Microvolt T-Wave Alternans in CHF and MADIT II Patients

Daniel Bloomfield, MD
Daniel Bloomfield, MD
This is a slide that by now everyone in an EP lab is familiar with it is the MADIT II slide I think this study dramatically changed the landscape of how we do risk stratification. Prior to MADIT II, the name of the game of risk stratification was let s pick out the needle in the haystack patients that need an ICD. Since MADIT II, the wave has turned towards doesn t everyone need an ICD, and which patients don t need an ICD. That was a dramatically important change in the way we approach risk stratification. I am going to show a couple of slides on data that were shown in the past. This data is from an article by Thomas Klingenheben and Stefan Hohnloser the article appeared in The Lancet in 2000. This was a series of 107 congestive heart failure patients with a mixed etiology that all had the alternans test done with a follow-up over 18 months. What you can see here is that the alternans positive group had, approximately, a 20% event rate over 18 months, and the group that was alternans negative had no events at all. This is one of the first papers to show that you could really pick out a low-risk group of patients amongst the heart failure population. The next paper I want to mention is a paper by Dr. Ikeda from Japan. This was a study of nearly 1,000 patients post-MI. Three tests were done here: alternans, ejection fraction (EF), and SAEKG. This was a straight post-infarction population, where over 98 percent of the population had an angioplasty as the primary therapy for revascularization, so overall it was a very well-treated group, and the mortality as you d expect in this population was pretty low. If you look at the top curve, this is alternans negative and positive. Despite the fact that this group is a very low-risk group, the alternans was able to distinguish between the low-risk and high-risk patients. This ejection fraction was cut in 40 percent, and once again it is as you'd expect: the people with an EF grade of 40 did pretty well, and the people with an EF under 40 don t do as well. What I found most interesting about this study, though, was that if you look at EF, it has a sensitivity of about 56 percent that means it picks up about half of the events and has a positive predictive accuracy of only 9 percent. This has been the problem with identification in the past there wasn t a test with a high enough positive predictive accuracy to be used to identify patients who need an ICD. When you combine ejection fraction with T-Wave Alternans notice that alternans has a positive accuracy by itself of 7 percent with a much higher sensitivity it picks up nearly all of the events of the patients. In this low PPV, it is in large part because the population is very low risk as opposed to my population. When you combine the two and look at EF less than 40 at alternans, you don t lose any of the sensitivity from ejection fraction. You double the positive accuracy. So if you thought you had a patient that was high risk because they had an infarct and an EF less than 40, you now had an alternans test in which you could double the rate at which these patients would need an ICD. This is a meta-analysis that Stefan Hohnloser, myself, and Dr. Richard Cohen have put together of about 132 patients that were all EF less than 30, had ischemic myopathy, and no prior events. Once again, the alternans was positive, abnormal, or determined here clearly than if it was a higher risk group; if it was normal, it was considered a low-risk group. These are the two different endpoints, including sudden cardiac death arrhythmic events as well as resuscitating cardiac arrest and sustained VT. This is just a background to some of the data I want to present from our own study. The preliminary data of this paper was first presented at The American College of Cardiology in 2003. This is a more complete analysis now of the MADIT II subset of that population. We have titled this study T-Wave Alternans Distinguishes Between Patients Likely to Benefit from an ICD from Those who are not: A Solution to the MADIT II Conundrum. We have a number of excellent investigators and coordinators that helped us with this study, and I am grateful for their hard work. I am also grateful to the X form data coordinating staff at Columbia University, including Tom Bigger and Tony Magnano, who is one of our cardiologists in EP. This is supported by a NIH grant, as well by a grant from Doris Duke and Cambridge Heart, Inc. In addition, this was a grant that was balance funded we controlled the data and the protocol, and they gave us the financial support. The background is that MADIT II showed that patients with prior infarction, an ejection fraction less than 30, and who were randomized to ICD had improved survival; the absolute mortality benefit was about 5.6 percent over 20 months of follow-up. The bottom line of that study was that you see a difference in those that have an ICD than those who don t have an ICD. Unfortunately, the difference was smaller than we would have hoped, and the fact was that many patients who got ICDs in MADIT II were destined to go on to use them. This put real pressure on Medicare services/CMS and the group run by Shaun Tunis. Last year they were asked to make a decision: should Medicare reimburse all ICDs in people that meet the MADIT II indication? What they decided was they couldn t come to grips every single patient with an EF less than 30 in ischemic myopathy needing an ICD. Therefore, they tried to hedge, and their hedge was that they wanted people to have a cure restoration that was greater than 120 milliseconds. They asked that people have a cure restoration that was greater than 120 this was based on the subanalysis of MADIT II criteria that suggested that the mortality benefit was greater in the group with less than 120. That has since been debated, but that was the data that was presented, and that was the decision that was made. The answer was that a third of MADIT II patients roughly have a QRS complex greater than 120; CMS said since that is a high-risk marker, and T-Wave Alternans is a high-risk marker, they were going to make it both. If you have both, we are going to put an ICD in. I think that puts commissions in a very tough position, because the MADIT II studies say that if your ejection fraction is less than 30, you ll likely benefit from an ICD; CMS says you also have to have a QRS greater than 120. The real question is, what about those patients that have a QRS less than 120 that are at really low risk? Can you tell those patients to go home, you don t need an ICD? That is what we are trying to address in our study. The objectives of our study were to compare the prognostic significance of alternans and QRS restorations in MADIT II-like patients, and again, this is a subgroup analysis from a large, multicenter perspective outcomes trial that we did. The eligibility criteria for this study overall were pretty broad: age greater than 18, sinus rhythm, EF less than 40, and no prior arrhythmic events. In addition, they could not currently be at high risk we knocked out Class IV heart failure, unstable coronary disease, persistent atrial fibrillation or atrial flutter. That was the criteria for the overall involvement of the study. It is important to recognize that this study was done in a mixture of community practices as well as some academic centers. In fact, two of our three biggest enrollers were community or private practices, so this is a very representative group of patients. In addition, this was a natural history study it did not require any treatment. Therefore, clinicians did not have to make a decision to worry that their patient was going to get an ICD and have to put them into the study. I think one of the issues from MADIT II was that in order to get into the trial, your patient had to be warned that an ICD might be placed, and some physicians would have sent those centers a higher risk patient group. The sole group analysis that I am going to show here is a group that had prior myocardial infarction and an ejection fraction less than 30. Out of all these patients, we knocked out the patients that had non-ischemic cardiomyopathy and an ejection fraction greater than 30 percent. They all gave consent, they all received an alternans test, primarily on the treadmill, although some were on a bicycle. They were all on current therapy that is important from my point of view. If they were taking beta blockers, they were tested on beta blockers. There have been some previous studies with alternans where beta blockers were withheld Dr. Hohnloser s group and my group have always studied patients on beta blockers. There have also been Holter recordings and follow up. The primary endpoint for this analysis is the same endpoint which has been used in MADIT II, which is all-cause mortality. This analysis will show you that we used the automatic computer generated interpretation of alternans from the CH2000 - the HearTwave. The interpretations here were based specifically on what the computer said, and they were not overread. The computer used the rules that were established to interpret the tests, so the tests were either for positive alternans or sustained alternans for at least one minute with the onset heart rate less than 110 and had an alternans voltage greater than 1.9 microvolts. If the patient achieved a heart rate of 105 or greater, with no sustained alternans, they were determined to be negative. We are in the process now of overreading all these tests with expert overreaders to analyze these further. In this analysis, we lumped together the positives and indeterminants, because in the previous study by Dr. Hamas group called The Heart Flutter Study patients who were indeterminant were often because of ectopy or non-sustained alternans. They also were at higher risk. Therefore, in this study, we are going to call positive and indeterminant tests abnormal, and the negative tests normal. We enrolled 590 patients, there were 38 patients with post-enrollment exclusions four didn t have the necessary permission to Class II classifications, so we had 548 eligible patients, of whom 77 met the MADIT II criteria. This is the group that I will talk about. The mean age was 61 years old, 85 percent were male, and the mean ejection fraction was 23 percent. This was a really sick heart failure population. However, despite the ejection fraction of 23 percent, we still had 16 percent of patients who never had clinical heart failure, and 11 percent who were only Class I. Forty-seven percent had Class II heart failure and 26 percent had Class III heart failure. Again, none of these ejection fraction groups were less than 30. About a third had diabetes. MADIT II infarctions were just in pass; the mean time from prior myocardial infarction was about 5 years. These were also all patients with ischemic heart disease. About half had prior CABG, and about half had prior admission to the hospital for heart failure. Seventy-four percent of our patients had used beta blockers, eighty-five percent had ACE inhibitors, and 82 had blockers again, a pretty well-treated group. The mean follow-up was 16 months, and the actuarial two-year mortality was 13 percent. The mortality was a little lower than you might have expected from MADIT II, so let me make one comment about that. The main differences, in my opinion, between our sample and MADIT II is that: 1) MADIT II included patients with atrial fibrillation, and we excluded patients with atrial fibrillation; 2) patients with atrial fibrillation and heart failure have a higher mortality, so that is going to give us a lower mortality without those patients. The thing I said before is that I think there is a selection bias in MADIT II as well as SCD-HeFT and MADIT I, which is that the people are going to get an ICD. Physicians often know who their high-risk patients are, if not by intuition, then by other criteria. Physicians refer their patients to a study with an ICD for those that they think need an ICD. Our study really was a true epidemiologic study we didn t treat patients at all, we just studied and followed them. I think that gave us a chance to enroll a much broader population. In the overall tests, sixty-eight percent were abnormal, and 32 percent were normal with alternans. Therefore, there were no surprises this was a group of patients with significant LV dysfunction, and only about a third of them had a normal alternans test. If you look at QRS restoration, in some of them it is remarkably the exact same split, although 32 percent are abnormal (greater than 120) and 68 percent have a normal QRS (less than or equal to 120). We are picking out twice as many high-risk patients than QRS restoration. Here are the results of the study. If you look at actuarial T-wave mortality, if you have abnormal alternans or QRS, the mortality was pretty high: nineteen percent for alternans and 15 percent for QRS restoration. Abnormal tests both pick up high-risk groups. However, the real difference here is that if you have a normal QRS restoration, the two-year mortality rates were still 12.7 percent, whereas if you have a normal T-wave alternans test, the mortality rate was two percent. This is actuarial two-year mortality. Remember, in the MUSTT study, the patients that were followed in the registry that had nonsustained VT and were not inducible had about a 12 percent two-year event rate. I think the message from that study was if you have a negative EP study and ischemic heart disease, you may not have patients who are really at low risk just because they have a negative EP study. I think you can make the same comment here about QR restoration. A normal QR restoration, at least in our sample, does not connote a low-risk patient group. If you look at the hazard ratio, because of this high event rate in the normal group for QRS, the hazard ratio is 1:5; this is clearly not significant with alternans, because of the low-risk group among the normals the hazard ratio is 7.4. The conference intervals were wide because the samples were all too small, and the event rates were low. However, it was clear that this is a test that not only distinguishes high and low risk, but also picks out a very low risk group (2.1%). If you look at the false negative rate, it is 1.8 percent in alternans and 7.6 percent in QRS restoration. Here is that same data now presented graphically. There are a couple of myocardial curves of mortality plotted over 24 months for alternans and for QR restoration. What you can see here is there is only one event in the alternans group at one year, and there are a number of events in the abnormal group here, so you get a significant mortality of almost 20 percent in two years. For QRS restoration, you see a significant number of events in the normal QRS restoration group as early as 3-4 months and increasing up to two years. To conclude, we believe that we have shown in our study that alternans is a much stronger predictor of mortality than QR restoration in the MADIT II population. Patients with a normal T-wave alternans test have extremely low two-year mortality about two percent. In patients with a QRS less frequent than 120 the so-called low-risk patients they have an unacceptably high two-year mortality rate of 12.7 percent. All patients that are classified as false negative QRS, and all patients who died had a QRS less than or equal to 120, had an abnormal alternans test, so alternans picked up on every one of those patients. Finally, I think alternans identifies a high-risk group that requires therapy to prevent sudden death, and I think in MADIT II-like patients, the number of ICDs implanted for every life saved is significantly improved if you use alternans it is 1 in 18 if you use MADIT II, and 1 in 7 if you use alternans. I think it is important to mention that alternans also identifies a low-risk group; that documented conservatively, we believe that MADIT II-like patients with a normal alternans test can be reassured and safely followed without implantation of an ICD.