtronic.
↵⁎ 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.
- American College of Cardiology Foundation
- Buxton A.E.,
- Lee K.L.,
- Hafley G.E.,
- et al.
- Goldenberg I.,
- Vyas A.K.,
- Hall W.J.,
- et al.
- Levy W.C.,
- Lee K.,
- Hellkamp A.,
- et al.
- Bilchick K.C.,
- Stukenborg G.J.,
- Kamath S.,
- Cheng A.
- Barsheshet A.,
- Moss A.,
- Huang D.,
- McNitt S.,
- Zareba W.,
- Goldenberg I.
- van Rees J.B.,
- Borleffs C.J.,
- van Welsenes G.H.,
- et al.