We live in an era marked by a dramatic broadening of the indications for implantable cardioverter-defibrillator (ICD) implantation for the prevention of sudden cardiac death. Recent randomized clinical trials including MADIT II, DEFINITE, SCD-HeFT, and COMPANION show that defibrillator implantation may benefit large groups based on simple selection criteria (myocardial infarction, left ventricular dysfunction, congestive heart failure), without the need for traditional risk stratification. The purpose of this presentation is to review the data that support the current paradigm for prophylactic ICD implantation, and critically analyze the trial results. We will review recent data regarding the use of microvolt T-wave alternans for risk stratification, and close with some personal comments and philosophies about risk stratification. One focus will be the potential for good risk stratification to benefit not just the patients who are identified as being at low risk and who therefore might avoid device implantation, but also to benefit the high-risk patients as well. William Osler famously wrote, One of the first duties of a physician is to educate the masses not to take medicine. In modern times, it is likely that he would have extended that dictum to implantable devices as well. Nevertheless, there is now a strong imperative to implant ICDs in the masses. The first trial that substantially broadened the indications for ICD implantation for primary prevention was MADIT II.1 MADIT II evaluated patients with a prior myocardial infarction and left ventricular ejection fraction (LVEF) below 30%. There was no requirement for any qualifying arrhythmia (non-sustained left ventricular tachycardia), and there was no systematic assessment of inducibility using electrophysiologic testing prior to randomization. The trial proved that defibrillator implantation resulted in a statistically significant improvement in survival as opposed to conventional therapy. Those treated conventionally had a 78% survival rate after two years vs. 84% survival with ICD implantation. More recently, this same stratagem has been evaluated in patients with non-ischemic cardiomyopathies. The DEFINITE Trial enrolled 458 patients with non-ischemic cardiomyopathy and LVEF below 35% (the mean LVEF was 21%).2 Patients did need to have a qualifying arrhythmia, either non-sustained ventricular tachycardia, or in a small minority, frequent VPCs. Patients were randomized to ICD therapy versus optimal medical management. The results did not make nominal statistical significance (p = 0.08), but the two-year total mortality was 14% with medical management versus 8% with defibrillator implantation. In other words, there was a 6% absolute survival benefit associated with defibrillator implantation in this population: the magnitude as the survival benefit for the defibrillator in the ischemic population in MADIT II. SCD-HeFT, which was recently reported but has not yet been published, is the largest of all the primary prophylactic trials by far, enrolling over 2,500 patients.3 In order to be eligible for the trial, patients had to have an LVEF below 35%, and mild to moderately symptomatic heart failure (NYHA Class II-III). Patients with both ischemic and non-ischemic cardiomyopathies could be enrolled, and the split was just about even: 52% had ischemic cardiomyopathy, and 48% had non-ischemic cardiomyopathy. The mean ejection fraction among this population was 25 percent; 70% of enrollees were in Class II, 30% were in Class III. Of note, if one thinks about the potential benefit of cardiac resynchronization therapy, in addition to defibrillator therapy, the mean QRS duration was 112 milliseconds, and 41% of the patients had a QRS > 120 milliseconds. SCD-HeFT employed a three-way randomization between placebo, amiodarone and ICD implantation. The first salient point is that there was no mortality benefit to amiodarone therapy whatsoever, whether analyzed overall or in any individual subgroup. Mortality, with either placebo or amiodarone was approximately 15% at two years, and by the end of five years had reached 33-36%. In comparison, in the group randomized to ICD therapy, total mortality was reduced to approximately 10% at two years; at the end of five years, total mortality was reduced to 28% with the ICD. Similar to MADIT II and DEFINITE, this yields an 8% absolute mortality benefit, albeit now after five years of therapy. These are the results, and at face value, they certainly seem to support the concept that widespread application of ICDs in patients with left ventricular dysfunction will improve survival. However, take a closer look at these results, emphasizing the number of patients requiring intervention in order to save one life. In the earliest primary prevention trials (MUSTT and MADIT I) trials that used an electrophysiologic study for risk stratification prior to ICD implantation in order to save one life at the end of two to five years, we needed to implant devices in three to four patients. In MADIT II, we have to perform 11 surgeries for every life saved at the end of three years. The SCD-HeFT data suggest a need to implant 17 defibrillators to save one life at the end of three years. As implanting physicians, one cannot be satisfied with a paradigm requiring surgery on 17 patients in order to benefit one. In addition, from a societal perspective, we need to be concerned about the cost implications of such a strategy; ask at what point does the cost outweigh the benefit? COMPANION, a trial of resynchronization therapy (CRT) for congestive heart failure is also instructive in this regard. COMPANION was a randomized trial of a CRT-ICD versus a CRT-pacemaker versus optimal medical therapy.4 In order to be enrolled in COMPANION, patients needed to have a QRS > 120 milliseconds, an LVEF 5 Sixty percent of patients in this trial had a history of syncope or pre-syncope, and 27% had a history of prior or sustained ventricular arrhythmia. Interestingly, only 6% of patients in this population were being studied for primary indication of non-sustained VT. If the EP study was positive, there was a 25% rate of sustained ventricular tachycardia or death within 400 days (in other words, a positive predictive value of 25%). That was comparable to a positive predictive value of t-wave alternans at 26% in the same population. However, those patients with a negative EP study still had a 10% risk of sustained VT or death not a number anyone can be comfortable with. On the other hand, if the T-wave alternans test was negative, the 400-day risk of either sustained VT or death was only 3%. A recent study by Hohnloser et al.6 pooled data from two previously reported trials and analyzed the subset of 129 MADIT II type patients (those with a prior myocardial infarction and LVEF 7 In contrast, the risk in patients with a negative T-wave alternans test was extremely low. The available data thus suggest that the negative predictive value of T-wave alternans testing may be sufficient to take patients who would have been eligible for defibrillators based on a MADIT II/SCD-HeFT criterion and identify a group of patients who do not require ICD implantation. What I want to propose is that excluding the low-risk patients not only benefits them, it also benefits the high-risk patients. In order to explain why I believe that, it is first necessary to acknowledge the underutilization of ICDs for primary prevention. Ruskin and colleagues analyzed Medicare and managed care claims data in 2002 and estimated there was somewhere between 736-1140 ICD candidates per 1 million population of the United States.8 However, the actual implant rate in the US at that time was only 416 per million population. Thus, only one-half to one-third of potential ICD candidates in the U.S. are actually getting the implant. Why should that be? There are three barriers to ICD implantation in potential candidates, all of which can be favorably impacted by doing good risk stratification. One barrier is reluctance among referring physicians to address the issue of sudden cardiac death in patients who are feeling OK. This may be a very difficult conversation to have with a patient who may have had a heart attack many years ago, but who has been feeling well over the intervening years. The availability of a test like the microvolt T-wave alternans test provides an entry point to discuss the risk of sudden death with a patient and to explain the utility of ICD implantation. A second barrier is the concern about quality of life in ICD recipients in particular, concerns about risk of inappropriate therapies in these patients. The best available estimates are that about 20% of ICD recipients experience inappropriate shock at some point. It is one thing if patients also experience appropriate life-saving shocks. However, it is quite another thing to have a series of inappropriate shocks and never need to use the device to treat VT or ventricular fibrillation. Therefore, risk stratification that enables one to pull out the low-risk population swings the quality of life benefit in favor of the ICD recipient. A third barrier concerns the issue of cost. The ability to pull out low-risk patients obviously improves the cost benefit trade-off in the high-risk patient, and therefore facilitates implantation of devices in the high-risk patient. Thus, a series of benefits accrue with risk stratification: risk stratification benefits the members of the low-risk group, who avoid an unnecessary medical procedure. However, it also benefits the high-risk group by providing an entry point for discussion of sudden cardiac death and the utility of defibrillator implantation. The greater benefit in the high-risk patients allays important patient/physician concerns regarding cost, quality of life, and procedural morbidity. Not only is there a continued role for risk stratification in the modern era if anything, results of trials like MADIT II and SCD-HeFT have made the development of tools for accurate risk stratification imperative.