[1] Thank you. I hope this talk would be easier than getting up on the stage. My chore today will be to start the discussion with the latest kid in the block when it comes to the world of pacing and defibrillators and that’s cardiac resynchronization therapy. I’m gonna talk about patient selection as well as optimization of therapy. [2] Why is this an important problem and something we’re concerned about? I think we’ve all been quoting over and over again from studies that there’s about a 30% nonresponder rate when we implant biventricular devices with CRT and in fact, this 30% responder rate is probably an underestimation because in placebo controlled studies, we see the placebo response rate of pacemaker is significant in the order of 20% or so. So possibly, up to 50% of patients may not be responding any better than a placebo patient for this therapy. I think another important point which I think has been very underestimated is the problem of hemodynamic function which may be worsened by CRT. Something that very few people have been thinking about that we’ve may actually be hurting some patients with this therapy. I should start out by saying that I am very, very enthusiastic about CRT but I am a little worried about how we may have gone overboard. [3] And it reminds me, not too many years ago, when we were having debates at this meeting and others about right ventricular pacing and the idea that if we maintained AV synchrony, there is no way we could possibly hurt a patient. These are the data from David and now people think that it is heresy to be pacing the right ventricle, it’s malpractice, yet this was our standard of treatment just several years ago. So we now well-accept right ventricular pacing as being deleterious. I think there are some situations where we do need to worry about CRT. [4] And this is an example. This is a patient. This is an acute hemodynamic study in a patient looking at the PVT or contractility in this patient and what’s being shown here is the acute response as a function of AV delay from this patient pacing in two sides. First, the lead was placed on the free wall which is a preferred site. We could see that on top that in fact with lateral wall pacing, there is a small but significant hemodynamic improvement. But in fact if this lead is placed in the anterior vein or in the septum. In fact, we get a very large negative hemodynamic response with CRT pacing. This is as large a response as you’ll ever see with right ventricular pacing. So clearly, a lead in a bad position is a bad thing. [5] There are a number of factors which affect our patient selection beyond which simply where we’re placing the lead, the severity of heart failure, the QRS duration and morphology, the etiology of heart failure whether it’s ischemic or nonischemic, as well as the degree of baseline asynchrony. [6] If we look at the overall heart failure population, a vast majority of patients are not traditional CRT eligible patients. In fact Class III patients which are the primary group in whom benefit has been shown before represent only about 25% of all heart failure patients. Another five percent are Class IV and I think there are certainly some discussions as to what the clinical studies have shown in Class IV patients. But there no data yet showing long-term clinical benefits of Class I and Class II patients. [7] If in fact we look at the data for implanting less severe heart failure patients, we do have some data available. This is data from the CONTAK CD study, the first study that tried to push the envelope a little bit and allow for less severe heart failure patients. And you can see here that if you take the advanced heart failure patients with wider QRS in Class III to IV heart failure, a marked improvement in quality of life, whereas if we take our other patients with less severe heart failure, as shown there on your left, you’ll see no change at all in their quality of life. [8] Similarly, if we look at peak oxygen consumption, we see a very similar phenomenon. A much greater response in those patients with severe heart failure compared to the group with less severe heart failure. [9] If we summarize from all the clinical studies which have enrolled patients with Class I and II heart failure, these are all patients with wide left bundle branch blocks. So it’s not narrowed QRS patients but wide left bundle branch blocks. We could see in the Class III and IV heart failure patients, six-minute walk is improved, peak oxygen consumption is improved, fairly consistent results. However, if we look at the Class I and II patients, in fact we’ve yet to show any benefit in that group of patients from small sub-study analyses from these clinical trials. [10] If in fact we look at simply reverse remodeling sort of endpoints. We can in fact see that Class I and II patients do appear to show at least an improvement in ejection fraction comparable to other groups of patients. There are a couple of large studies going on, REVERSE and MATID-CRT, trying to assess the benefit of CRT in these less severely ill patients. But, at this point, I think clearly the response rate has been smaller. We do not yet know whether these are patients who will benefit. [11] The hemodynamic responses, the patients in acute hemodynamic responses could be classified into two types per shown by the PATH-CHF investigators. Shown on top is a Type 1 response, where they have again this very profound positive hemodynamic response to CRT either looking at dP/dt as a measure of contractility or pulse pressure as a surrogate of stroke volume and it is a relatively broad concave relationship over a wide-range of AV delays which is what is shown on the abscissa there. On the bottom panels, are shown the so-called type 2 patients. These are patients in whom we only see any positive hemodynamic response of long AV delays if at all. In fact, here’s a patient in whom they have a negative hemodynamic response again, both pulse pressure and dP/dt. There are not a whole lot of reasons why we think this patient may get better if we’re making them worse with pacing hemodynamically. And if we try to differentiate these patients at least the PATH-CHF investigators show that these were the patients with the relatively shorter QRS’s, 120 to 150 milliseconds, again not the 190 milliseconds we’ve been trying to push through now but the relatively short but still wide eligible patients. [12] And if you look at the relationship between mechanical response and baseline QRS, we can see that the wider the QRS, in general, a lot of noise in the data, the better the hemodynamic response. So while QRS is not a perfect marker by any stretch of the imagination, certainly as we get narrower and narrower, the response rates get smaller and we start to see adverse response. [13] If we take a large group of patients and start to look at using QRS threshold, we can see in this somewhat complicated slide here that when we get out the very large QRS’s, if we look at the far right at 200 milliseconds or so, we can see that there’s about 100% response rate. However, if we restrict ourselves to QRS’s of up to 200 milliseconds, we would only benefit about 5% of patients who may potentially benefit from CRT. If we go to the other extreme of, let’s say 125 milliseconds, we will have gotten virtually every patient who may benefit from CRT, will get that benefit. But the response rate at about 120 milliseconds is only 20% or 80% of patients are either nonresponders or adverse responders. So we sort of once again come to the same sort of middle ground that somewhere around 150 milliseconds which is much wider than most people usually use as their cut-off is where we can best optimize choosing patients for this therapy. [14] Another group of patients I think, which have often been implanted with very little data to support it, are those patients with right bundle branch block. At least the guidelines and reimbursement criteria say you have to have a wide QRS. It doesn’t say what it has to look like. And here is some data on right bundle branch block patients showing that in fact they have much less interventricular delay, a marker of patients who can benefit from CRT. [15] And in fact the some data which we are presented at this meeting last year, showing that if we look at the left ventricular dP/dt with right bundle branch block patients shown on the left, you see very little hemodynamic response no matter how we paced the BiV, RV, LV only, no matter what, it doesn’t make much of a difference. In fact on the right-hand side, once you do by pacing right bundle branch block patients is to improve right ventricular function. That’s were the delay is in the right ventricle. For patients with right heart failure, this may be a useful technique but it doesn’t appear to be very useful from proving patients with left heart failure. [16] One simple way of trying to measure hemodynamic response is to measure the distance from the onset of a QRS to the local electrogram of her LV potential. If we’re gonna create electrical re-synchrony of patients who were desynchronous, the farther away the LV electrogram is in time from the beginning of the QRS, when we pace it, we will bring those together. And we could see a nice linear relationship here between those two phenomenon. And in fact our nonresponders in our study here turned out to be those patients with the shortest Q-LV times. So if we have our lead in the septum, that’s where normal activation. For a start, it’s gonna be a very short Q-LV time, that’s why we don’t get much of a response there. But what this also implies, if we take patients once again with truly narrow QRS’s, if the QRS is only 100 milliseconds, it’s gonna be pretty hard to get a Q-LV time any longer than 100 milliseconds to set the end of your QRS. You’re gonna fall into this nonresponder range where I personally think, I am somewhat skeptical of our ability to improve patients with truly narrow QRS’s based on these kind of data that may have mechanical desynchrony but they don’t have electrical desynchrony. [17] Well we certainly know that the QRS is far from perfect. Patients with wide 150 milliseconds or greater do better than patients with relatively less wide QRS’s. But there is a lot of scatter in the data as shown in this curve. [18] Therefore, there have been other methods being used and other ways of trying to identify these groups of patients and there is now a whole cottage industry of techniques and various echo software that have been developed to try to look for desynchrony. And clearly we know that patients who have mechanical asynchrony with desynchrony tend to respond better to biventricular pacing. [19] Some of the early data were shown here, from Yu and his colleagues in Hong Kong, looking at pressure waves, pressure curves and we can see if we compare RV and LV pressures, that when they are out of sync, as shown on the left-hand panel there when you create that RV-LV pressure curve loop, you see there’s a lot of area under there. When it’s synchronous, as shown on the right, the lines virtually overlap and there’s very little area under that curve. [20] They’re able to show quite nicely that the responders tend to be those patients who had desynchronous pressure changes between the right and left ventricle as one way but a rather invasive way of trying to measure or predict the synchrony. [21] Other more elegant ways include MRI, in which looking at MRI, we could very accurately see wall motion abnormality. [22] And in fact we can see that the asynchrony that we pick up by MRI technique correlates nicely again with the hemodynamic response that we see in this group of patients. [23] But really I think MRI and invasive pressure monitoring is not gonna be a standard way that anyone does this but maybe a very nice research method. But it’s really the echo technology which is taking over. And there are a number of echo techniques, essentially there seems to be as many echo techniques that have been described as there are investigators looking at this problem. I certainly am not an expert enough to tell you which one is the best one. But if we look at, at least one that is easy for me to understand which is looking at the simple measurement, an M-mode from the septal to posterior wall motion delay as shown here. [24] We can see that when there is a delay between septum and wall motion here being very large with a 330 milliseconds that tend to be a responder and when we see when we have septal wall motion shows very little delay against synchronous contraction, we tend not to get responders in that group of patients. [25] And in fact if we look at reverse remodeling or the improvement here in end systolic volumes, we see that it is inversely correlated with septal wall motion delay. The larger the delay, the more the ventricle is gonna shrink with pacing, the greater the response you’re gonna get because there is more desynchrony. [26] A much more sophisticated method has been developed by Yu, looking at 12 different segments rather than just the septum in the free wall, more complicated, using tissue Doppler to do that. [27] And once again, without going in any detail because of time and because of my ignorance of this technique, we can see quite clearly that this is a nice separation using this tissue Doppler technique between the responders and the nonresponders with a very high sensitivity and specificity at least in their hands. This is a difficult technique and others have had trouble reproducing this just because it’s challenging technically to do. [28] I think the third technique, I’ll just briefly mention, is that by Sogaard who has looked at delayed longitudinal measurements, again another measure of asynchrony using spectral methods of tissue Doppler echo. [29] And in fact once again, we see a very nice correlation both in terms of change in ejection fraction and systolic performance which correlates well with their echo measures of desynchrony. [30] I wanna switch gears with just a couple of minutes here to talk about the role of CRT programing. I think clearly in patients with left bundle branch block and severe CHF, this is very, very good therapy. I’ve already talked about patient selection and lead location but I think that we need to be programming the patients appropriately if we’re gonna maximize the hemodynamic response. Once we selected a patient with a potential to develop what we don’t want to do is inappropriately program them and lose that benefit. [31] Again if we use acute hemodynamic measures, this is just an example of some of our hemodynamic studies in which we put a catheter into the left ventricle, a lower catheter with a transducer. And in fact can measure hemodynamic response acutely over a number of AV delays and we can calculate what the maximal achievable AV delays. So in this patient, 15.7% improvement was the greatest we could achieve in this patient. [32] One thing that became clear, quite obvious to us early was that atrial pacing has profound effects on the hemodynamic response. We usually don’t think about atrial pacing thus all of our clinical studies were in fact in an atrial-sensed mode. But in fact you see two things here, when we paced the atrium shown in purple layer, we see a much larger hemodynamic response. So this is still biventricular pacing but with atrial pacing. The other thing we see is that the optimal AV delay has a marked rightward shift, so much longer AV delays. [33] If we look at a series of almost 50 patients here, again we see this response. This was known in physiology as the treppe effect, that the faster you pace the heart, the greater the contractility of the heart. So overdrive pacing or faster pacing which we used to do for years thinking that stroke volume was fixed and therefore, we get a greater cardiac output in faster rates. Well in fact, stroke volume is not fixed. It’s improving as we increase heart rate in these patients. [34] But to improve we have to allow adequate filling of left atrium to left ventricle. And in fact, what’s shown here is this marked shift in the optimal AV delay in the atrial-paced mode compared to the atrial-sensed mode. [35] For those of you familiar with programming pacemakers, out of the box, our offset from sensing to pacing is either zero or at most, 30 milliseconds in devices. But in fact in heart failure patients who are receiving CRT, the optimal offset should be in the order of 70 milliseconds. I don’t think very many people are putting in that 70 millisecond number but it’s what we will give as our optimal AV delay. [36] And the reason why that’s important is shown here. If we take this group of patients and in fact give them this 70 millisecond AV delay, we get a 60% improvement in dP/dt with atrial pacing. That’s as much as we get with CRT alone. We added the effect of atrial pacing. Yet if we just leave them at a nominal 100 milliseconds, in fact there’s no improvement at all and no change with atrial pacing which were so far down the curve, optimization curve for atrial pacing and then we’ve lost any of the benefit of atrial pacing in that group of patients. [37] One simple method that we’ve been working on to try to optimize this is rather than trying to use echo techniques simply looking at electrogram measurements between right atrium, right ventricle, and left ventricle so that we can synchronize contraction, so that when we’re getting normal activation down the right bundle we can pace the left ventricle. [38] And in fact, what we find is that intrinsic PR interval is not very useful. What we really need is the intrinsic AV delay directly measured from our electrograms which gives us a much more accurate way of being able to do this. [39] And in fact when we do this, we get an almost too good to be true correlation between optimal function and what’s predicted here. So this is a group of patients, I believe there are 41 patients in this series in whom we directly measure with an invasive catheter what their maximal dP/dt rise was and then we set the AV delay to in fact what was predicted just looking at atrial and ventricular electrogram. And in fact, we get a very, very good correlation. If we could set AV delays in both sensed and paced mode with nothing more than some simple measurements of intracardiac electrocardiograms. [40] And in fact we compared this now with a variety of different AV optimization techniques either empirically just setting patients at a variety of AV delays or using the most common method, the Ritter method for ,mitral in-flow or the aortic VTI method. And we can see here that the number is lowest with our expertise which is our electrogram method, expertise for heart failure, the lower number means exposure to the optimal AV delay. So the best technique is a simple one that can be done in a minute in the lab compared to doing all this echo optimization afterwards. [41] So just in summary, who benefits most from CRT? I think certainly patients with New York Class III and IV heart failure tend to show the greatest therapeutic benefit that’s surely greater than those with Class I and II, remains to be determined the benefit in the I and II patients. Patients with left bundle branch block respond consistently, whereas those with right bundle branch block do not. So clearly the left bundle, wide left bundle Class III patients are wonderful patients for CRT therapy. I didn’t show you data but certainly it’s fairly consistent that patients with dilated cardiomyopathy tend to have a somewhat larger response than those patients with ischemic cardiomyopathy. They don’t have dead walls and we don’t have to worry about it. About 80% of patients with QRS greater that 150 milliseconds improve hemodynamically with CRT. Improvement in asynchrony seems to be the determinant of the benefit noted with CRT and this may be independent of QRS width. But I think it’s becoming quite clear, well more and more clear that we not only need desynchrony as shown mechanically but we also need electrical desynchrony. If there isn’t enough delay electrically we can’t do anything with the pacemaker. And optimal programming is necessary to achieve the maximal hemodynamic benefit of pacing. Thank you. [42] [42]

Michael Gold 151AB Cardiac Resynchronization Therapy: Patient Selection and Optimization Selecting the Right Device for the Right Patient: What Type of ICD to Implant