[1] Thank you. I am going to discuss patient selection and I hope that I am not going to upset too many people in the audience, but I think I am pretty well going to guarantee to upset at least one. [2] We have got a number of ways that we can think about patient selection. We can use some logic. Maybe carefully logical that some people, let us say if they have severe dementia, you just look at the whole patient, they are obviously not a candidate for device therapy, so that is pretty logical. Somebody whose ventricle is not that sick, it is logical that we should not be using cardiac resynchronization, so we can use logic. We can also use experience, assuming that you are actually doing a reasonably large volume of devices, so experience will begin to tell you where devices work and where they do not. Well, fortunately, as a professional, we are very prone to developing myths and then we manage patients according to the mythology that we create and I am very concerned that we are creating a wrath of mythology around cardiac resynchronization and we need to sort out the facts from the myth and some of what we have been doing this afternoon is to try and point out we are in danger of creating myths. They might be facts, but we really need some harder evidence and ultimately, there is experiment and experiment is how we make sure that the mythology we create is true or not. [3] Well, if we look at how we evaluate the evidence, we can look at the clinical trial process and we can look at the outcomes. We can do observational trials and we see at this meeting and many others, a large number of observational trials and there are some severe flaws in interpretation of these observational trials and they are in danger of treating a mythology around this subject, which may or may not be true and we really need to sort this out. There are the randomised controlled trials, the much more robust, but there is far fewer of them. They do have some of their own problems. We have single-center studies and multi-center trials and there are lots of problems with single-center trials and why they are good hypothesis generating, but they do not constitute sort of evidence that we should base our clinical practice on. We have also got outcomes, so we seem to be quite keen on measuring response in terms of cardiac function and jumping to the conclusion that improvement in cardiac function with results and benefits for patients. We have been through this again and again in the pharmacological field and in fact ventricular function is not such a great surrogate for clinical benefit from a patient’s perspective, and we should be quite cautious about assuming that improvements of cardiac function will translate into clinical benefits. [4] Just to look at the single-center observational clinical trial data. If you do not have a control group, you really do not know whether it is the intervention or the disease that is determining the outcome in the population. And I will show you some quite startling evidence in just a moment that perhaps will worry a few people in the audience. Observational trials are really not a good way of finding out if a treatment works. They can be used to identify what I call met and unmet needs, but if you put a device into a patient and they do well, then perhaps you do not need to do anything else. It does not of course tell you that the patient needed the device in the first place. The patient could have done perfectly well without it. If you put a device in a patient and they do badly, then you know that the patient needs something more than the device that you gave them. On the other hand, you do not know that they would have done even worse if you had not put the device in. So, there are severe limitations to observational data. For single centers, we got no protection from what I call a subgroup effect. This means that as an individual investigator, you are perhaps seeing a very selected group of patients, you have an idea in your head which is not quite correctly expressed on paper, and you deliver up a group of patients and they look different to the reader from what they are in your head, so we cannot really be sure that the single-center patient is the same patient as in our clinical practice. We do not know how many analysis were done very often, there are not usually good statistical analysis plans for single-center trials, and you do not know which patients were excluded and it is often three years of analysis with 500 P values before you get one and then you report it. Of course that P value has no meaning, so we have to be very careful about that. I think single-centered studies require replication in many centers before you can really have a great deal of faith in them, and of course hard outcomes are very rarely reported and when they are, there is a lack of statistical power, so we should be very cautious about interpretation of these studies. [5] This slide is going to upset some people I think. This is data from CARE-HF, so there are over 800 patients on the study and there are over 300 endpoints shown on this study. What I want you to draw your attention to is the blue line. Now the blue line is people with little in the way of cardiac desynchrony, who did not get CRT. And you can see it is the people who do not have desynchrony, who do worse in the control group. If you have desynchrony and you do not implant a CRT device, the patient does better not due to the device. Patients with cardiac desynchrony have an intrinsically better prognosis if you match ejection fraction for ejection fraction and age for age unless we get radical and obviously we are getting different information from the cardiac desynchrony data on imaging than we are from QRS. And notice that although the benefits of cardiac resynchronization are somewhat greater in patients who had more desynchrony at baseline, the difference between yellow and green is the benefits of cardiac resynchronization in those patients who had little evidence of cardiac desynchrony baseline in CARE-HF and perhaps the problem is with the name CARE-HF, it means cardiac resynchronization and heart failure study, but perhaps we jump to conclusions and perhaps we should be talking about atrio-biventricular pacing because is this therapy really about resynchronization or have we created a myth? [6] To turn to the surrogate endpoint, but I think I have pretty well gone over this. Are we sure that cardiac function is going to predict the effects on symptoms and prognosis? [7] And we can go back to the CARE-HF study. I will just concentrate on the right hand side of the slide here. We have the control group and we have the non-ischemics and the ischemics, and you can see that ventricular function did not change much in the control group. A big improvement in ventricular function in the non-ischemics and much less improvement in the ischemics, and that is what we all know and CARE-HF confirms your suspicions on data from observational studies. [8] But anyway look to clinical outcomes. This is all-cause mortality now in the three-year followup extension study, 40% reduction in all-cause mortality with cardiac resynchronization, but the benefit is greatest in the ischemic group. The group who got the least improvement in ventricular function. So, there is something else going on with cardiac resynchronization that is not predicted by the improvement in cardiac function. And you can see down here, we got interventricular mechanical delay somewhere down here. If I can see it from here, lost it for a moment, here it is. This is the people with more or less interventricular mechanical delay. You can see what a poor predictor interventricular mechanical delay was of prognosis. The one thing that did sort of stick out and it is not statistically significant, and of course there are lots of subgroups here, so something is going to stick out. The people without mitral regurgitation did not seem to benefit quite so much from cardiac resynchronization as the others. [9] So there are some problems with the randomised control trials as well. The entry criteria designed to give the study the best chance of success. We design trials to succeed. We do not design trials to fail. They often do fail and that just shows our inadequacy of the tools that we use. Of course any clinical trial will include groups of non-responders and exclude groups of responders and give the illusion of an average effect, but of course the average effect is made up of lots of groups of patients responding in different ways. Very importantly, we design entry criteria for clinical trials but then investigators, especially when those entry criteria are subjective, do not necessarily do what we ask. So do not believe that every patient recruited into a trial actually fulfills the entry criteria of clinical trial. And that is important because clinical trial entry criteria should be used for general guidance and principle. They are actually not very good for making rules. [10] Here, the entry criteria for CARE-HF. This is the largest clearest result from cardiac resynchronization. So these patients have to have heart failure for at least six weeks and require loop diuretics. They were supposed to be in class III-IV heart failure and they were supposed to have a high standard of pharmacological therapy, systolic dysfunction and dilatation. They were suppose to have a QRS greater than 120 milliseconds and if they had a QRS between 120 and 149, then they did have to have some desynchrony criteria, so some are placed in desynchrony in the CARE-HF study, and we had to exclude patients with atrial fibrillation or those requiring pacing for other reasons. Now when it came to objective criteria like QRS and ejection fraction, the CARE-HF patients pretty well follow the rules, but then we got some subjective criteria in here and that makes life a bit more difficult and interesting. [11] Here are the baseline characteristics. The first thing, I do not know about your clinical practice, but in my clinical practice, if you have got class III-IV heart failure, usually on 80 mg or 120 mg of furosemide. We just do not think you are trying hard enough if you will only give them 40 mg of furosemide and the patient is still in class III, but you can see that less than half of the patients in CARE-HF were taking more than 40 mg of furosemide per day. So they were not on high doses of diuretics, these were not really end-stage heart failure patients. [12] This is looking at the investigator-assigned NYHA class. So 763 were assigned to class III and 50 to class IV, but then we asked the patient a series of questions which allowed us to allocate the NYHA class according to the patient as opposed to the investigator. And you can see that about a quarter of the patients said that they were asymptomatic or had very little in the way of symptoms. [13] So these patients were all class III-IV patients. They did have quite bad cardiac function. If you look at the ejection fraction, the average is 25%. These are the interquartile ranges, you can see that 75% of these patients have an ejection fraction under 30%. So they undoubtedly have severe underlying disease, but not necessarily severe symptoms. [14] Okay, so let us go back to these trials and look at the evidence that cardiac desynchrony predicts the outcome in randomized controlled trials. And here we have the CONTAK CD study with Marcus et. al., and the septal posterior wall mechanical delay had been shown in observational trials to predict the benefit of cardiac resynchronization. So they are trying to replicate this in CONTAK CD and if you read the report it says the feasibility and reproducibility of performing this measurement were poor. They had a huge interobserver variability. And in fact, they could find no evidence that this measure of dyssynchrony predicted outcome in CONTAK CD. [15] If you look at miracle, we have only got a report on the effects on ventricular function. I have told you that that may be flawed. It does appear that interventricular mechanical delay is the best predictor of improvement in left ventricular function in a randomized controlled trial with some evidence that aortic pre-ejection delay might also help. [16] But I will take you back to the CARE-HF study and just remind you that interventricular mechanical delay are indeed the other standard echocardiographic measures of desynchrony, did not predict prognosis and were not great predictors of outcome. And I can also see in a sub-study with tissue Doppler imaging, that it is statistically there, but it is not great. [17] To move on, perhaps we need to go through more advanced technology and here is cardiac magnetic resonance imaging and if you use delayed enhancement gadolinium, you can look at the extent of myocardial scar in the wall. And if you have got lots of posterolateral scar, then logically, you would think that this group of patients could not respond to cardiac resynchronization because there is nothing to pace out here. Well, we do not actually know if that is true but it sounds good. [18] We can use CMR just like echocardiography to do wall motion studies until we get desynchrony index. [19] And when you do that, it is quite interesting because you get a very accurate way of assessing cardiac volumes and function. And when you do that, you just find that desynchrony is a very good way, well there is a high correlation between ejection fraction and cardiac volumes and dyssynchrony on CMR, so you might as well just measure the ejection fraction and be done with it. [20] So in terms of criteria for implantation, I think that we can divide them into these, at least four groups. We have got fairly secure ones, we have got rather insecure ones, we have got a group that is too early to say, and we have got a group of new markers. I think left ventricular systolic dysfunction is fairly secure for the moment. There are a group of people out there that think that diastolic heart failure might benefit from resynchronization and I would challenge them to make a confident diagnosis of diastolic heart failure. I do not think we can do it at the moment. Blood pressure and mitral regurgitation are not decision makers, but if you have got a patient with a low blood pressure and moderate mitral regurgitation, these are patients who, in my experience, do better and indeed you can show from CARE-HF study, these are soft indicators of better outcome. I think it is too early to say about QRS width. I wonder whether QRS 120 is just another modern myth of cardiology. We invented 120 and then having invented it, went out of our way to prove it works, but we have not really bothered to look at less than 120 and CRT may work equally well in people less than 120 as above 120. We have no evidence to say otherwise. I think that NYHA class is another way of having, sure if you want to improve symptoms, the patient has to be symptomatic, but if you want to improve the disease and patient outcome, I think the symptomatic state is irrelevant. Sinus rhythm, I think there probably might be some added advantage to being in sinus rhythm for this technology, but I think that atrial fibrillation patients benefit as well. Echo desynchrony, I just wonder whether it is going to survive into the next decade as a tool to select people for cardiac care or atriobiventricular pacing. And in terms of optimal pharmacological therapy, I am distressed by the number of people who die having their therapy optimized before they get a device, and perhaps we should be junking this as an entry criteria for device therapy. There are some new markers, BNP does not predict who is going to benefit from resynchronization, that is pretty clear, but what it does do is identify that you are an at-risk patient, that something more needs to be done for your therapy. So by identifying risk or at least identifying low risk patients, you can say, well this patient is low-risk, he is not going to benefit from the treatment, do not give it to him, and myocardial scar perhaps by CMR. [21] So if I can conclude with the CARE-HF entry criteria, how would I change them because they are the basis for my clinical practice and the basis upon which I choose to refer patients for implantation of device. I am beginning to abandon the class III-IV concept and if I see patients with other features indicating that they need cardiac resynchronization but they are not symptomatically severe, I am strongly tempted to go ahead. I no longer wait for optimization of pharmacological therapy in those patients because of my experience and what I believe to be desynchrony does not go away with the pharmacological therapy. So I do not wait until the end-stage before going for the device. I think that systolic dysfunction and dilatation is still in there. We have no data in people with less severe QRS, so I do follow this CARE-HF entry criteria as my measure of desynchrony. I have dropped atrial fibrillation and I put in elevated BNP, so if somebody has got a low BNP, I am very reluctant to go for a device because they have a good outcome anyway. More and more do CMR as a routine in our patients and if they got extensive posterolateral scar, I am reluctant to put a device into those patients. And I guess tying up the pharmacological therapy, the one thing that I do look at is systolic blood pressure. If the patient’s got a well preserved systolic blood pressure, they actually do not gain so much advantage from cardiac resynchronization and perhaps as an indication that they have not had enough pharmacological therapy. Thank you. Okay, I’ve got to do my disclosures on the... CARE-HF study which was sponsored by Medtronic and I have received research grant funding and speakers on the radio. [22] skipped slide [23] Skipped Slide. [24] Skipped Slide. [25] Skipped Slide. [26] Questions and answers. [27]

John G. F. Cleland 253ABC Patient Selection Success with Failure: CRT from Implant to Optimization