Welcome everybody, good morning. Thanks for coming to the session. This is gonna be a rapid fire topics in lung cancer and pulmonary procedures. I'm gonna moderate this session. My name's Alex Chen, I'm a pulmonologist in Washington University School of Medicine in St. Louis. So we're gonna do this a little different kind of game show style. So we've got a panel of experts here and we're gonna sort of go through. Now normally this game show, if you're familiar with this, has got four panel members. Instead of these folks, we've got these folks. So before I get to the rules session, I'm gonna introduce the panel. And I think it's kind of nice because we've got people who represent a diversity within sort of the space at different levels of their careers, but all experts in the field of interventional pulmonology, thoracic oncology. So I'll start with our most junior member, Max Wayne here from the University of Michigan. The next person up, I'm gonna do this in chronological order, is Diana Yu from the University of California, San Francisco. Fabian Maldonado is from Vanderbilt, who represents our Deep South and our European colleagues. And then Mohamed Wahidi is from Northwestern University. So we've got people who are sort of at the earlier phases of their professional career. We've also got people kind of in the twilight or the sunset of their career as well. I was talking to Dr. Wahidi. So the idea is we're gonna sort of present some topics that are contemporary. So the idea is pull some articles that have been published over the past year around lung cancer, pulmonary procedures. We're gonna have the panel members weigh in. We're gonna randomly assign some points here, although it might be more fun at the end just to do sort of an audience response of just clapping to see who gets the most claps and just to see who wins. But the idea is that they've not seen the topics. This is all off the cuff. We want things to be honest, controversial, sort of a little bit prodding, and that's the idea here. All right, so let's get to it, round one. We're gonna start off with biomarkers. Biomarkers is a very popular topic here at the meeting. We've seen this a lot. I'm gonna preface this by saying I'm gonna show you some articles. The idea here is not to talk about one product or one device, but really use this topic as a sounding board to get opinions. This is something that was published within the past year. It was a prospective registry data, so they enrolled patients prospectively. All who underwent this biomarker assay, physicians were able to use it and then determine clinical judgment based off of that. They did a propensity score matching process where they then took a retrospective analysis of some database patients they had, and then they were able to sort of say, look, if we applied this criteria to that database, would it have made a difference? All of this is for patients who have indeterminate pulmonary nodules with a pretest likelihood of malignancy, less than 50%. And in this study, they reported a 74% reduction in invasive diagnostic procedures with that approach. So, first question is biomarkers, are they fitting into your practice right now? I think we'll start off with Diana. Diana, what do you think right now in terms of the ability to throw biomarkers in? And then we'll get into some of these questions. Do you think the data's good enough? And what are realistic and meaningful endpoints to convince you? So, let's just start off, how do you practice? Where do they fit in? So, I was gonna plan on wearing a t-shirt that says, where's the evidence? So, I think I'll start off there. I think in terms of how does it fit to our practice, we don't routinely use biomarkers to determine or rec-stratify. I think that we rely heavily on physician's judgment based on our quantitative risk model calculators along with our multi-D approach in terms of getting all of our other colleagues involved in sort of managing long-term nodule management. And in terms of data being enough, I mean, currently, I don't think there's enough data for us to apply this clinically in our clinical practice to really weigh in and make a difference in our decision-making process. Because a lot of these are discovery validation studies. There's a lot of new kids on the block where it's just maybe some clinical utility data that's limited with limited number of patients as well as follow-up. I think the longest may be about a year follow-up. So, I think there's not enough data. I think the big picture is, I can go. So, that's a great question. So, we talk about data and talk about wanting more data. We talk about that all the time. Fabian, what do you think is meaningful data at this point in time? Because we ask for it. So, what exactly are we asking for? How will this prove useful? Diana just mentioned she's not using it routinely. There's a ton of resources getting poured into this space. How do we make it useful? Yeah, that's a great question. So, shameless plug here. We have a paper coming out from our lab on an update on biomarker in lung cancer coming out in chest, I think, in October. And we talk about that a little bit at the end. So, the history of studies for biomarkers has been essentially consisting of studies looking at what would happen in the best case scenario. So, you have a biomarker with particular test performance and you try to see how much you can reclassify patients from intermediate pretest probability to higher low pretest probability. And you infer from that, theoretically, what would the benefit be in terms of reduced unnecessary interventions and reduced time to diagnosis. That's what you're trying to do. And to this day, we don't have a good study showing actual clinical utility. Do you actually reach these endpoints with application of particular biomarkers? So, until we get this, and there are at least a few studies that are currently enrolling patients that are looking at this, until we get solid patient outcome measures, patient importance centered outcome measures, I think, as Diana said, I think we're probably not gonna use these biomarkers. One thing that we are looking at in our lab and we're gonna R1 studying this is to look at a gated approach. By that, I mean these studies are complicated to do. A randomized control trial, particularly multi-center, randomized control trial is a lot of work, it's a full-time job. And so the problem with these studies is that if you don't adjudicate your primary endpoint very quickly after the intervention, then it becomes a really big monster of a study. And so what we're trying to do is to apply an endpoint that we can adjudicate immediately after the biomarker has been obtained. And the way we do this is by looking at the change in management. And we published a few papers on the intervention probability curve. If you look it up, it's pretty interesting. We look at how often implementing a given biomarker will result in a change in management from the physician, which obviously is not just the physician, it's the physician and the patient that discuss what to do next. If that endpoint is not reached, there's no point in looking at unnecessary interventions and delayed diagnosis, because if it happens, it's pure luck. You haven't changed what you're doing in practice. And so that's the way to do it. I think we need randomized studies looking at a gated endpoint approach, looking first at the necessary but insufficient endpoint of change in management, and then actual patient-centered outcomes that are improved. And by changing management, Vivian, you're talking about following patients based on the pretest or the likelihood that they have cancer. So not doing a biopsy, for example, or not moving to surgery, not moving up the chain, basically, based on these results. That's a clinically meaningful response. Is that right? That's right. And it's tricky, because if you never biopsy, you're never gonna do unnecessary interventions, or you're gonna decrease the unnecessary interventions. If you biopsy all the time, you're always gonna reduce your time to diagnosis. So you need to balance these two endpoints. That's what's tricky about it. But yes, you need to be a patient-centered model. Well, and just to chime in very quickly here, I mean, we know that practices across the United States and probably globally vary. So it isn't as if physician A is gonna react the same way to a result as physician B. And so how do you look at heterogeneity of practice and where these biomarkers fit in? Because what Fabian just described is assuming everyone practices the exact same way. So how do biomarkers account for that, or do they sort of get washed away because practices are so heterogeneous? And I think that's the problem with the current data, right? We have registries, we have some bias, because people are not blinded, which is hard to be blinded. You have comparison to historical control. So we don't know. Really, we don't know. And I think where do these markers fit in our practice? I don't know. They're looking for an indication. Is it up front, every lung nodule that presents? Or maybe is it after a negative biopsy, right? A negative bronchoscopy or TTNA. And maybe that's where there's more utility. You know, we don't have data. They're somewhat expensive. And I guess if you're somebody who sees lung nodules every day or a lot in your clinic, is this for you or is this for somebody who just sees lung nodules intermittently and maybe they need that help in that judgment? So you're right, it's a very varied practice. Pulmonary critical care is such a broad field. People do a million things. So not everybody's gonna practice the same or react the same to this test. All right, let's move on then to the next topic. So pulmonary procedures, this is a meta-analysis. This was published this year, released in CHEST, and this is actually a follow-up to a previous study that was published in 2012. We're getting into sort of the nitty-gritty about pulmonary procedures, specifically for lung nodules right now. For those of you who aren't familiar with this initial study, it looked at all the studies up until 2012 for guided bronchoscopy for lung nodules. It showed a pooled diagnostic yield for each technology being about 70%, which really was the benchmark for quite a long time. The same group updated this, so thousands of lung nodules at this point in time through multiple studies. And let's talk a little bit about that. So what they found over the past decade was there was no significant difference in diagnostic yield. This table's probably too small for you to see, but it actually includes some things even on robotic bronchoscopy. And you see that over time, there's been no significant difference. Now, one thing that this study did that the previous version didn't was they did a quality assessment of the data. So all of you know that meta-analyses are only as strong as the studies that get fed into it. So the quality assessment, really using this QUADIS method, said, all right, what is the strength of these studies? Are they primarily case series, retrospective studies, or are they more rigorous, like something that's prospective and comparative? And they show that if you had a lower risk of bias, meaning a so-called higher quality study, if you will, that the diagnostic yield was actually lower than if you had a higher bias. So let's sort of get into this. I'm gonna ask Max to chime in on this first. I don't think he was born when the first meta-analysis came out. So let's get his opinion about sort of what this means. I mean, look, this is a pulmonary procedures community here. And so showing no significant difference over that 10 years of time, is that really the case? I mean, have we really shown no difference? Were we literally fooled by that initial yield with some of the studies that went into it? And then we can talk a little bit about bias, but first impressions about this information. I think first I'd like to caveat that these studies were done before I was practicing, so we can blame our more senior colleagues maybe for the less than ideal yield on some of those older ones. All kidding aside. Well done. I think that it's kind of hard to compare a lot of these different things, just because the quality of the studies of these are so variable and how they're defining yield makes such a dramatic difference in terms of how you're qualifying these data. So like, if you have one definition for what a yield is in a different study, it makes it really challenging to compare across the board. And so I think probably there is a slight improvement with some of these newer technologies, but how much of that is driven by a robot versus a cone beam CT, especially when we're combining all these different utilities together makes it a little bit challenging. And I think just the quality of data to begin with is not great for any of these different technologies. It's really, these are safe to do, but do they add benefit to what we're doing clinically is a little bit challenging to assess. And I think that's something we need to work better to achieve as a society. Momen, your thoughts? Well, I, according to my source in social media, the yield is 95%. So I don't believe this stuff. I mean, social media is, you know, it's the truth. It's 99%, so. But beyond that, you know, I think this is what it is. Whether you don't like the study or not, this is what's been published. If you have better results, publish it. You know, this is it. I don't care if you tell me your yield is 97% because that is not what the data, the published data are saying. So I think we're still trying to figure out this issue. I think we all think the technology's better. We all use it. It feels better. We're getting to smaller, you know, nodule. But there are many, many other factors. And I think maybe the game changer is do we have real-time guidance? Is it the radial EVAs? Is it the CT scan that's now becoming mobile? And that's where maybe, you know, the rubber hit the road. Is that, I don't think these studies had actually CT, added CT confirmation. So is that the next frontier? Maybe, maybe not. But it is what it is. This is the truth. All right, well, let's start talking about some of the individual technology then. So let's talk about robots. Look, you can't avoid talking about robots in this space. It's not new anymore. We've been talking about this for five years. There's a lot of presence here and in the community. So when I did a PubMed search for robotic bronchoscopy, and I limited to the past year, this is what I found. 79 articles of those 15 original investigations. And the mean number of patients enrolled in some of these studies was between 20 and 30. So there are some certainly larger studies. The majority of them, smaller studies. Out of those, only two prospective studies, okay? There were eight reviews. And in that same period of time, three meta-analyses. So there were more meta-analyses published in the past year in robotic bronchoscopy than prospective studies. So let's talk a little bit about this. Where are we with this? And this is a loaded question for everyone. Do robots help? Are they enough? And do they add value? So let me qualify this. So how do they help? Are they enough? Meaning, just as Momin said a second ago, we're seeing combinations of robot plus this. If we're hearing about yields in the 90% range, why do you need anything else? So let's talk a little bit about that. And then the value question, of course, is also a loaded topic. Fabian, you wanna start us off with this one? Sure, I'd be happy to. So I think it's prime time for anybody who wants to publish in a good journal to publish their crappy data. So if you've got this horrible yield with robotic bronchoscopy, that will get some editors to look at it twice. We just don't do that. We don't publish crappy data in general. We publish our fantastic, best, awesome data that makes us look good. We don't publish complications. And I think it would be nice to have a more representative. So all this study, or the vast majority, are single-arm retrospective studies. And just like the meta-analysis we just talked about, there's a problem with this, not just in terms of the definition of diagnostic yield, but in terms of the selection of patients, what kind of nodules to go after. Do they have a bronchus sign at the upper lobes? Is your pathologist in a good mood in general, or in a bad mood? Are you a crappy bronchoscopist or a really good bronchoscopist? There are so many variables that affect your diagnostic yield besides the diagnostic yield that comparing these studies and trying to infer from these different diagnostic yield what the value of a particular platform is, which is one of 100 variables, is pointless. So these studies cannot be compared, period. The only way we know to account for known and unknown confounders, meaning these variables that you don't care about, is to do randomized studies. We don't have any. We have one randomized study in bronchoscopy in the U.S. over the past 15 years, and the results are not great. So we need randomized studies. It's hard work. And the problem with randomized studies in this space is by the time you get the RRB to even look at the first page of your application, you get a new robot on the market that you need to study or a new version of the robot you had at the beginning that's changed. And so we need to be smart in the way we design trials. We've got another shameless plug here, but we're doing a cluster randomized study of robot versus electronic navigation right now. We've enrolled over 200 patients in six months, which is amazing because the last study I did was 258 patients in three years, and we just finished recruiting for that. So we need these studies. We need to be smart. We need to, as a community, to get these studies going. That's what we need to do. Let's talk a little bit about the clinical aspect. So we've talked about at the research end of it some of the challenges with the data. How do they fit into your practice? Are you using them right now, Diana? Well, we just, I mean, despite that we're an academic center major, we've met a lot of roadblocks to get the robot. So we do have an eye on. So does it help in terms of utility of robot? I do think that it helps in certain features of the robot that it offers in terms of articulation, your ability to get to the periphery with some visualization, not less of a deflection in the catheter for biopsies. There are certain features that are very helpful, but to everyone's point, we're supposed to be contentious and fight each other at this, on this panel, but I agree with everybody in a sense that, you know, it really does need image confirmation, right? So we talk about digital tomosynthesis, and we talk about fluoroscopic guidance in terms of you get close enough, are you actually in the lesion? So addition of these image confirmation does seem to help with the diagnostic yield. And to go back to the diagnostic yield in terms of, you know, how we define it, that's like the issue that we have in our field as a whole, because we're not really defining it the same way to give credit to some of the studies that other folks on this table have done. It's actually that depending on how you set your diagnostic yield, it can really change your outcome by 20% or even more in terms of, you know, this whole 97% diagnostic yield. If you look at that meta-analysis going back, you look at the range of diagnostic yield. It goes from 40% to 97%. I mean, look at the huge range, depending on what you define as the diagnostic yield. And so that study actually, it's not as depressing. When I first looked at that study, I was quite depressed that the, we haven't made any improvement in 10 years, but you have to realize that that does include all the other studies, including ultrathin, virtual bronchoscopy, some of the old modalities that went into the meta-analysis, right, after 2012. The robot, you know, it's in the past couple of years at best, right, with limited data that we have so far. So I do think that it's a little bit promising based on what we see, also anecdotally, and obviously we need more randomized studies, but really defining what that diagnostic yield means, because depending on how you define it, every study, major studies within IP, NAVIGATE, BENEFIT, the MSK study published in CHESS 2022, even ACQUIRED data, they all had different diagnostic yield criteria, right? Did you get it on the first tissue passed with bronchoscopic biopsy, or did you get a non-diagnostic material, but the lesion resolved, so therefore, you could consider that a diagnostic true negative malignant. So moment, I'm gonna ask you for one second here. Take, put on your chief financial officer hat here. Diana wants to take the half a million dollar system and then add another half a million dollars to that. It's 350 for a company. Well, she gets a discount because, whatever, but so moment, I mean, on top of the capital cost, and then you've got the cost per case. Look, is this something that is an every case type of thing? Is it a selected case? If you're sort of organizing a healthcare system, which you know something about, like what is, how do you approach that? Yeah, if this was a standalone technology in a strip mall, nobody would buy it for you because it's expensive, it doesn't really pay back. Expensive disposables. I think, as part of a whole picture of lung cancer care, as a CFO, I'll listen to Diana and I'll sort of make it part of this multidisciplinary approach that you're gonna, hopefully, diagnose these patients, staging them with the EBS, and then get them to my surgeon in radiation oncology. So it's a team approach. If you're alone, Diana, I'm gonna ask you to sell some tacos in the strip mall. That's a great way to end that segment. Sell some tacos. Okay, let me see here. Round two, which is very meaningless because there's no rounds here. All right, we've talked a little bit about new technology. This is the only old paper I'll show you, but it's really, really interesting, and it's relevant to what we're talking about. So this was published 10 years ago in Mayo Clinic Proceedings. This is an investigator who looked at 1,300 articles that were published in the New England Journal of Medicine, all original investigations, looking at established practices and new practices that were investigated over this 10-year period of time. And what they found was that 23% of new practices were found to be of either no benefit, they were worse, or they were inconclusive. So these practices that were adopted, standard of care, screening for certain types of cancers, BIS monitoring for paralytics and the OR, these things were reversed in that period of time. 60% of established practices were either no better, they were worse, or they were inconclusive. So 60% of the things that people were doing were actually not better than the things that they actually replaced. The way that these were demonstrated, to Fabian's point and I think everyone else in this panel, these were demonstrated through randomized control studies. So was this sort of the Nostradamus effect? I love that David Oss is here, because he sort of led this effort with the Acquire Registry. For those of you who don't know, this was a chest effort where a proceduralist logged all of their procedures for a period of time. So they were able to capture 500 cases, or more than 500 cases of guided bronchoscopy using things like Radial E-Bus, which was available, electromagnetic navigation, which was available. And the Radial Probe E-Bus, so these numbers were way lower than the 70% range that we just quoted in the meta-analysis. And at the time we said, look, this is a methodological issue. This is a registry, there's all sorts of potential biases here, and this just doesn't make any sense. But was it predictive? So moving forward a number of years, just a few years actually, this is a randomized controlled trial that sort of demonstrated the yields for guided bronchoscopy techniques using thin bronchoscopy Radial E-Bus was just barely, not even 50%. And that's a lot closer to the Acquire Registry information that is meta-analysis. There's also additional data forthcoming on electromagnetic navigation, which I think is gonna be telling also. So what is the right number for diagnostic yield? Is this just a silly academic exercise that people in ivory towers sort of talk about? Or does it actually matter to know what the actual diagnostic yield is for these technologies? Max, let's start off with you. I think it matters to know that because when you're deciding to do or not do a procedure, you should have some idea of like how likely are you to be successful at it. And I think that's one of the issues with, I guess it's not really answering the question you're asking, but as in part of the interruption debate, you kind of change the topic here. I think a lot of the issue I sometimes have with the robot is that it makes it possible for you to do, you try to biopsy things that maybe you shouldn't be biopsying. So really small ground glass lesions, maybe these don't actually need to be biopsied. And so I think in terms of what's the right number for a diagnostic yield, I think depends on what your aim here is for. So I think if you're trying to be better than transthoracic biopsy, it's gonna have to be quite high. If you want it to be safe, maybe you're okay accepting a little bit of a lower answer there. Momen, what do you think? I mean, for a long time we've been holding 70% as the gold standard for what we're trying to do. We look at CT guided biopsies being 80%. Of course, we know that that's somewhat cherry picked data as well. But how do we know if we're doing better than we were without a baseline? So where do you think that sort of sits? Well, first of all, Max, you're gonna go in the first round of the draft. I'm liking your answers of the football draft. But so I think we should ask it differently. I don't think it's the right number is, do we agree on a definition of diagnostic yield, right? And I think I'm gonna punt here to Fabian in a little bit because he led the effort with others for ATS and ACCP on writing that document to say, hopefully anybody that will publish on diagnostic yield moving forward should stick to these standards because diagnostic yield can be a very fluid definition. You can change it. You can say things are diagnostic or non-diagnostic. And so we really need to stick to these standards. Once we have that, Alex, I'll take anything. I'll take a good one. All right, Fabian, why don't you pick it up from there? Okay, yeah, so we have a consensus statement that's coming up in the Blue Journal, hopefully soon that will explain how we think we should define the diagnostic yield. But I would say that the definition doesn't matter actually all that much. What matters is having comparative studies because then you apply the same definition for your two groups. So for our randomized study of navigational bronchoscopy versus CT-gated biopsy, we're using a different definition than the one we propose in the Blue Journal paper. It doesn't matter. As long as you're comparing apples to apples, you're gonna be fine. In terms of what's a good diagnostic, what depends what you go after. So if you are in a center where, you know, interventional radiologists look at CAT scans and they say I'm just biopsying in tumors over two and a half centimeters, the easy shots that are, you know, where I can biopsy them, and you have to biopsy all the tough ones, then you may have a diagnostic of 55%, which incidentally was our diagnostic yield before we started using real-time imaging with a new technology, and we were pretty darn good at it, and we got a ton of patients, and we did a lot of good for patients because there was nothing else. What are you gonna do after that? If you don't get the diagnosis, then they get surgery, and if they get surgery 20% of the time for benign disease, that's not a good thing for patients. So I think the definition of diagnostic yield varies based on your population, and that goes back to the single-arm prospective studies. I think they're useless for me if you tell me that your center has 95% diagnostic yield. They're great for you because that means you know how to select your patients, you know which patients you're gonna get a good diagnosis for, and you exclude the ones that you think you're not gonna get, so that's good. All right, very good. Let's move on, AI. So this is something that's very prevalent outside of the medical community, very much so within the medical community. There's lots of different definitions for machine learning, deep learning, those types of things. Let's talk a little bit about it. So you've got things like nodule management software where a software system is gonna pull anyone's chart that has pulmonary nodule in it so that you can capture them so they don't get lost to follow up. With pathology, is there the ability to use AI to improve the uniformity with which pathologists grade things such as cancer? Because we've learned recently, there's a lot of different ways to grade this, and there is some subjectivity when you talk to different pathologists. You might get cancer on one, atypical cells on the other, or the exact opposites, which is very, very concerning. Pleural disease, this is a very cool article, differentiation of malignant from benign pleural effusions based on AI. Proof of concept, what about differentiating benign from malignant disease during bronchoscopy? So what if you could just put the scope on a mediastinal lymph node and it says, like, this is cancer, this is not cancer. If it's not cancer, don't bother sticking it, move on to the next one. So there's lots of different potential applications here, and so let's talk a little bit about where we think these applications are best suited, and then the things that concern you the most about this. Diana, you wanna start us off? So I think, I mean, I can't say that I have too much knowledge on AI at this point. I think there's a lot of cool things that's being published, including one of the pleural disease that was published, looking at sort of segmentation patterns in the CT scan. They look at special imaging to see if they can predict whether or not this is malignant versus benign. So I think it all kind of boils down to tissues. The issue is always gonna be the case. You know, I don't think we're ever gonna feel comfortable looking at something to predict what it is without actually getting some tissue and diagnosis. So where AI would play a role would be, I think the big picture that we're trying to answer is to improve patient outcome no matter what we do, right? Minimize invasive procedures to minimize the cost for the patient, for the hospital, et cetera, and safety for the patient at the same time. Do watchful, do surveillance enough so that we pick up early disease at the same time, do no harm, right? I think that's the big picture. And we're all focused on patient outcome. So with that, I think AI can play a role in certain things. Like we are human, so we do make human errors in terms of nodule follow-up. Some of these tiny, tiny micro nodules that I see in patients that incidentally get discovered and we're following at a certain point, but when you go back five years on the CT scan, that has been there for all these years. So if we can figure out a system where we can integrate some of more human-prone error sort of system where we can use AI for deep learning to incorporate that into the physician's assessment and what we can offer from diagnostic standpoint to get the tissue. I think combination will be sort of the best strategy because we need to find better risk stratifications models to begin with. I think the way things are now, it's not individualized to your patient, right? You wanna say Mayo model, Brock model, whatnot, but that's specific for a certain patient population. So you need to be really aware of what patient population that you're plugging that calculator into. So where AI would be helpful is like, you obtain all the data from individuals, including non-smokers. I have plenty of EGFR positive non-smokers, no risk factors. You put them in a calculator, it'll say low risk. And I diagnose cancers all the time. So how do we incorporate all these individualized patient data into the AI system that's deep learning to incorporate that into us making a better risk stratification model, and also to kind of condense this whole range of intermediate risk. Five to 65%, that's a huge range. Anybody falls into that bucket. Is there any way for us to define that further with maybe assistance of AI? I don't know, but I think there's a lot of possibilities where we can incorporate both and not to take away the human aspect of it to really kind of practice what's good for the patient. Moment, talk to me a little bit about some of the concerns you might have about AI. I mean, I'll give you an example of this. So let's say a nodule management software is sort of picking up lung nodules from the ED, all these patients that come in. I mean, is there a possibility of over diagnosis, those types of things? I mean, what type of oversight do you think is gonna be needed to sort of combat that? Well, before I talk about this, I think I wanna talk about some really great examples. I think AI is fascinating, we should embrace it. I think I'd like to see it as a diagnostic aid to us. I have a funny story. I was teaching on a simulator, EBUS simulator. You have them here if you've attended the session. And I was teaching pulmonary fellows, and in the EBUS simulator, you can turn on the labeling feature, and then all the structures will have the name on them, like 4R, SVC, and literally one of the fellows thought that was like the normal EBUS. Like in real life, he'll go, and he'll put the EBUS scope, and it will label to him like 4R, SVC. I mean, wouldn't it be that cool if AI does that for us? Like teach AI, you can teach AI. You can give it like thousands and thousands of EBUS images and videos, and eventually it will label it as 4R and SVC. I think that's great. And something like the lymph nodes, like teach it where is the best area to sample. Avoid necrotic area. There's elasticity things. Maybe it will light up and tell me, hey, sample here. So I think there's so many fascinating applications. I think oversight, yeah, of course. I think we're the superior race humans, and we have to make sure that we approve it. We kind of like validate it, approve it. I think the last word is to us humans, I think, in how the decisions are made. All right, let's keep on going here. So the next topic, very sort of of the moment here, bronchoscopic ablation. This is a study that was published by an international group of investigators, Safety and Feasibility of a Device, a dose escalation study for essentially ablation. So again, we're not here to talk about any one way to say that that's the best way. These are different ways that you can do it. So you can microwave it, you could RFA it, cryo-steam it, vapor it, pulse electrical field, or you can inject it with pharmaceutical agents. So look, there's a lot of options out there. Some of these you could do right now. So that's the question. So would you do it right now? You have these options available. Would you do it now? Would you do it ever? Why would you not just send the patient to stereotactic radiation therapy if they're not a surgical candidate? Let's talk a little bit about that. And then what evidence do you really need in order to start adopting this practice? Fabian, what do you think? Is this something that you would, would you do it now? I'll sort of start with the low-hanging fruit. So I would not do it now. Would I do it ever? Probably. I think it's coming regardless of what we think about it. I think there are good potential indications for this. And I think the most likely indication is probably not oligometastatic disease, as the FDA was considering last year in the public workshop that some of us participated in. It's probably gonna be for indolent adenocarcinoma and multifocal adenocarcinoma. I'm thinking if I try to predict the future here. What do we need to make this a realistic option for us and our patients? Well, we need data. I think we have none right now. And when you publish your good data and you don't publish your crappy data for diagnostic bronchoscopy, it's one thing. When you don't publish your crappy data on ablation, it's a different one. And so until we get a little better handle on what's going on, then we really shouldn't be doing it. Now, you have to come to the session tomorrow with Momen. What time? I think 8.30 maybe? 8.30, because we'll talk about this. We're in a very different situation than SBRT was when SBRT was introduced. We needed an alternative to surgery. So we didn't have SBRT. Now we have SBRT. So the landscape is different. The thought process is different. And we need to find either niches in which SBRT is not good, which is arguably almost never. Maybe you got a radio-resistant tumor like a renal or whatever. Or you have to have a comparative study that shows me that we're at least not inferior to SBRT. And if we're not inferior, then great, because actually you could potentially think of diagnosing, staging, and treating in the same setting which I think is a potentially great advantage. But you need to show that you're not inferior to SBRT. There is a competitor now that did not exist when SBRT came about. All right, very good. I won't call on Momen, because he's gonna talk about this tomorrow. Diana, what do you think about this whole process? I mean, so you have a patient who is a non-surgical candidate. You know this before you do your bronchoscopy. So you have the option of going in and taking the time to do your full staging exam, get the diagnosis, and then you treat. Is that a viable pathway for a lot of these patients as opposed to stopping the procedure and then sending them on somewhere else? So to echo Fabian's point, I do think non-inferior to SBRT is critical. The answer is no. I mean, this whole holy grail of interventional pulmonary, when I started fellowship, I think the whole holy grail was we have the opportunity to diagnose, stage, and treat at the same time. That's still our holy grail, I think, as our field moves forward. I think that that's something that's a foreseeable future. But we do need more data to capture, to say at least prove non-inferiority to our radiation oncology colleagues. Otherwise, just because we can doesn't mean we should. I think that's sort of the theme here. I mean, we could biopsy everything, including a two, three millimeter long nodule. Does it mean that we should be doing it? The answer is no. So in terms of catheter-guided treatment, I think there are a lot of potential out there, including pulse electric field, formulating sort of tertiary lymphoid structure. There's a lot of promising data that's coming up. I do think that we do need to see it through and see where that leads us. But I think currently it's just premature to do any catheter-directed therapy. All right, Max, I'm gonna play the devil's advocate for one second here. So let's talk about this. Do you think non-inferiority is good enough? If you have a process that works very well, that's typically safe to do, and is not incredibly uncomfortable for patients, is a bronchoscopic approach that is non-inferior good enough for you to say, all right, I'm gonna take on this responsibility of curing this person's cancer? I honestly think you would have to be better than SBRT to treat one of these lesions because you're putting that patient through a more extended procedure. And so if it was my family member, why would I tell them to go through this prolonged procedure if they could get external radiation in a couple days that is incredibly safe and effective? I mean, it's hard enough to get, like the VA is currently participating in a trial to compare SBRT to surgery. And that's, I mean, we've been doing SBRT for quite some time now. Like, I can only imagine how long it's gonna take us to get comparative data between SBRT and ablative bronchoscopic techniques. If that's even something that's feasible. Let me just, pardon the interruption. So. That's a different show. No, I do think, okay, I agree to some extent, but I think there is a potential rationale for having a all-in-one test that is cheaper and for the patient, just a, you know, two hour or one hour and a half procedure as opposed to coming back for three days or, you know, SBRT is not that comfortable for patients either, to be honest with you. When they're sleeping on the table, it's a different story. I just wanted, actually, that's why I interrupted. I wanted to say, when I say crappy data, I've been saying this all the time. I'm not talking about qualitatively crappy data. It has to be good data, solid, robust data. But what I'm talking about is having the courage to publish about outcomes is what I'm talking about. So I just wanted to clarify this. Fair enough. All right. Can I add something? Again, shameless plug for our session, PROCON debate tomorrow. You know, I'll review the path to SBRT becoming sort of standard. It took 15 to 20 years. So we gotta be patient. Patient, do the studies, and you'll see tomorrow. I mean, it's not overnight. It's not a year. It's not two. You gotta follow these patients and really, you know, have a rigorous studies to prove that this thing works. Fair enough. All right, very good. We're gonna move on. Look, we've talked a lot about pulmonary procedures. This is something that a lot of people are not as familiar with, but the people on this panel definitely are. So this is a process by which new technology actually gets in your hands. And it is worthwhile just spending a few minutes to talk about this because so much has been talked about in terms of generating evidence, clinical evidence that says this is how well a product worked, this is who can use it on. So I'll just go through this very quickly, but emerging technology being granted to the FDA is very dangerous for a pathway to sale through an expedited process known as the 510K process. So if you're coming out with a product that has a predicate device, and you can show substantial equivalence, meaning, hey, look, what I'm putting out in the market is really close to what this means. And by the way, really close can be a big range here. They'll say, yeah, you can go to market, but you don't have to generate quite as much evidence as if you're brand new on the market with a new device entirely. That requires a bit more evidence, basically, clinical evidence. And so this is the way a lot of this information and material gets into your hand in this technology. So I guess the question to the panel is the 510K process helpful in getting equipment into the hands of healthcare providers, or is it harmful in lowering the expectations for generating clinical evidence for new technology? And I suppose it's safe to say that, look, if we're talking about new technology, and we want evidence, in theory, if something went in sort of the non-510K pathway, we might have some of that. It's fair to say that we don't, because some of these products have gone this way. So Fabian, why don't you start us off with that? Yeah, so I'm gonna take a little bit of a different approach than what I've published on the topic. I've published some about this. So if we did not have the 510K pathway, we would not have any devices like robotic bronchoscopy and electromagnetic navigation and da Vinci. We would not have them, and arguably, I think it'd be a disaster. The reason why is because, as opposed to pharmaceutical companies, which go through a very stringent pre-market approval pathway for drugs, which require randomized studies, device manufacturing companies typically don't have the deep pockets of pharma. They just don't. The vast majority of device companies, I work with some of them, have very little money, and they gotta obey to such international standards to begin with that by the time they have a product that they can submit to the FDA, there's no money for a $10 million randomized study. It's not gonna happen. What this means is that we're in a system where over 90% of all medical devices in the US are commercialized without any evidence of imperfection outcomes, and that's not gonna change. That's the way it is, and that's the way it will be for the future. This means that it's shifting the responsibility to producing patient outcome data to us, and we have to do it, and we're not paid to do it, but unfortunately, we have to do it. And so, again, kind of belaboring the point, we need to be organized as a group of interventional pulmonologists and have an infrastructure, a network that can produce quality randomized data in a timely fashion without being outpaced by innovation. That's very important. A moment. What do you think about that? You've been in the game probably longer than anyone in this room, so talk to us a little bit about the 510K process. Yeah, I mean, I agree with Fabian. I think it's a two-edged sword, right? You need it because otherwise you stifle technology. Nothing comes to the market. On the other hand, there are loopholes, and then technology that could be harmful are following that path. So, and this is a political issue for the FDA, and then we last year participated in an FDA panel on this issue for bronchoscopic ablation, because they're struggling with this. So actually, they invited experts in radiation oncology, oncology, interventional pulmonology, thoracic surgery, to help them decide, does bronchoscopic ablation of tumor, should it go through 510K? I mean, it was not the question particularly, but they wanted to hear what is out there, and they wanted to learn. And so, it's a tough, tough area. And again, on the industry side, it's welcomed. And we welcome it for certain things. Otherwise, like Fabian said, we would not have a lot of things in our hands today. So what is Moment, then, the motivation for industry, then? They have their product on the market. What is the motivation, then, to continue to do investigations at that point? Why would you do it? Zero. They don't. They don't. And that's the problem. It's a risk, I think it's a risk, right? I mean, but things like understanding whom the technology works best in, whatever that information is. There's definitely opportunities to learn more so that we can use the product in the best possible and safest way. So, sorry, pardon the interruption to my show. But since you called me old, and by the way, I looked it up, middle age is 40 to 60. I'm middle age, so thank you. So, you know, I've been in this field for 20 years, and I've seen technology come and die, die a miserable death. They were 510K approved. Some of you may remember, we had, does anybody remember autofluorescence bronchoscopy? Raise your hands. Okay, those are the middle age people, thank you. Autofluorescence bronchoscopy was gonna change our life. It's this special light that identifies pre-cancerous lesions, and we're gonna diagnose cancer on everybody and cure them. And they got 510K approved, no study. Then the investigators, the interventional pulmonologists started doing the studies, and even without study, the technology will speak for itself. After you use it a little bit, you know, out there, you're like, I'm not getting any benefit. It's picking up everything, from chronic bronchitis to this and that, and so these technologies die. And so sometime industry is mistaken not to invest in research studies that show the value of their technology. And the problem is sometimes technologies die by error, because we haven't produced the data that would have shown that it was effective. To your point, Alex, you know, there are smart ways to incentivize good studies after commercialization, and it's been done before. So one way is to agree to a conditional FDA approval, and you have to produce data showing that within five years you're getting improved patient outcomes, and you have to do it, otherwise we pull your clearance, your 510K clearance. Another way to do this is to only provide reimbursement from private payers or Medicare for patients who are enrolled in the study, which will be aiming to show improved patient outcomes. We have precedent for this, the RVRS study, the NET trial was the first trial that was funded by Medicare, and they only reimbursed the procedure if the patient was enrolled in the randomized control trial. So we can do that, it's just a little bit of advocating and policy change. Very good, all right. We're gonna close this out in the final round here. We have 10 minutes left. We're gonna bring this back to a clinical aspect. So we've talked a lot about the different technologies here. Some people have mentioned cone beam, mobile CT, robots, things like that. What is the ideal bronchoscopy suite in 2023, or let's hopefully not date ourselves, for the next 10 years look like? So what does a bronchoscopy suite of the present and the future look like? Here are the different tools you have to choose from. It's like picking off a menu. Convex eBus, thin scopes, ultra thin scopes, radial eBus, robot, mobile CT, cone beam, 3D, fluoro, electromagnetic navigation, and let's not forget about the anesthesia component as well. Are you gonna do these procedures in the future in the operating room with general anesthesia? An endoscopy suite with general anesthesia? Are you gonna use things like moderate sedation? So in this day and age, let's just sort of go through this. Max, you have unlimited resources. You won Powerball, and you get to pick whatever you want. So off that menu, where are you getting? If I have all the money in the world and I wanna buy it all then and have all the fancy toys, then I can make a name for myself and advertise to everybody that they should come to me. I have every single fancy equipment. I think realistically, I think certainly linear, convex eBus, a radial eBus probe, thin bronchoscope. I think probably a robotic platform and some type of 3D fluoro machine makes sense to me. I also think. Everything. Everything, yeah, pretty much everything. I think you just listed everything on there. But I also think importantly, that's something that we kind of don't think about is having the space to do these procedures, having a dedicated suite with enough pre-op and post-operative beds so that you're not spending, waiting for your patient to recover in the room and you can have the turno and the staff. And I think sometimes not having the staff is maybe more important than having the equipment sometimes. Very good. Diana, you have champagne taste and a beer budget. You don't have Max's Powerball. You just want like a little bit of his pot. So what do you pick if you're trying to be strategic, I suppose, with the choices that you make? What can you sort? What do you think you need to have? And what's a nice to have? I think need is. So recently converted to, and my colleagues, they're converts of robotic bronchoscopy. There were a lot of skepticism around it, but I think so far so good. So I would say robot, in terms of, do we need a comb beam CT? It would be nice to have a comb beam CT. We currently have a fluoroscopy that seems to work pretty well for confirmation. And so far, a good diagnostic yield, whatever that definition may be. Staging for sure. So we do need a E-bus. I do teach my fellows that you have to figure out a way to do this in an old-fashioned way. Machines fail, as in a robot could jam. The power won't turn on. I've had a lumicide. I won't name specifics, but I've had experience in sensor leads malfunctioning, the board malfunctioning, the CD not uploading in compatibility because of thin cuts or thick cuts. So there's a lot of things that can fail on you. So I do think that learning really the radiographic, sort of study the radiographic like anatomy and know exactly where you're going. You should be able to get to a certain place. I know that you had a pretty good track record of radio E-bus localization, Alex, but that's important. I think that being able to use your P-190, navigate out to the periphery to really mark and nail that, that's when I feel most rewarded instead of just following my tools and getting there and sticking a needle in. So I do think that you need to have a backup system for sort of old school way of doing things, for sure, because machines fail. I think that we rely too much on the fact that robots never gonna fail you or malfunction, but they do. So I think that good old technology and good old sort of skillset is important. So I don't really love being referred to as good old, but that's fine. The good old. Not as old, the old way she talks about. Let's talk a little bit about the old way, Fabian. I mean, you have a fellowship program at your institution. How do we teach upcoming generations about these old ways when so many places are converting to doing most of their procedures with robots or navigation or other types of things? Well, that's a good. Does that worry you at all? Not really. Okay, so what worries me is rigid bronchoscopy skills, which will always be needed, but decreasingly so, and that's an issue that bothers me. It doesn't bother me that somebody is very good at robotic bronchoscopy and will be using this for the rest of their life. I guess it depends if they go to a practice where that's not gonna be available. They'd better be ready to do it in a different way, but we can adapt. You know, in terms of, I was gonna say to the question what my ideal bronchoscopy suite would be, it would be me sitting at home drinking coffee and watching the autonomous robot do the whole case, maybe with a fellow nearby, and then I realized that I just crashed my Tesla these past few days and it's totaled and maybe that's not such a good idea. So I think we're gonna have to think about that. I think the way I think about it is really three areas. First one is an assortment of scopes, sizes. Two, a tool for central bronchoscopy to sample lymph nodes. Third, a tool for peripheral bronchoscopy, and you choose. These three is what's needed for a functional bronchoscopy suite. Again, you can do it in a stage approach. Maybe you're new, you get some of these, and then you show that you built up the volume, and then your bean counters can say, okay, I'll get you that third thing you want. But it should be a stage approach. But these are the three areas that I think about as I build bronchosuites in my health system. You gotta have one in each of the categories. All right, I think that those are great answers. I mean, I personally am concerned about any one person or center being too reliant on any one given technology. That scares me. We see it in the community when a system goes down. It sort of makes those procedures undoable, essentially, for a period of time. So I think redundancy, until we find out which product or pathway is a superior pathway, very clear definitions, I think it's important still to be able to do things in different ways, because the data so far looks like there isn't a clear winner just yet. And that's gonna probably change in the next few years. But until good investigations are being done, that's probably not a bad idea to have. All right, we made it through all the topics, which is fantastic, over the past hour. And so we can choose a winner. I think Diana wanted it the most, I feel like. I think you want it, Jeff? I want it, but I think you'll get it. All right, we're gonna do applause for these people. So we want their support.

lung cancer

interventional pulmonology

biomarkers

clinical practice

diagnostic yield

pulmonary procedures

new technologies

robots

clinical data