Good morning, everyone, and great to have you here. Almost the last session, I think, in the meeting, so thank you for sticking around and coming to this session. So we're gonna talk about the peripheral lung nodules, where are we now, and provide some updates on the lung nodules. We have a fantastic panel of speakers here that will cover new areas in developing tools, bronchoscopy, and then Dr. Vera will end with some areas to lung cancer screening, so I'm gonna start. All right, so my name is Mohamed Wahidi. I'm a professor of medicine and interventional pulmonologist at Northwestern Medicine in Chicago, and I'm gonna talk to you about evidence-based approach to peripheral lung nodules. So then, a minute after I submitted this session, I immediately had regret, how am I gonna review evidence-based approach to the lung nodule in 10 minutes? So bear with me, I'm gonna just kinda go quickly through some of the guidelines, particularly the CHESS guideline and BTS guidelines on lung cancer, and try to answer a couple sort of controversial questions. But what I can tell you is this, the evidence-based approach to the lung nodule hasn't changed a lot, so the principles are still there and still remain valid. These are my conflict of interest. So the peripheral lung nodule that we all talk about is a bit dreaded by us, right? Obviously, it's bread and butter for us, but when we see this in clinic, and this is chest X-ray, most of the time we see it on CT scan these days, we are agonizing a little bit because we have this dilemma. If this is cancer, we wanna get to that diagnosis, we wanna get the patient to treatment as soon as possible because of the survival rate of stage one, stage two lung cancer. The flip side is unnecessary procedures, surgeries, cost, morbidity, and all the anxiety to that patient, I left that little cartoon. This is what our patients go through. They're asking themselves, do I have cancer, am I gonna die, is this probably a scar, my aunt had it, things like that. So this is a real struggle for the patient, too. So we have detection, what are the tools? Again, our tools, still kind of basic stuff, clinical history, very important to get the old imaging studies, PET scan in the appropriate patient population, and then we have the prediction models, which I'll talk a little about. But really our goal in clinic is this, is what is the pretest likelihood of cancer in this patient? Is it low, intermediate, or high? And based on that, we decide in our head that this is likely benign, likely malignant, or this intermediate, indeterminate, and this is where we wanna deploy potentially our tools. So this is our sort of approach. You're trying to put everything together in that funnel and come up with that probability. Again, less than 5% is low risk, 5% to 65% is intermediate, more than 65 is high. According to the guideline, it is important to do this pretest probability of lung cancer, and this is the chest guidelines that says, in an individual with a solid indeterminate nodule that measures more than eight millimeter, we suggest that clinicians estimate the pretest probability of malignancy either using your clinical judgment or using one of these validated models that we'll talk about. So this is an important approach to the lung nodule. So what are these risk factors that affect your determination of the risk? Patient's factors, age, smoking history, history of prior cancers, and nodule risk factors, the growth of the nodule, the size, the border, and morphology, and I'm going to review some of these here. Growth rate, obviously, is something that's important, and it's usually expressed in volume doubling time, or VDT. Typically, one doubling in volume of an SPN represents about 26% increase in the diameter on chest CT. Okay, so it's not a tremendous growth, but even a small growth can represent a volume doubling time. Typically, the VDT for malignant nodule is anywhere between 20 to 300 days. This is where that time-honored, two-year stability of the nodules came, right? Because the longitudinal studies in radiology showed that it's rare for a solid nodule to stay stable for two years because of that VDT that we know. That is a different story for ground glass, as you know, which can grow very slowly over five years or so. So for solid nodules, this is where the two-year stability comes from. Examples of growth, you know, we see this patient with a nodule, didn't want to do much about it, small. Three months, it's still growing. Six months, still growing. So the question, why don't we do anything at three months? It was patient's preferences, and finally got resected. It was an adenocarcinoma. And on the other hand, this is on the left-hand side is a nodule detection, and then on one month follow-up, you can see tremendous growth and appearance of other nodules, and this was a case of an infection. So typically, cancer doesn't grow or explode this fast, typically. Now, there's always exceptions. So, you know, have your heightened alertness about other causes of lung nodules other than malignancy. We know that nodule size is important. This is one of the trials, and we actually summarized it when we did the lung cancer guideline in 2007, but clearly there's correlation with size. You can see in this one study, less than five millimeter was zero to 1% risk, five to 10 millimeter, six to 28%. And this is when, you know, I get a little anxiety. If I have a nodule that's more than 20 millimeter, there's a good chance this could be malignant. This should spring you to action and more aggressive, or at least figuring out the likelihood of malignancy in this patient. And then we talk a lot about calcification. We don't see them a ton in clinical practice, but I think awareness of them are good. So the top row is that benign calcification pattern. So the central calcification, just the dot in the middle, the laminar calcification, the fused calcification, the whole thing is a blob of calcium, or the popcorn, which I'll show you. Those are usually benign patterns. The lower panel is the malignant calcification pattern. So the teaching point is malignant nodules can have calcification. So this is the speckled calcification pattern for malignancy and the eccentric calcification. And I was curious why eccentric calcification is a representation of malignancy. It turned out this is the classic, what we've heard about scar carcinoma, right? So you have a scar and then malignant cells transform and become malignant, and then they push the calcification to the side. And that's your scar carcinoma pattern, eccentric malignancy. This is our favorite. You see it on board exam, popcorn calcification on CT. And this is hematoma. Obviously, it's not the only pattern that hematoma presents it with, but it makes you very suspicious of hematoma and it might alter your management. Margins, we have the smooth, very smooth margins. We have the lobulated, like a broccoli, I call it, or spiculated, where you have rays coming out of the nodule. And we know from various studies that smooth borders are less likely to be malignant than lobulated, spiculated. Spiculated is the higher risk in all the prediction models. So much higher risk of cancer in spiculated nodule followed by lobulated, followed by smooth edges. And then morphology, we know that pure grand glass opacity are more likely to be malignant than solid nodules. This has been shown in many longitudinal research studies. And more importantly, development of a solid component in a previously non-solid nodule is a strong presumptive evidence of invasive malignancy. So that should be a call for action. Just kind of showing you various patterns here of pure grand glass on the left-hand side and part solid adenocarcinoma. All right, a little break here. We don't have an obvious response system, but are you able to spot the abnormality in this CAT scan? Everybody sees the lung nodule? Yeah? But you guys didn't see the gorilla. You see the gorilla? And this is actually a study, this is not a joke. This was a study that was given to 20 radiologists, I think. 83% of 24 experienced radiologists missed the gorilla because they're scrolling up and down. Saw the nodule down there, but not the gorilla. So this is a tough job. And that's why maybe AI may play a role in the future. And maybe Anil will touch on some of that. But I wanted to ask a couple questions for you here today. How about these pesky prediction models? Mayo, VA model, Brock model, Bayesian method. Is there a favorite? Everybody has their favorite. Everybody's using Mayo Clinic. So I wanted to just kind of dig into that a little bit and give you sort of a summary. So the Mayo Clinic actually, although it's sort of our favorite, it's actually a model that was developed in 1980s in one institution, Mayo Clinic. And so it's not really representative of the entire population. And they excluded patients with history of lung cancer, history of extra-thoracic cancer within five years. VA model, as you expect, its population is mainly older male smokers, nodule size of seven to 30. And they did not have a lot of clinical predictors included in the model. And then finally, the Brock is really a screening population and is sort of a newer model, but it did not incorporate PET scan findings. So no model is perfect. In fact, I love this graph. This is same scenario. 70-year-old man with speculated nodule in the upper lobe. And look at where the risk is per model, right? There's a lot of variation. But at least the British Thoracic Society thinks the Brock is a slightly better model. Look how it performed in sort of the lower size, right? The nodule between zero to 10. It gave the lowest likelihood of cancer. And there's a feeling that maybe Brock is slightly better. And I'll show you the recommendation, at least from the British Thoracic Society. So again, British Thoracic Society say, use the prediction model just like Chess did. And this is the second line is a practical tip. I actually like to use the prediction models and put that probability in my notes. It's good to communicate with other physicians who read my notes. It's good for the patient to see, right? I often show it to the patients, show them to see and say, hey, by the way, when you put it in these models, this is your risk of cancer. And that's why we're doing X and Y and Z. And I think this is a really good approach to involve the patient and showing them what we're doing. So at least BTS said, use the Brock model. I'll leave it up to your preference. Maybe in a discussion, we'll talk about it. But the message should be use any model better than not using model at all to form your pretest likelihood of malignancy. I'll end with this area that has kind of been bugging me a little bit. Are there any important differences in nodule characteristics according to the route of presentation, right? How do lung nodules come to us today? Could be somebody had a respiratory symptom, so they had an X-ray CT that led to discovery of the nodule. Could be incidental finding for chest X-ray CT being done for other reasons. You know, it's amazing that we're having all these cardiac scans and we're finding all these nodules, right? That's an interesting new source of nodule discovery. And then we have lung cancer screening. And then we have these patients with known cancer undergoing staging investigation or follow-up imaging and something pops up in their lung. So the summary is we actually don't have any sufficient evidence to say that you need different approaches to these nodule based on how it was detected, right? So you're still gonna approach it the same. You know, screening studies obviously recruit asymptomatic people at high risk of lung cancer. So there's a high risk of cancer in general prevalence. Incidental nodules could be either, could be these young patient getting some other studies, but it could also be these people at high risk who may have similar risk factors to lung cancer cleaning population. And then again, the prevalence of malignant nodules is higher in patients with extra pulmonary cancers, but the relationship is not very well defined. So that risk factor is not as clear to us. So I think the conclusion is the root of presentation should not be an important factor in how you manage these pulmonary nodules. Approach them similarly once you have that nodule and decide on these management. I'm not gonna talk about management today. We know that we're talking about either watchful waiting, biopsy with bronchoscopy or TTNA, or potentially surgery. I'll end up with this. How do we do in following the guidelines? Not great, right? We don't follow guidelines. And so this was a nice paper from Tanner, multi-center observation study, 377 patients with nodules. And 9.5% of the patients had nodules that had a low risk for malignancy. But you see what happened here. The rate of surgical resection was similar among all groups. 17% in low risk, 21% in immediate risk, and 17% in high risk. So we're not following guidelines. We are doing interventions potentially on low risk patients. And we need to do a better job of making sure that we follow the guidelines and the evidence. And finally, the most important thing is talk to your patient. As I mentioned, offer accurate information. Tell them about the risk. Tell them about how you concluded that that's the best option for them. I love this. This is some education materials from ATS. And this gives you a scale of the size of the nodule to show it to the patient and sort of make it real life. Is it a pea, a cherry, an apricot, or a lime? I'll finish up with this. I know this patient's nodule is really high risk. Can you tell me why? Can somebody spot it? Doug did. There's a packet of cigarettes in the CAT scanner that is showing up on the patient's scan. So immediately, the pretest likelihood is probably pretty high. So this is not officially in the guidelines, but something to watch for. So thank you so much for your attention. So now our next speaker is not Alexander Chen, because Alex had to leave last night. So it's a pleasure to introduce Dr. Chris Kapp, who's at Northwestern Medicine, Interventional Pulmonology. And he's gonna talk to us about bronchoscopic and biopsy approaches to the lung nodule. Thanks, Momin. I'm obviously not Alex Chen, as Momin said, so if you don't like the talk, please direct all inquiries at Alex. These are his disclosures, and they are significantly more than mine. So keep these in mind as we're talking about them. So I think Momin started the conversation about this, but with these nodules that are being detected on LODO CT scan, we're finding more and more. And now the question is, when we decide we're going to biopsy them, how do we do that? So navigating out to the lesion, we've shown, and we'll get into some of the data about that, that in cadaveric models and then also in actual patients, that we're able to get out to nodules, but that doesn't necessarily guarantee diagnosis. And then anatomical relationships, still, despite the fact that we have all these fancy new toys, still strongly influence diagnostic yield. And then we're kind of discovering, as we use tool and lesion modalities, like cone beam CT, that CT to body divergence might be a bigger thing than we first recognized. So robotic bronchoscopy, I think I still called it a novel tool in an abstract I submitted, but I guess we're kind of going further away from that being novel, as in they were initially approved in 2018 and 2019. But the robotic, all of the platforms, this is an agnostic talk about the platforms, but all the platforms are able to navigate out very well into the periphery. They have very good articulation, and it can allow for more precise sampling. So I think that's the biggest benefit of the tool, is that you were able to get out further, and then you have stability, instead of having to hold a bronchoscope and throw a biopsy needle. So Alex and colleagues did this study back in, it was published back in 2021, and it was the benefit study, and it was evaluating the safety and feasibility of robotic bronchoscopy on live human subjects. So note the endpoint was not diagnostic yield necessarily, but just successful navigation to the lesion, and confirmed with radiolibus. And then there were also safety outcomes. If robotic procedures were non-diagnostic, then patients crossed over to, on rows, then patients crossed over to conventional bronchoscopic approaches. So as you can see, the primary effectiveness endpoint was reached pretty significantly, with 53 of the 54 cases having a radiolibus view, and lesion localization was 96.2%, and the time to localization, with any adoption of new technology, the question is, how long does it take to kind of integrate that, was only 13 minutes. The concentric view occurred in 60% of the patients, and then the eccentric, essentially the other 40%. And then, despite the fact that we were able to get a radiolibus view in most of the cases, the diagnosis was still only about 75%, roughly, which has kind of been consistent with a lot of the tools that we've used over the years. Malignancy was diagnosed in 33 out of the 40, and non-malignant were in seven patients. We're pretty good when we have a concentric view, but we're not as good when we have an eccentric view. And there's kind of the distribution of the malignant, with most of them being non-small cell. So there are perspective observational studies looking at the robot, and kind of identifying it, because I think a lot of the data that has been published thus far has been single center. So there's a couple of trials, you can see the listings down there at the bottom, but one of them is looking at the ion, and looking at navigation success, biopsy success, sensitivity for malignancy, and then complication rates. There's also transbronchial biopsy, the target study there. And then I think Anil will probably touch on this a little bit more than I will, but read the fine print about diagnostic yield as you're evaluating anything. How did they treat atypical cells? What was the patient population and the prevalence of malignancy? So I think as a group of bronchoscopists, we kind of need to do a better job of standardizing what we count as a diagnostic procedure, and there's more coming down the pipeline about that. So what's happening now, and I touched on this just a little bit, but there's far too many single center case series reports. We don't have comparative effectiveness studies, so ion has not been compared to, Monarch has not been compared to NOAA, and I think, I don't know that those trials will ever really happen, so we don't really know which robot is the most effective. And then we need to combine the robot with these other tool and lesion technologies that we're using, so whether it's cone beam, augmented fluoroscopy, needle-based confocal laser, endomicroscopy, to kind of use that as a way to confirm that you're actually in a cancerous lesion. And then there's a couple other tools that I'll discuss a little bit more. The problem is, is there a value add there? Because a lot of these tools cost a lot of money, and reimbursement for these tools have not really changed. So I think a couple more challenges with robotic bronchoscopy is, by and large, we still rely on ENB and virtual platforms, and there's really no real-time confirmation of lesion sampling in the periphery. You know, you're taking your vision out, and you're putting your radial EBIS probe in, or you're using cone beam CT, but in real-time, we don't have that, and we're still using the same tools that we've kind of always used. So touching on the real-time guidance in the periphery, so this is a linear EBIS. We have real-time confirmation that we're going in and out of the node, but we just don't have that in the periphery. And one way that this is trying to be remedied is with the INOD catheter. So there's a kind of a schematic up there on the top, but you have your radial EBIS probe that kind of ejects out, and then you have your needle that comes through. And then you can, in using the radial EBIS view, you can see your needle going into the lesion in real-time. So this is one possible method of real-time confirmation that you're in the lesion, but you still have to find that concentric view, which, as we saw from the previous study, doesn't happen every time. So Lonnie and Alex and company did a pilot and feasibility study of this technology and placed targeted lesions, peripheral lesions, that were in all segments of the lung, and then used this INOD technology to kind of to determine whether or not this is a feasible tool. So lesion was confirmed in 24 out of 24 patients, and then the lesion, the needle was visualized in 24 out of 24 patients. So it's promising technology. It's still kind of being studied and adopted. The problem with it is it's $3,000 per catheter, and you just don't get reimbursed all that much. So for doing your CPT 31654. So it's kind of hard to justify this technology, although it is promising, and as it gets more developed, it'll probably become cheaper, hopefully. One of the other instruments that is being developed is a steerable needle. So you have a needle that goes in a conventional bronchoscope and I believe they're also developing it for the robotic bronchoscopes, but it has the ability to turn in 360 degrees, and there's a little bit of ability to flex and extend in it. And the goal would be to help, and when you get out to that lesion and you don't necessarily have a concentric view, you can target and move it just a little bit in order to make those fine movements, so you can really get in the center of the lung, or the center of the nodule, I'm sorry. So this is an example of it under fluoroscopy, and you can see the needle bending on that. See if it'll recycle there. Yeah, so you can see how it's shaping and moving around, which certainly can be useful for some of those lesions where you don't have a concentric view on radial ebus. So again, Alex and company just published this trial, actually, and I believe Momin was involved with this, called the Bullseye Study. It's a little bit of a jump to call it, to make that acronym Bullseye from that. So they put simulated targets in cadavers that were in all different areas of the lung, and the whole goal was to see whether the steerable needle could get out and biopsy the nodule. So 15 targets that were between 10 and 30 millimeters were placed pretty much throughout the lung, and you can see up here kind of the distribution and what they looked like. So the way that the study was done was using 2D fluoroscopy and navigating out with a conventional bronchoscope, the trained bronchoscopist through the needle, and then cone beam CT was spun to see where the needle was in space, and then a fiducial marker was placed. If the needle was in the lesion, and then they redirected the needle and tried to, and put another fiducial marker in to see if they could redirect from the original goal. So with one fiducial marker in this cadaveric study, the fiducial marker was in place in 93.3%, and then the second marker was in place in 80%. There's an example of the fiducial marker sitting in the lesion there. So methods versus procedures to improve diagnostic yield. So there's a study that is under development. Some of the preliminary data has been published, and I'll get to it on the next slide, but looking at thin and ultra-thin bronchoscopy, so taking kind of the robotic bronchoscopy out, you drive out to the lesion. If you have a concentric radial EBIS view, you take a TBNA. If you have eccentric, then you convert to the ultra-thin and see if you can improve your diagnostic yield. So again, as we've seen time and time again, our diagnostic yield sits somewhere around 70%. There were five cases that were converted after an eccentric view, and all of these were diagnostic with the ultra-thin bronchoscopy. So you can see it's still a pretty small study, only 25 patients, or 25 nodules, I'm sorry, but using these new different technologies still has our diagnostic yield at 72%. So in terms of kind of tool and lesion confirmation, ConeBeamCT has kind of been the new kid on the block that has been used, is starting to get increased use. The 3D fluoroscopy uses your, one of the platforms uses your existing fluoroscopy, 2D fluoroscopy, and kind of augments it and tries to integrate where the nodule is. and these are both kind of being developed and being studied, but we don't have great trials or any trials that are comparing these head to head. So while it's important to try to confirm that you have tool-in lesion, especially as we start thinking about diagnosing and treating these lesions in the same procedure, we still don't really know if this stuff helps our diagnostic yield. And then these are some of the various modalities. So this is your typical CT scan, and then this is a needle-in lesion on CT scan that you can see in a couple of different planes there. And then here's our radial EBIS view, and then this is your 2D fluoro view. So with that radial EBIS view, you could argue maybe you don't need to spin and confirm that your tool is in lesion, but these technologies have shown some promise, but they're still under study and we don't know what the appropriate amount of spins per case is and things of that nature. So I'll wrap up by saying the majority of all of this technology is aimed at improving diagnostic yield adjacent to peripheral airways. We do a pretty good job when we have a bronchus sign, but we really need to improve it, finding these lesions that we don't have a bronchus sign. And despite all of this stuff, the continuing reporting of diagnostic yield shows nothing better than about 80%. And so all of these things need more study, which is not the best way to end the talk about peripheral bronchoscopy, but as we move into these newer technologies, we're gonna have to define better ways to study them. Thank you. So we lost, oh, Dr. Vitani, your talk is back. All right, good morning, everybody. Thanks for coming to this late session. Moomin, for whatever reason, asked me to talk about three technologies and more within 15 minutes, but so given that challenge, these are my disclosures. To be clear, I will actually mention a technology biotelum and I'll discuss some biodesics data on their biomarker, so just to be clear on that. And so based on the amount of time we have, here's a very broad overview of the technologies that are emerging. This is not comprehensive in any fashion, but I think these are the emerging tests that are gaining some traction, have some data, and worth reviewing. So we'll talk about NotifyXL2 as a part of really a suite of tests now being developed by Biodesics, but we'll focus on that test for today. We'll talk about the PerceptoNasal Brush briefly. I'm gonna skip the Bronchial Brush because that's been around for a little while now and not necessarily new, but it is a partner test related to the technology developed in the airway during bronchoscopy. And we'll talk about two emerging tests in the imaging AI space. So the first one is the Biodesics XL2 test. I'm sorry, let me make one other point. I'm gonna very quickly just put this into the paradigm of are they at verification validation, clinical validation, or have they gone to utility? And of course what comes after utility is in most circumstances, if the data looks good, is guideline acceptance. We recognize that the chest guidelines haven't been updated for about 10 years, and these technologies have really emerged in the last 10 years. And so I recognize as a practicing clinician myself, I struggle with are the data quite there and ready to adopt, and what do the experts say? And I would love to know what the experts would tell me to do in this situation, but I hope that the guidelines will address some of these tests in the next iteration, which will hopefully come out, fingers crossed, Frank, 2024, so we'll see. So XL2 is a blood-based test that combines two proteins, it's up in the far right corner there, two proteins with five clinical factors to determine the likelihood of cancer of a baseline lung nodule. This is not a test for a lung nodule that's been observed over time, or that you definitely know is new from an old scan. This is like a classic baseline CT, you have a nodule, you don't know if it's new or old, and you don't know anything about its growth. In this study, and I'm gonna skip through the details, but in the clinical validation study, which is shown here, published in CHESS a few years ago now, the prevalence of cancer in a patient population that the clinician determined to have a pretest probability of 50% or less, this is gonna be a mouthful, 50% or less, and had a nodule of eight to 30 millimeters in size on a baseline CT was 16%, 29 cancers out of 178 total patients, so this was their validation cohort. The results are shown here, and I won't go through this in great detail, but the integrated classifier, this classifier with two proteins and five clinical factors is the top line on these various ROC curves. It had the highest AUC relative to PET, the VA, and the Mayo model, and that ultimately forms the basis of their validation data. They, based on this, have an operating point where the specificity is 44%, the sensitivity is 97%, and a negative predictive value of 98%. Based on this data, it was estimated that if this test was used to change the post-test probability of a lung nodule having cancer, you could save on procedures in those with benign nodules, but you would risk a small number of malignant nodules being called benign and going on into watchful waiting. So what actually happened with this data? So the first clinical utility data was published by Michael Pritchett and colleagues just earlier this year, and so if you look at their data, and I'm not gonna go through the nuances of the trial design, but this is essentially those tested with the integrated classifier, 197, versus a retrospective cohort of control patients where the test wasn't available to them, matched by propensity score matching to essentially balance all the other risk factors for cancer here. As you can see here, the prevalence of cancer was about 18% in this population, and this is what, at least just in the benign patients, what the post-test probability looks like, right? The model, the Mayo model would say that the probability ranges from somewhere between three and 45%, but that the test in this benign population of 72 patients where the test result suggested it was likely benign can go down. But a couple of points to keep in mind, right? Only 72 patients had a benign result, but there was 162 benign patients, so those patients are still going to have an indeterminate result where a decision needs to be made by us on what to do. And the second part that the study didn't tell us about was the cancers. There's no data on what the cancers look like in terms of their test results. So this is a nice start to clinical utility data, but I would encourage the folks working on the utility data from this registry study to give us more, because I don't think that the guidelines are going to look at this data and say that it's complete enough to be able to say we can adopt it. That's my prediction. I'm not on the guideline panel. We'll see what those folks say. But I do think that we need to learn a little bit more about this. This is a commercially available test, paid for by Medicare and most commercial insurers. So, you know, I think that if you are interested and want to use this in your practice, just make sure you learn about what it is doing and what it's perhaps not doing. So we'll jump through the rest of that data. The nasal classifier from Verisight. Many of you have heard about the bronchial classifier that I did about previously. This is just coming back up into the airway at a higher point in the nose. The idea being that the changes that we see anywhere along the airway tree are potentially reflective of the risk of whether a nodule can be malignant or not with this airway hypothesis. I won't go through the classifier in any great detail. Here's some slides that I borrowed from Bill Bowman at Verisight. Multiple genes in the classifier. Essentially, the idea is to show that in a patient with a long nodule, brushing their nose and using this classifier can distinguish benign from malignant. In some very early clinical validation data shown by Carla Lamb at CHEST a couple of years ago now, these are the sensitivity and specificity numbers. I won't read them to you, but potentially a very high negative predictive value and a reasonable positive predictive value with two different cutoffs for this test. I can also reveal, I think, that the clinical validation study that shows this data in greater detail is in review. And I think I can also share that a clinical utility study where we would actually see how it works in practice. How many benign nodules does it save from procedures? How many malignant nodules might it miss? Or how many malignant nodules might it capture based on the two cutoffs is ongoing. The Nightingale study should be done enrolling in about six to 12 months and should have some early data. I would bet by CHEST next year would be my guess. So let's, what am I doing here? Five more minutes, yeah, that's like a lifetime. Let's talk about the two imaging studies, technologies that I mentioned. One is a technology by a company called Optelum. This is work that I've actually done with my colleague Roger Kim and others at my institution. Essentially, it takes a nodule and gives you a risk score based on validation data shown on the right. So I'm not gonna go through the prior validation studies, but for example, this nodule, if you were to click on it with your system, would give you a risk score of nine. And a risk score of nine, as you can see from the bars, would suggest a probability of malignancy of around 84% if you believe the validation data. And it's a little bit of a stretch, of course, because I'm not showing it to you, but let's just say that that is what the data and how the tool works. So in a study then that took that tool and presented it to 12 readers, six radiologists and six pulmonologists, they all each read 300 nodules that ranged in risk and said, well, what was your risk before? Or what is your risk now without the tool? And then gave them the LCP score and said, well, what is your risk now? And if you looked at the results from that, this is how it came out, that there was an improvement in their assessment of the nodule being malignant or benign, the yellow line, relative to without the risk score technology. Artificial study, not embedded in clinical practice. This is 12 random readers reading 300 random CTs divided equally in half. So we do certainly wanna see some additional data come forward. A secondary analysis of some of this data suggested that for those with benign nodules here on the left, that the decision and the decisions that are shown on the bottom axis are go from very conservative, like no follow-up to imaging to biopsy, changed ever so slightly towards more conservative management and for malignant nodules, changed ever so slightly to a more aggressive management, which is what we want. But the effects look a little small, and again, it was an artificial setting. So this also needs a little bit more utility data in a real-world setting, and there are some trials at Vanderbilt, at my own institution, that are actually embedding this tool in practice to see what happens with the risk score. And then the last one I think worth mentioning is this one from Harvard, which was published in JCO just this year. It's received a lot of press and interest. Sibyl, a validated deep learning model to predict future lung cancer risk from a single-load OCT. So this takes individuals who are screen-eligible, it was developed in an LST data, and just uses the baseline scan to predict what your six-year risk of lung cancer will be. And I won't go through the black box because I still don't always understand the black box of imaging AI, so we'll save that for another day when we have more time. But here are the results across NLST validation data and two external cohorts from MGH and a third center. The one-year risk of predicting cancer is quite good, the AUC at one year. Ranges from .86 to .94 across the three cohorts. Not surprising, right? If you have a suspicious nodule and the AI works, it can probably predict risk pretty well at one year. What they didn't tell us is how does it compare to clinician or lung RADs or anything else, so we'd like to see that going forward. But even at six-year risk across the two cohorts where they had it, an AUC of .75. So not terrible to be able to just use one data point in time with no clinical risk factors, just imaging data, you can predict the risk of cancer at six years across NLST and one other external cohort. So very promising and I think worth following because you'll probably, I would anticipate we'll hear much more about data from this technology in the coming years. And this is just a couple of examples from Sibyl, right? So you can see where at baseline, you can see these very small nodules that might have almost been missed by radiologists and they don't really look all that concerning, perhaps, and that the technology predicted these as being cancers and they all ultimately were, right? These are sort of four cherry-picked examples, of course, but you can sort of see the promise of what AI can bring to our practice. And with that, I will end and say thank you. I'm sorry, I will just say this. Without the guidelines telling you how to use these technologies and they're already available to you, I find it to be the hardest thing I do in my practice. Do I use this test that's now available to me, it's paid for by insurance, some patients are even starting to ask about these when they come into clinic, but I have no guideline to guide me. And so I'd like for you all to think about, well, what is it that CHESS, beyond updating the guidelines, how can we support, Frank and I just had a conversation about this an hour ago, how can we support clinicians beyond just writing a guideline to help you think about how to use these new emerging technologies? So what can we do to support the membership? Because we'd love to hear more back from you to be able to understand where I think the society needs to go in this regard. Thank you. Thank you. All right, it is my pleasure to introduce Dr. Patricia Rivera, who really needs no introduction. She's the Chief of Pulmonary at University of Rochester. She's the current ATS President, and she's gonna talk to us about lung cancer screening and tackling barriers. Thank you very much. I've got, I'm gonna do this quickly so we can get some questions. These are my disclosures. I have no financial relationships to disclose. It's hard to believe that it's been about 50 years since the first lung cancer screening trials were conducted, three in the United States, one in Europe, two showed improvement in survival, but none showed mortality reduction. It took 40 years for us to have the results of the National Lung Screening Trial, the NLST, the first trial ever to demonstrate reduction in lung cancer mortality in individuals undergoing lung cancer screening with a low-dose CT compared to a single-view chest X-ray. It's been a decade since the USPSTF approved lung cancer screening in the United States with the letter B, and in the last two years, the recommendations for screening have been upgraded by the USPSTF. This is in response to data from the general population about differences in risk, as well as data from additional screening trials, including the Nelson trial in Europe, which enrolled younger patients with lower smoking history, and again, demonstrated a reduction in lung cancer mortality. The 2021 updated USPSTF recommendations to screen individuals age 50 to 80 with a 20-pack year smoking history or greater, and if quit, quit within 15 years, was very important because it allowed us to expand eligibility to high-risk individuals in the population that were previously excluded. That change in the recommendations increased the number of individuals in the United States eligible to screening by 81%, and it is reported that this change will significantly improve eligibility among black, Latinos, and women. We will pick up more screen-detected lung cancers, and we will avert more cancer deaths, particularly in women. Despite all the benefits that we know exist in terms of lung cancer mortality reduction with lung cancer screening, I think it's important to recognize that the USPSTF added these comments. There's uncertainty about the relevance of screening in the general population. We don't know if it's going to translate to the benefits that we're seeing in the screen population in clinical trials because they tend to be younger and healthier. The individuals who did the analysis for the USPSTF, this was a group at the University of North Carolina where I was formerly at, stated that populations eligible for lung cancer screening may be less likely to benefit from lung cancer screening because they face more comorbidities, they're older, and they may have competing causes of death. So there's this, we believe in screening, letter B recommendation, but there's caution. We don't know about how applicable this is in the general population, and this is not just a struggle for us in the United States. This is what Europeans are also facing after the Nelson trial, which demonstrated significant mortality reduction, more so than the NLSG. European Union and across various countries have been trying to establish more defined screening policies. This is a meta-analysis that looked at all of the screening trials that compared low-dose CT to another modality, in many cases, chest X-ray, or screening to no screening. And it was an intention to show, what would these trials say in a meta-analysis? Well, lung cancer screening decreases lung cancer mortality. It does save lives. We can't be arguing any more about that. Lung cancer screening, like mammography screening and cervical cancer and colon cancer screening saves lives. And why do we see this mortality a benefit? Because we diagnose more stage one lung cancers, or early stage cancers, where we have more that we can do to improve long-term survival and decrease mortality. And at the same time, the number of advanced lung cancers is decreased. So lung cancer screening results in stage shift. That's been argued for years. But the studies that have demonstrated this all show more early stage, less stage. So let's get it out there. Lung cancer screening saves lives. Lung cancer screening detects cancer at an early stage. Lung cancer screening results in less advanced cancers being diagnosed. But that's just one aspect of this story. There's a lot more. In order for us to really realize the benefits of this modality, a modality that took 40 years, 40 years we waited for a letter B recommendation for lung cancer screening, we have to screen the eligible population. And we can have a whole discussion on eligibility. I didn't include it in this talk because I would be verklempt. I could talk about it for hours. But we have to screen at least those that meet USPSCF criteria. We have to ensure that individuals who get screened come back for annual screening. And if they have a positive screen, they come back for the appropriate follow-up for that positive screen. And we have to ensure that individuals are enrolled in tobacco treatment modalities. Tobacco cessation is the most important intervention to decrease lung cancer mortality, period. But that's it. You combine tobacco treatment and tobacco cessation with lung cancer screening, and retrospective analysis of the NLSD demonstrated that you double the mortality benefit. But let's not forget, if you're screening individuals, you've gotta be talking about tobacco treatment and helping individuals get treatment for that addiction. There are many factors, and I listed some of them, that really go into play and that influence if and how individuals are traversing the lung cancer screening continuum. Because it is a continuum. It's eligibility assessment, it's getting the low-dose CT, it's getting follow-up annually, it's getting follow-up, it's getting enrollment in tobacco treatment. It isn't just, here, get the test. It is truly a continuum. This is work that several of us did as part of an ATS proposal to address barriers in lung cancer screening. And the problem here, it's not a problem, the reality is, is that we're talking about multi-level barriers. We have to remember that they exist at the patient-level stigma. They're afraid, they don't trust us, they don't know about it, no one's told them, they don't have access. It's at the clinician-level bias about knowledge, lack of knowledge among primary care clinicians about the benefits of screening, and importantly, concerns that the risks outweigh the benefits, and geographic location and access to screening programs. We talk about racial and ethnic factors impacting disparities in healthcare. Cancer care is plagued with the most significant disparities, particularly lung cancer, across all cancers. But it's the socioeconomic factors that are much more important. You eliminate socioeconomic factors, and you can reduce lung cancer mortality significantly in high-risk, vulnerable populations. So we have to remember that it's more than just race and ethnicity. There is probably, in black men have the highest risk of lung cancer, and the highest risk of dying from lung cancer. There are differences to how people metabolize the tobacco carcinogens. But socioeconomic factors are critically important, and they are the most important factors to define disparities, and access to care, by far, are one of the most important factors within the socioeconomic parameters. This is data from the American Lung Association from 2022. This is lung cancer screening rates across the United States. We are screening 5.8% of eligible individuals. 10 years after we first got approval for the one test that probably saves more lives than mammography and colon cancer screening, that is more efficient than mammography and colon cancer screening, 5.8%. Some states are doing a little bit better. Woo-hoo, we're at 6% in New York State. Some states, like California, holy Toledo, what's happening, right? So this is a problem. We're not screening people who need to be screened. And when you look at this, it's just one of many factors, I'm hurrying, the factors that impact who gets screened, sociodemographic. For screening to work, people have to come back annually, period. If you don't come back, the benefits are screening. This is a meta-analysis that looked at multiple studies that reported on screening for annual follow-up and after a positive CT. The pooled rate is low, it's low. It's only 55%, and when you extract the papers that only look at annual adherence at 12 months, 30% of individuals are coming back. That is not acceptable. Multiple groups, including our research group in North Carolina, we've looked at adherence to follow-up after a positive CT. It gets better as the higher lung rats, it gets better if you extend the time. Okay, the lung rats has come back at six months, but it's okay, you can go back at 12 months. And oh look, we're doing so much better. But for the first time, we have data from the real world that this is not okay. The thoracic surgeons have been saying this over the course of many years. Don't delay surgery for lung cancer, even early stage lung cancer, because delays in follow-up do result in upstaging. That is not what lung cancer screening is supposed to do. Who doesn't come back for follow-up? Again, black individuals and individuals who are currently smoking. There are multiple strategies, and again, because the barriers are multilevel, the strategies have to be at multilevels, and they probably need to be deployed at the same time. These are strategies that are currently being investigated at our institution. We are doing studies looking at the impact of community outreach, telehealth, and mobile technologies. But I just want to make a pitch for centralized lung cancer screening. And what does that mean? It means a group of individuals who have expertise, not only across the lung cancer screening continuum, but across the lung cancer continuum, so that you can identify nodules that are suspicious, you can keep track of nodules, when does this patient, you can make the best decisions about diagnostic procedures for positive lung cancer screening CTs, and importantly, involving the primary care physicians, because incidental findings are not insignificant, they're important, and we need to partner with our colleagues. The process can be really, really simple. We've been doing this for how many years now, multidisciplinary lung cancer care? I mean, since what, 1995? We've shown that it works, we've been doing it. We can do multidisciplinary, centralized lung cancer screening across community hospitals and academic. Why is it important? We have shown that centralized approach improves the rates of annual screening, and others have shown that a centralized approach improves the adherence rate, particularly after a positive CT among vulnerable populations. So in summary, it works. Let's not be talking about anything else. We know who we need to screen, there are a lot of barriers, there are multiple level barriers, and it's going to require a great deal of effort in order to improve increasing the rates, and importantly, adherence to annual screening and follow up after a positive CT. Thank you.

lung cancer screening

low-dose CT

mortality reduction

barriers

screening rates

follow-up adherence

USPSTF recommendations

comorbidities

socioeconomic factors