Ladies and gentlemen, it's my pleasure to speak about global contact mapping for atrial fibrillation. My disclosures are on the bottom of this slide. Mapping for atrial fibrillation really in many ways parallels mapping in general. At first we started off with a general concept of where things may or may not be with a lot of unknowns, then we moved forward through a general sense of where things were but without precision to more recent mapping approaches which aim at least to identify specific targets to guide ablation to improve outcomes. We're not quite there yet, but global contact mapping definitely has a part to play in that picture. I'm going to divide this talk into four different sections. The first, clinical evidence, the second, rationale and motivation, the third, and the fourth are my ideas of future improvements for global contact mapping to find true from false positive sites, to improve sensitivity and specificity, and to understand things such as why driver sites are transient and seem to fluctuate. The first important point is what's the clinical data? The first multi-center randomized trial of driver-based ablation was presented at Heart Rhythm Society as a late breaker last year, 2019, by Hannes Brackman. The reaffirmed trial randomized 350 patients at 18 centers to either PVI or PVI plus firm. The intention to treat results are shown and as you can see there was no adverse safety signal but no real difference between the groups. What was not so well appreciated is that in fact in both limbs nearly half of patients received additional off-protocol ablation. This resulted in four treatment groups, not two, and reduced power unfortunately for the trial to separate these groups. Now if you look at the on-treatment limbs, and this is data from Johannes Brackman and David Wilber, then in fact the pre-specified groups, which were PVI plus firm, achieved a 77.7% one-year freedom from AF and AT in persistent and long-standing persistent patients, versus PVI alone that was 65.5, which trended to be superior, but obviously the trial was diluted by the four subgroups. A second important point, hypothesis testing only at this stage, is that in the limb which received a lot of additional ablation, including CAFE and Lyme's, the success rates were actually lower, raising the question that was shown in STAR-AF2 and in BOCA and in CHASE-AF among other trials, that too much ablation could actually worsen outcome. Now interestingly, this trend for improved outcome is different to previous adjuvant ablation approaches. In this recent meta-analysis and systematic review by Clarnett and Pratch-Sanders, every other strategy trended lower than PVI alone, and if you look at the recent results of posterior box wall isolation by Lee and Kim from South Korea, you can see that even there, there was no improvement or trend for higher results. Is it true that AF mapping can improve these results? Well, of course, we don't know yet, but if you look at the number of approaches, there's certainly interest in the area. All approaches are now looking broadly and globally, because the old approach of looking at very small, fine areas would not capture spatial differences and would miss parts of the atrium that could be important. The contact mapping methods, there are at least six, I've shown a few of them here, are in the right of the slide and non-contact in the bottom left, which I won't cover because my colleagues here will be doing that. Interestingly, despite very different algorithms, mapping catheters, filtering approaches, and so on, each of these methods shows rather similar results. They show that there are driver regions in almost all of these methods, that there are two to three or maybe four of these in the left atrium and a couple in the right, and that ablating these areas might improve success, so localized regions of interest. The biggest difference between them might be in the temporal fluctuations between sites we'll come to at the end of my talk, and this could be technical in some cases and also potentially semantic, because if a site fluctuates but comes back to the same site, is that different from a spatially stable site that shows temporal variation? It's unclear. If we look at the results of smaller multi-center non-randomized or single-center non-randomized trials, you can see that overall the results were fairly promising across many of these methods with success rates of 60 to 70% in long-standing persistent and persistent a-fibrillation. Now, this is not true at all centers. We're all well aware that there was a great deal of variability with some very poor results at some centers, and clearly any future improvements must try and address this variability. Why map drivers at all? Well, to me the biggest rationale driver has been that human atrial fibrillation optical mapping shows drivers. Here you can see work by Vadim Fedorov and Brian Hansen from Ohio State University showing in the atria of patients explanted, these patients often had atrial structural remodeling and other cardiac comorbidities. You can see with the addition of a voltage-sensitive dye, optical mapping here of the right atrium shows a counterclockwise driver here in the superior and here counterclockwise in the inferior right atrium. Several papers from this group also showed left atrial sites. Interestingly, this group actually placed contact baskets, commercially available ones on the atria, then used in this panel commercially available clinical software to actually analyze these maps, blinded to optical maps, and as you can see showed quite close correlation of both of these driver sites, and overall had an 80% correlation. To me that's very important. Even if the AF in these atria is not completely identical to what you see in vivo in patients, it does show that what the features you would map optically from a dye looking at action potentials is faithfully reproduced by some clinical mapping systems. Many of those sites actually in the human atria were anchored at microfibrosis, and so could fibrosis be a viable mapping target? One of the methods of doing this, of course, is by a gadolinium-enhanced MRI. This is a figure from the DCAF1 trial, Naseem Maroosh et al. The DCAF2 trial targets these sites, which are shown in the lighter colors, which indicate structural abnormality for ablation, and the results of that are pending. A more accessible approach might be to map electrical surrogates as markers of structural abnormality. Of course, we all are familiar with voltage thresholds of 0.5 millivolts or 0.35 in recent studies in AF and lower. How reliable are they? This was an interesting study by Wong and Kullman published in JAKI-P at the end of last year, and they actually showed that there could be a great deal of variability based on rate and direction. Here you can see in one patient, column one, when the atria was paced from the coronary sinus at 600 milliseconds, there was tiny amounts of low amplitude, so red low voltage here near the left inferior pulmonary vein. When you then paced at a shorter cycle length, this expanded, and when you paced from the left superior pulmonary vein, it occupied most of the posterior wall. Now, obviously, the structural substrate is not affected over the course of minutes by pacing, so clearly we would have to take these factors into account, and this would pose challenges to reliably interpret these substrates by electrical indices, in my opinion. But clearly, work is being done in this area. Another area that's interesting to us has been to try and reconcile differences between the epi and endocardium. We know they're different from the Fedorov work and from work by Mazzalessi looking at longitudinal dissociation in the atrium. We've been looking for some years now, all of our patients doing body surface mapping. This is a nice case where we can see a counterclockwise rotational element in fibrillation on the body surface in left and intracardiac on the right. Clearly you can see that, but you can also see the quite difference in spatial resolution. The one on the left is much less grained, it's much coarser, showing activation that really encompasses the whole of the body surface, whereas, of course, in intracardiac, you can see competing wavelets and breakthrough. Similarly, when we looked at this consistently, and this was done, published just recently by Miguel Rodrigo in Cirque EP a couple of weeks ago, you can see that while there is a correlation between intracardiac and body surface, there is a great deal more localized activity on the body surface than intracardiac, and the spatial resolution and meander may be greater. I've summarized this a bit. The spatial resolution VCGI maybe is 10 to 15 millimeters from some studies, but nevertheless, there is a correlation. Now moving on to ways to improve this in the future, because we clearly have to improve the technology across the board and improve mechanistic understanding for any of these concepts to be accepted into clinical practice. How do we find dominant or true positive or relevant sites from those which are not? This is work that George Leaf presented at APHRS in 2018, and won first prize at the Young Investigator competition, and basically, there are three different approaches you could use empirically. You could look for fast sites. This is a dominant frequency map. The higher the frequency, the faster. Unfortunately, in clinical studies so far, this has not really been robust, but clearly, theoretically should work to be determined. The next approach would be to look at temporal consistency. Here we have a couple of sites represented by how much they fluctuate over time. You can see, for instance, there's a blue site, which comes and goes, and a red site, which comes and goes. You could look for how long each of them is here, and target the one that's there for the longest, so that's also potentially viable. An area we've been looking at a lot is a slightly different approach, which is, how much of the atrium does a site control? In the left panel here, you can see rotation activations clockwise, where my pointer is, towards the bottom center, about six o'clock. There's also a site that you'll see periodically at about 12 o'clock. What we did here was to develop new software, which looks at propagation vectors dynamically, and then shows them in a persistent look-back fashion, so that we can identify how much of the atrial surface is linked to that site. You can see there's two areas here, and in fact, the top area, which occupies a very small area of control, was not a site of termination. The one at the bottom that occupied larger areas was. A book now by Neil Bartier and A.J. Rogers looked at this, and this was just published, as you can see. The reference is here. Here's a case where we had two sites. We ablated this one, shown in white. You can see it's an X, now in panel C, at which point the area that was initially small has expanded. The area of control has grown, because the competing site is no longer present. When we looked at the electrograms, raw electrograms, in this site, you can see that they now show spatial linking. It's almost like a mini atrial tachycardia within a sea of fibrillation. Because this was now big, ablating the site terminated AF to sinus rhythm. This was the site here in the posterior wall, outside the PVs. Here's a converse site, where you can see that there was a smallish area within complexity. Ablating that, the X now shows two more. They didn't really get bigger. Ablating one of those left another one that didn't get bigger. When they were all gone, there really wasn't another large coherent site. This patient needed cardioversion. If you now look at our series, in the recent publication, patients who terminated had much larger areas of control than those that didn't. If you look in the right column, that lasted for longer periods of time. This will be the examples that I just showed you. The final point I want to look at, which I think plays into dominant versus non-dominant, is why do drivers fluctuate? Of course, nobody yet has a completely definitive answer. One possibility is competition from competing sites. This is work by Chris Kowalewski, who's now training back in Germany with Gert Hendricks. You can see here that this is a patient with persistent AF, in whom ablation at site A terminated persistent AF prior to PBI. You can see a basket in place. AF terminated to sinus. Emitting the basket, we could find at site A a rotational site. You can see here clockwise in panels A, B, C, D around here. Because we had a global map, we could look simultaneously in all sites. We also saw a focal area that's emanating here. You can see this progression, centrifugal emanation. In fact, these two weren't present on all the time competing equally. In fact, they came and went. We plotted them in these colors, blue and red-orange. In fact, you can see over a minute, the blue site was sort of present for two-thirds of all cycles, then went away to zero, then came back, then stayed for a while, then went away. Conversely, the red-orange, the focal actually had a completely reciprocal relationship. When it was present, the reentrant circuit had gone away. You could argue it was suppressed. Then when the reentrant became dominant, the focal was suppressed and vice versa. You could imagine saying there's no rotation. You'd be right if you looked at from four to, let's say, 16 seconds. You could look and say there's no focal, and you'd be right if you looked at 24 to 40. Looking at periods where you have greater consistency, or in this case, ablating both sites ultimately, ignoring the fact that one caused termination and one didn't, might help to explain how we can get all drivers eliminated that are relevant over time. We're working on that too. In summary, ladies and gentlemen, contact AF mapping is a promising method to target ablation. AF driver sites and contact maps show high correlation with optical maps, and their ablation tends to improve the results of PVI and randomized controlled trials, although at the moment this is still a hypothesis to be tested. Further approaches should focus on finding true positives to standardize mapping if possible, to improve and automate mapping interpretation, and then to define the best ablation approaches for testing in randomized trials, which may include not ablating too much of the atrium, which may be a more consistent signal of patients not doing well. I'd like to thank our colleagues and funding sources. Thank you very much.

global contact mapping

atrial fibrillation

mapping techniques

ablation targets

clinical evidence