Good morning and thanks for attending our session today. So we're going to talk about an important topic for most of us pulmonologists, oncologists, radiation oncologists, and radiologists, which is lung cancer screening. I think we all realize the importance of lung cancer screening and we know that we are facing a lot of difficulties. So some of our speakers today will discuss different aspects of lung cancer screening. We will have eight minutes for each presenter. So please, I will let you know when there is one minute left. We are limited when it comes to timing. So at eight minutes, we will have to cut you off if you're not done. But I hope all of you stick to eight minutes. We'll give two minutes to questions after. So the first speaker is Dr. Haley Tupper and she will talk to us about, does preoperative mediastinal lymph node staging affect early stage non-small lung cancer management? Okay, hello and thank you to CHEST for the opportunity to present our work. So I'm Haley Tupper. I'm a general surgery resident at UCLA and I have no disclosures. So the purpose of, I'm here to share a study that looks at how the receipt or non-receipt of invasive mediastinal staging changes their treatment management. Guidelines recommend invasive mediastinal staging, a test with variable sensitivity in a large proportion of early stage lung cancer patients with a relatively low pretest probability of having mediastinal or N2 disease. So the purpose of this talk is to challenge the dogma that all early stage patients necessarily benefit from invasive mediastinal staging, particularly with regards to how these results change or do not change their treatment management. So early stage lung cancer is increasing. We have an aging population and they're increasing expansion of lung cancer screening access and use of diagnostic CT with incidental findings. In 2019, 38% of non-small cell lung cancer was stages one or two. Most surgical candidates have an indication for invasive mediastinal staging. However, less than half of these patients with an indication actually undergo the recommended invasive staging. So procedural risks are low, but they are not negligible and the patient is still subjected to delays in care and increased anesthesia time. Meanwhile, surgical resection pathology provides the most accurate staging information. So what are the guidelines? So for small peripheral cancers that are N0 on CT and PET CT, there's general consensus that invasive mediastinal staging is optional in these patients. And for patients that have N1 or N2 disease on any imaging, it is clearly recommended. It's also recommended for patients that have tumors that are more than three centimeters or central. However, I think this is actually more of a gray area than we acknowledge, especially when less than half of providers are following invasive staging guidelines and they're using more of their clinical gestalt. So in light of this, the aim of our study was to characterize the association of clinical to pathologic stage change with receipt of invasive mediastinal staging in early stage lung cancer. Our underlying hypothesis here is that routine invasive mediastinal lymph node staging rarely results in a treatment-relevant stage change. So this is a multicenter retrospective cohort study of 2,577 early stage non-small cell lung cancer patients with preoperative imaging who underwent surgical resection with curative intent. Some of these patients underwent preoperative invasive nodal staging, either EBIS or mediastinoscopy. Who gets preoperative invasive nodal staging in our health system is a combination of radiographic indications that are concordant with NCCN guidelines as well as a heavy dose of clinical gestalt. So our primary outcomes of interest were the clinical to pathologic stage change, both overall TNM stage and nodal stage change, as well as the stage change relevance for treatment. For patients that were upstaged to less than 3A, we consider this to be irrelevant in the sense that surgical resection was appropriate in these patients and they can still receive adjuvant therapy based on their pathologic staging. For patients that were upstaged to 3 or higher though, this was an upstage that was considered to be relevant because upfront surgical resection was not necessarily appropriate in these patients. This was a data-only study and we performed descriptive and association analysis and we performed a sub-analysis of patients that were clinical 1B or higher. So when we look at the results for all early stage patients, we see that 18.7% of them underwent invasive nodal staging. There was differences between who underwent staging and who did not. In particular, there was more squamous cell histology, possibly representing more central tumors and higher stages in those that underwent invasive staging. On this slide, I also want to highlight the 522 patients who were 1B or higher who did not undergo invasive staging despite recommended guidelines because these will form the basis of our sub-analysis later. So we look at the outcomes for all early stages. We do see that there was a difference for patients that received preoperative nodal staging in the sense that they were both more likely to be upstaged and downstaged. Even though almost a quarter of patients were upstaged in a relevant manner or upstaged at all, only 0.4% were upstaged in a relevant manner, meaning 3A or higher. And when we look closer at this, we see that there's no difference in relevant upstaging on the receipt or non-receipt of preoperative nodal staging. So in our sub-analysis of 1B, we did this because there's a subset of 1A patients for whom invasive staging is already optional. Furthermore, 1A patients are more likely to undergo wedge resection with the potential for a decreased surgical lymph node count. Consequently, we wanted to look at the 1B plus patients for whom guidelines currently recommend invasive staging. So we still saw that these patients were different, but that they were more balanced between groups, and we did see more squamous cell histology in higher stages in those who underwent invasive staging. However, 66.8% of our 1B plus patients did not undergo invasive preoperative nodal staging. And when we look at the outcomes for the clinical 1B plus, we see actually that there was no difference in TNM or end-stage change based on preoperative invasive nodal staging. So they were both upstaged and downstaged in relatively equal proportions in both groups. And only 0.8% of all patients were upstaged in a relevant manner, meaning to 3A or higher. And again, we see no difference in relevant upstaging based on the receipt of preoperative nodal staging. So in summary, even though 81.3% of early stage lung cancer patients did not undergo preoperative invasive mediastinal sampling, only 0.4% were upstaged in a potentially treatment-relevant manner, meaning pathologic stage 3A or higher. And for clinical 1B plus, there was no association between the absence of preoperative invasive mediastinal staging and subsequent upstaging, even though in these patients guidelines currently recommend invasive staging. So this data challenges the notion that preoperative nodal staging is routinely necessary for appropriate early stage lung cancer management. The take-home message here is that preoperative invasive mediastinal nodal staging may not change the management of most patients with early stage lung cancer. And existing guidelines may over-recommend invasive mediastinal staging in this population. Thank you. So next speaker is Dr. Lambert, going to talk to us about identification of individuals for lung cancer screening using geospatial analysis. Good morning, everyone. So my name is Christine Lambert. I'm from the University of Minnesota. And today I'm going to be talking about identification of individuals for lung cancer screening using geospatial analysis. I have no financial disclosures. So I think we all know that lung cancer screening is a way that we can decrease lung cancer mortality. But unfortunately, the uptake of lung cancer screening has not been as robust as we hoped for. And there are sometimes challenges in identifying individuals for screening, trying to get a sense of the individual smoking histories. And so what we wanted to do with this project is look at the population level. If we combine smoking information with demographic information, can we get a sense of the number and distribution of individuals that we predict would be eligible for lung cancer screening and use that to try to guide some of our efforts at increasing lung cancer screening uptake through directing resources and outreach to the best populations. So we used several sources of data for this study. We took smoking prevalence information from the BRFSS, the Behavioral Risk Factor Surveillance Survey, combined that with demographic information from the American Community Survey, and then ultimately used the Smoking History Generator from CISNET to get a sense of PAC years for individuals which allowed us to get a sense of then who would be eligible for lung cancer screening based on those PAC years criteria. An additional aspect of this project was to look at the travel distance between individuals eligible for screening to what we would anticipate would be their closest screening site. And this was done by taking the American College of Radiology's Lung Cancer Screening Registry as of 2021 and modeling the closest facility for individuals, breaking it down into time periods. And this is work that built on prior studies along these lines, but this time trying to expand it with our population of former smokers. So getting into some of our results, this table shows our anticipated population of people that we predict would be eligible for lung cancer screening. And this table breaks down the smoking status by racial ethnic group. And so ultimately what you can see is we anticipate that there would be a little over 29 million individuals that would be eligible for lung cancer screening. And you can see the largest, both in terms of raw numbers and percentage, are white individuals. So we're narrowing that in terms of numbers, African Americans. I think you can also see that there's certainly a large percentage of American Indian and Alaska Native individuals that we anticipate would be eligible for screening, likely representing the very high smoking prevalence in that population. This is another table looking at smoking status, this time by gender. You can see among current smokers there's fairly equal numbers of men and women that we would anticipate would be eligible for screening. Once you get into former smokers, it becomes a higher number of men, possibly representing sort of the lag in terms of women starting to smoke and then quitting smoking. This is a map that shows the distribution of people that we predict would be eligible for screening. The red areas are the areas with the highest numbers of individuals. At the census track level, I think we all can kind of look at the area of the country that we're most familiar with and sort of try to make sense of what we see. I think there's a combination of areas, high-density population areas that show up red, but there are also areas of the country where there are not a lot of people and not a high population density that still have a lot of people eligible for lung cancer screening. This is a table showing the travel distance that we predicted for individuals. About 62% of people are within about 10 minutes from a screening facility, but there were about 15% of individuals that would have to drive at least 20 minutes to get to a screening facility. And then this table graph sort of overlays the prior map that I showed you, this time with travel distance. So the gray areas are areas where there's both a smaller number of people eligible for screening and they are very close to a screening facility. And then the problem areas are the sort of brownish areas. Those are areas where there's a lot of people that we predict would be eligible for screening, but they're also the farthest distance away from a screening site. So in conclusion, we used geographic analysis to combine demographic and smoking information to try to estimate the location and characteristics of people who would be eligible for lung cancer screening. We predicted that there are about 29 million here in the U.S. that would be eligible for screening based on the USPS guidelines. And there are, as I think we saw from the maps, a large number of variations by state in terms of the travel distance or some of the characteristics of eligible populations. And so our hope is that this work could help give people in healthcare systems or public health groups information that could be helpful as we try to plan out programs for mobile screening, outreach and education to increase the number of people aware of lung cancer screening and potentially incorporate other aspects of preventative healthcare services like smoking cessation. So I want to acknowledge my collaborators at the University of Minnesota, thank our collaborators at the University of Michigan for helping us with the smoking history generator, and then thank you to CHESS for inviting me to speak. All right. Our next speaker is Dr. Andrea Burnett Hartman, and she's going to talk to us about using electronic health record data to calculate lung cancer risk and inform lung cancer screening. So I'm Andrea Burnett Hartman, I'm an investigator at Kaiser Permanente Colorado, but I'm here presenting on a much larger group. And so I'm representing the Prosper Lung Consortium here, and this includes sites all the way from Hawaii to Pennsylvania. So we have Kaiser Permanente Hawaii, Kaiser Permanente Colorado, Marshfield Clinic in Wisconsin, Henry Ford in Detroit, and then University of Pennsylvania Health System within our consortium here. And so just to give a couple disclosures, our consortium is funded by the National Cancer Institute, and then one of our co-authors has grants from some industry sponsors as well as some honoraria and some paid travel for that as well. So I probably don't have to give too much background to this room, but with the U.S. Preventive Services Task Force recommendations there are certain criteria laid out on the screen where people are eligible for lung cancer screening. And as we look at these criteria and we study the populations that are eligible, we have found over time that there's both inefficiencies and inequity in who is eligible for lung cancer screening. So some people have thought about using some prediction models that have been available prior to even the guidelines that are out there. So one of these models is the PLCO model. And that model predicts 6-year risk of lung cancer within populations. And so this model works quite well, but it requires collection of a lot of different variables. And so people have looked at this using survey data to say, we can predict 6-year risk of lung cancer using a combination of characteristics rather than just strict criteria as laid out with the U.S. Preventive Services Task Force. So what we tried to look at was people have used survey data or self-report. Can we use electronic health record data that is, quote, laying around within our systems? So across our five different systems, are we able to kind of take out the barrier of collecting this separately through survey and use what we know is semi-dirty electronic health record data to do the same thing? And how well can we predict 6-year risk when we're actually using electronic health record data? So this slide shows our study population, how we select it. You had to receive care within one of the health systems that I mentioned on the first slide. We restricted to the 50- to 80-year-old range, similar to the current guidelines. You had to be an ever smoker, have ever smoked cigarettes, partly because that's the way this model was created. It's supposed to be used within people who have ever smoked cigarettes. And then as we go down the line, I want to focus in on that last criteria. You had to have complete data. And that means all the variables in the model, and I'm going to show that on the next slide. We had to be able to assess that using your electronic health record data. You can see that we lose about 70% of the population at this step. And so that's one of the barriers. I'm going to talk about that in a second. So these are the predictor variables that I mentioned along here. We used EHR data to collect these variables. We started in January of 2014 and then looked forward, connecting to our tumor registries that we have within each of our health systems to identify six-year risk of lung cancer, so people who actually developed lung cancer within those six years. In our analysis we used Receiver Operating Curve characteristic analysis. And so that allowed us to estimate the performance. I'll show the results in a second. And then we looked at a 1.5% cutoff. And this is a cutoff that's been used in some other countries like Canada to say, you can apply this model using a 1.5% risk. So we applied that and then estimated some other performance characteristics. And then we compared what we call a modified PLCO-2012 versus the current U.S. Preventive Services Task Force. And it's modified because we took out education. So that's the one variable that's not on here that's hard to assess in the medical record. So this just shows a little bit of the characteristics of the population. We had, you know, a predominantly white population, but you can see we also had representation from different race and ethnic groups including about 16% non-Hispanic black populations. When you look at the health system, partly the difference is partly due to differences in screening uptake, differences in size of the health population, as well as differences in complete capture of data. So we lost more in health systems that had less capture of these data. Again this wasn't people who had ever smoked cigarettes. And so we see more COPD. We see more personal history of a prior cancer, a prior non-lung cancer in this case because we excluded those who had prior lung cancer within this population than you would see in a general population. And then this is what we found when we used the EHR data. I'm going to orient you a little bit to this. This is a receiver operating characteristic curve. So if you were in the top left corner and you had a value of 1, you perfectly predict whether or not someone will develop lung cancer within six years with this model. If you're on that diagonal line, it's like flipping a coin, right? So you can flip a coin and predict it. So we came up around 77%, which is actually pretty good for prediction using the electronic health record data to do six-year risk of lung cancer. And when we applied the 1.5% cutoff, we found that we had similar sensitivity, meaning about 70% of people who have lung cancer we would capture using this prediction model at 1.5%. And U.S. Providence Services Task Force is pretty similar, like .69, .68, 69%, 68%. But better specificity, which means our positive predictive value was actually better using the PLCO model with the 1.5% cutoff. We also had better accuracy. One of the problems when you think about risk modeling as well as applying criteria, these models have often been developed in white populations. And so we really want to see, is there differences when we start looking at different subgroups of the population and performance? And we did find some loss of performance in our black population. So I think there is some work to be done to increase performance across all populations within these models. And then I'm getting back to the 70% loss. So we said, you know, you have to have complete data in order to be in our primary analysis. But we did some imputation work to say, you know, can we borrow information across different groups and be able to impute? And this allowed us to use the entire population. We saw some loss in performance of our model. Remember it was .77 earlier in the AUC, the overall model performance, and now it's .71. But what this really does is allow you to estimate in the entire population as opposed to just that 30% or so with complete data. And so what we found was that you can use the EHR data for the PLCO model, that we actually were able to perform better than U.S. Preventive Services Task Force using that 1.5% cutoff. And what that means is that you could potentially screen a smaller subset of the population and not necessarily miss numbers of lung cancer. So similar numbers of lung cancer by screening a smaller subset of the population. And just as I said, I'm here on behalf of a large team. Not everyone was at even on the first slide. So just thanking all my collaborators on the project. So I'm Christina Spears. I am a radiation oncologist at Cancer Center of Hawaii and I'll be moderating the rest of the session. So the next session will be given by Dr. Anant Mohan, who will be discussing discriminating chronic respiratory diseases from lung cancer using exhaled breath signatures by ENOS, a pilot study. Thank you very much. And I am Anant Mohan. I am working at the pulmonary medicine department at the Alden Institute of Medical Sciences in New Delhi. The title is how to discriminate chronic respiratory diseases from lung cancer using exhaled breath signatures and ENOS. It's a pilot study because exhaled breath is still a relatively newer topic comparatively in respiratory medicine. And we had started this work a couple of years back and this is some of the preliminary data. Work is still ongoing. This study, this work was actually funded by the Indian Council of Medical Research as part of a Center for Advanced Research and Excellence in Exhaled Breath Research. So I think introduction, we all know that lung cancer is one of the commonest cancers with the highest mortality. In our part of the world, we have other factors and other diseases which really confound the diagnosis of lung cancer, especially tuberculosis and also the rising prevalence of COPD, which are also significant risk factors for lung cancer, thereby causing a delay in the diagnosis, which we have also previously documented. We really need non-invasive tools to discriminate lung cancer from the other mimicking disorders. And among these, some of the ones which can be tried and are being tried now are the volatile organic compounds, which can be measured using the electronic or the e-nose, or the non-volatile compounds, which can be done or tested by using micro RNA assays. So we'll be talking about the first one, the volatile organic compounds and the e-nose right now. And that was the objective of our work, to evaluate the utility of electronic e-nose to discriminate chronic respiratory diseases, other diseases, tuberculosis in this case from lung cancer. So here we took 80 healthy controls, we took 70 COPDs, 60 TB, and 113 lung cancer subjects. And we collected the exhaled breath in the special Teddler bags, which are meant for that purpose. I'll show you the figure in the next slide. Then they were analyzed by means of the e-nose, which was the Cyranose 320 in this situation, which was its kind of a patented brand, which we purchased. And then using dichotomous comparison of the VOC breath signatures, that this gives a typical breath signature. And then we tried to discriminate lung cancer from the other diseases. And we established a discriminating algorithm, and also an area under the curve to interpret the data. The collection of the sample was done like this. There were some instructions to be given, no coffee, no food for one hour before, or alcohol for at least 12 hours, avoid smoking. Then a bit of lung washout was given to remove the exogenous VOCs by three minutes of normal inhalation through a filter cartridge. Then followed by slow exhalation through the mouthpiece into the Teddler bag. And then this Teddler bag was connected to the electronic nose device, which you can see over here. So I think the cursor is not moving. So you see on the top left is the e-nose device. On the right side, Teddler bag and the other attachments. And then in the below right, you can see it is connected to the system. Analysis of data was on healthy controls on one side, disease on the other. We had two sets. 80 per cent comprised the training set first. Then we plotted a cluster plot, 20 per cent were the test set on which we applied the machine learning algorithm using a vector method, a supervised vector method. Then we used a confusion matrix and ROC curve to try and find out the sensitivity, specificity. And we did the same on the other side, on the disease side. The demographics of the people, healthy control, lung cancer, COPD and TB, as you can see obviously the TB people were relatively younger. These were the spirometry results, which again I think we'll not spend much time in that. There was nothing much to choose between lung cancer and COPD. A couple of radiological features here, the lung cancer, COPD, hyperinflation and TB, as you can see, a destroyed left-sided lung. These are the main results here, the VOC cluster analysis. This is in the training set of these subjects. And here we can see lung cancer versus healthy control. Again, the cursor is not moving, but I'll try to show you that on the left-most side, the discrimination between the healthy control, the healthy controls are the green, the lung cancer are the orange ones. So the cluster is there on the upper right, but on the other parts they are kind of discriminated. In the centre panel you can see versus COPD, they are really very much all clubbed together and clustered. So they are not getting discriminated from each other. And on the right-most side, again, from lung cancer to TB, again, there is a kind of high overlap and there is not too much of discrimination. So basically the discrimination of lung cancer from healthy controls was relatively better, but not so good when we compared with COPD and TB. The next step was in the test set of these subjects, which were 20%. Again we did the same thing, we did both the things. We did a confusion matrix to calculate sensitivity, et cetera, and an ROC curve. And you can see over here that the ROC curve, if you go on the right side, lung cancer was healthy control. The ROC curve did pretty good about discrimination with the tune of almost 83%. Lung cancer versus COPD was able to be discriminated by almost 0.93, so 93%. But it did not do very well comparing lung cancer with discriminating lung cancer from TB. And also you can see on that figure that lung cancer, TB are almost similar. But on the left side, both the panels, lung cancer versus healthy control and versus COPD is doing pretty, pretty OK. So the sensitivity and the specificity also of the lung cancer versus COPD, as you can see, sensitivity is around 90%, which also did pretty OK in this particular situation. So we summarized that, and this is still work in progress. So we are now going on for GC-MS also to try and find out which are the exact VOCs which are kind of upregulated or discriminating. So that result, I think, will be available in a couple of months from now. We'll be in a better position to tell you the exact VOCs which are discriminating. So to summarize, the ROC analysis of the VOCs does show the ability of eNOS to discriminate lung cancer from healthy controls and also from COPD with reasonably good accuracy, but not so between lung cancer and tuberculosis as such. So we concluded that in this pilot preliminary study that eNOS can discriminate lung cancer from healthy controls and from COPD using certain VOC signatures in the exhaled breath. The poor separation between lung cancer and TB needs to be investigated, probably it may be because of very similar breath signatures between both the diseases, and we will need more work in that particular field. These are a couple of our preliminary papers that we had related to this particular aspect. And this was our team and some people who were involved. Of course, there are some people, the analyzing team was different because there was a different set of collaborators to actually do the VOC analysis and the ROC and cluster analysis. And on the right side, this is our team, the clinical team, which is responsible for recruitment and collection of all the things. Thank you so much. Thank you. So our next speaker will be Dr. John-Michael Sweetenum, who will be talking about Two Birds With One Stone, a cross-registry analysis of women undergoing lung cancer and breast cancer screening. All right. My name is Connor Sweetenum. I'm a third-year fellow at the Medical University of South Carolina in Charleston, South Carolina, and I'm going to discuss cross-registry analysis of women undergoing lung and breast cancer screening. I have no disclosures. So the objectives I wanted to discuss today were to identify and describe the population of women who have undergone dual screening, in other words, who are both eligible and have undergone lung and breast screening, to demonstrate the characteristics of that dual-screened population, and to consider methods to leverage other successful cancer screening modalities, such as breast cancer, to improve lung cancer screening uptake and adherence. So we all know that lung cancer screening is effective and that it's particularly efficient compared to other forms of cancer screening. However, that lung cancer uptake is exquisitely poor compared to others. Further, we know that adherence to follow-up is low. So part of the mortality benefit conferred in the NLST trial depended on a one-year adherence of 95 percent. However, larger analyses of the cancer screen population have shown that adherence is more likely 22 to 55 percent, depending on how you look at it. Interestingly, though, lung cancer screening-eligible women do undergo other cancer screening modalities. So this graph is actually I made using data from this 2019 NHIS analysis, but if you look at the striped column that's in the box, these are lung cancer screening-eligible women who have undergone a combination of either breast and or colorectal cancer screening at a rate of around 70 percent, compared to the same group who have undergone lung cancer screening at less than 10 percent. And so we were curious to know what was unique about this population, why they weren't undergoing lung cancer screening, even though they were receptive to other preventative services. So we did do a pilot study in our statewide academic health care system in South Carolina, looking at women who were both eligible for lung cancer and breast screening, and we found that predictors of lower lung cancer screening uptake included former smoking status, younger age, black race, underinsurance, and distance from our main screening center. So in this study, we wanted to define and characterize a national cohort of female patients who were eligible for lung and breast screening and had, in fact, undergone dual screening. And we wanted to determine adherence rates in that population and identify other factors associated with dual screening. So our hypothesis was that dual screeners would display a higher rate of adherence than the general screened population. This was a retrospective cohort analysis using American College of Radiology data. So we looked at the Lung Cancer Screening Registry, as well as the National Mammography Database, and this has sort of been touched on earlier, but this includes over 2,000 U.S. cities that are participating in these registries. The population we analyzed were female patients aged 50 to 74. We used that age cutoff to make sure that everybody in the population was eligible for both breast and lung screening, and who underwent both lung and breast cancer screening with an initial exam between January 2014 to January 2020, cut off at 2020 so that we could look at adherence further down. And at this point, we've just done some descriptive statistics. So our initial group were females who were 50 to 74 that were dual eligible and had an initial lung cancer exam before January of 2020. That was 674,881 people. Then when we looked at who had had mammography, we actually lost 652,000 of them because some of them may truly not have had mammography. It's more likely that the vast majority simply didn't have the proper identification linkage between the two registries. But we were left with 22,629 patients who had, in fact, were eligible for both screening modalities and had undergone breast and lung screening. And then further breaking that down, 5,524 of them met annual adherence criteria and 17,105 did not. So who were the dual screeners who were both eligible and had undergone breast and lung screening regardless of adherence? So that was that 22,000 group. So there's some missingness here as a result of just the registry data. But you can see the preponderance were of white race at 47%. The largest insurance group was Medicare at 27%. 61% were current smokers and 52% were aged 55 to 64. And were they more likely to be adherent to lung cancer screening? So this is that group of people who, the 5,524 who met criteria for annual screening, which we've extended to 15 months as has been done in previous studies. That was about a quarter of the population or 24.4%. You compare that to the million screen study that was in Silvestri et al and Annals recently. That looked at the first million screened and the adherence there was 22.3%. So it's pretty similar. If you look down the columns, this percentage is actually the proportion within each group who were adherent. In other words, for example, of the people of white race who were dual screened, 23.9% were adherent. Of people of black race who were dual screened, 18.5% were adherent and so on. There really aren't any really notable differences going down the column. The education status has a tremendous amount of missing data. So I don't make too much of these slightly higher numbers. And 65 to 74 group was slightly higher at 29.7, which is of course the Medicare group. One of the more interesting findings we felt was the timing of screening exams. So if you look at this graph on the right, these are people who were dual screened who had lung cancer screening performed first. And then we looked at the number of days before they had breast cancer screening performed. So the x-axis is days before breast cancer screening. The y-axis is percentage of breast cancer screening performed. So about a third of these patients who had LCS performed first had BCS performed within 10 days. Whereas if you had BCS performed first, it took about six months before a third of that population had lung cancer screening. And then if you actually extend that out, probably we meet a majority around a year or so. So we thought that was a really interesting finding. So it seems that providers ordering lung cancer screening or ordering breast cancer screening at the same time, but not the other way around. So that would clearly be a place we could try to intervene in terms of providers who have breast cancer screening as top of mind, but not lung cancer screening, to try to encourage them to take a more holistic view of cancer screening in general. Also somewhat surprisingly, to me at least, was that although these people were engaged in preventative screening behaviors and were dual screened, that didn't seem to be a predictor for adherence. Or at least their percentage of adherence wasn't higher. And we need future work to understand how to leverage that dual screening to improve uptake and adherence. And it seems overall that barriers to adherence in the dual screen population are likely similar to those in the single screen population. So ultimately, we found that over a third of the LCS exams performed within 10 days of BCS. Over a third of LCS exams are performed within 10 days of BCS when BCS is performed first. Adherence rates were similar in this dual screened population versus the all comers in the LCS screen population. And it seems like factors associated with poor uptake and adherence in the general lung cancer screening population are similar to those in this dual screened population. So I'd like to thank my research mentors, Dr. Tanner, Dr. Silvestri, as well as our collaborators at the American College of Radiology, Lincoln Goldman and Lars Grimm. For our final speaker, Dr. Chinmayi Jani will talk about CALM, coordinating a lung screening with mammogram. To continue with our presentations from the morning, my topic is coordinating a lung cancer screening with mammography. It was an ACS and National Lung Cancer Roundtable supported study carried out at Vanderbilt University Medical Center and Mount Auburn Hospital at Harvard Medical School. I'm a former chief resident over there where we conducted the study and currently a clinical fellow at University of Miami Sylvester Comprehensive Cancer Center. I do not have any personal disclosures. So objectives, these are my SMART objectives of this presentation. I hope by the end of the presentations all of you will be able to assess the effectiveness of our strategy aimed at increasing lung cancer screening uptake among eligible women who participated in mammography programs, as well as you'll be able to assess the feasibility and long-term validity of the proposed outreach and notification system for identifying LCS women participants, eligible LCS women participants. As we all know, lung cancer is a leading cause of mortality, not only in men but also in women. Multiple different trials have shown that lung cancer screening decreases lung cancer mortality. However, uptake has remained low, especially for women as well. There has been clear evidence of benefits, USPSDF guidelines, lots of scientific societies acknowledging and recommending, despite that this uptake has remained significantly low. However, at the same time, we saw in various different presentations today as well, the breast cancer screening population is really well acknowledged and it has been profoundly impactful in increasing the uptake as well. Overall, it has been well accepted in the society too. Therefore, our idea was to use this current data set and infrastructure and increase and try to increase the lung cancer screening population in both these centers. So our study aims is to identify mammography program participants eligible for and not yet enrolled in the LCS. Our goal was to measure the efficacy of outreach and synchronous interventions to improve enrollment rates of women eligible for LCS by leveraging the existing infrastructure of breast cancer screening as well as eventually, our hope is to pair it on the same day. And to evaluate the increasing LCS uptake by implementation of the centralized as well as decentralized screening programs. Our methodology was mammography participants from November 2019 till December 2022 at Vanderbilt and Mount Auburn were enrolled in the study. Three-prong approach was used and we'll go into details in a few slides, EMR reports, PCP outreach as well as patient surveys. We excluded all the patients who had metastatic cancer who have already had lung cancer screening or were unable to give consent. Eventually, outreach was performed to all these eligible patients as well as to the primary care providers to first confirm their eligibility and to offer lung cancer screening. For our eligibility criteria, we used the updated USPSTF guidelines with 50 to 80 age, current smoker greater than 20 pack care smoking history and former smoker who quit within 15 years. So this is one most important slide of our entire project, three-prong approach as I mentioned. The first one was EPIC report. So we did EPIC report and slicer-dicer analysis to identify these patient populations. A lot of those patients did not have accurate documentation. Second was we did PCP outreach, so we sent multiple emails as well as flyers to all the PCP sites in our network so that they can identify the patients in their upcoming visits as well as in their own database of the patients. And the third was to create awareness. So this was one example of the flyer used at Mount Auburn. This was kept at all the breast cancer screening sites, the mammography sites at the radiology department and they were handed over the surveys to fill it up whether they are eligible by themselves or not. Based on this, patients were identified. Once those patients were identified, we did an outreach to all those patients either by phone call or by email. Once they were identified that they should be enrolled, shared decision-making was done and they were offered LCS. So our results, as we can see, total 32,165 patients were in our study who had mammography out of which 1,516 and almost 5% patients were LCS eligible. However, almost 70% had not undergone LCS. These are our results. A little bit busy slide but you can see over here for females at baseline, the increase in uptake was from 11 per month to 18 per month at Mount as well as in Vanderbilt it was up from 28 per month to 37 per month. We used a natural population of male LCS as a control cohort study and in this, you can see that for male population, the LCS uptake was not increased. However, overall increase was also there probably driven because of the female increase in LCS. So a significant increase in LCS exams was seen for women who observed during the intervention period at both institutions and as I mentioned, at the control group, the male population did not see a similar increase as they were not part of our overall clinical trial. This is a graphical pictorial presentation and over here, clear evidence about an increase in uptake during the calm period at Mount Auburn Hospital. So going to some discussion points, this is for our overall project. Our main focus was on increasing lung cancer screening but we are also trying to identify various different aspects on why this increase is not happening or why patients don't want to have lung cancer screening. First and foremost, patient-associated causes. This was identified through some of the qualitative data that we collected on outreach. First and foremost was patients were not aware about lung cancer screening. There was still misconception about lung cancer is equivalent to death. Lots of patients are told that they are not interested in knowing whether they have lung cancer or not. And the second thing was, this was a system implementation. Again a previous speaker had identified in their EPIC reporting as well and the similar results were in our EPIC reporting system as well. So if you can see particularly for former smokers or current smokers, only 35% had accurate documentation of back years. And similarly for quit years, only 60 to 62% had accurate documentation. So we are having, we are taking further steps to increase this documentation accuracy in our hospital systems as well through various different modifications along with nationwide modification EPIC reporting as well. And along with that, increasing PCP awareness just with regular flyers, regular emails, or putting all this information of lung cancer screening at different places. So as a conclusion, one thing I would like to identify, the two circles which I have kept as just a nomination about breast cancer screening and lung cancer screening. Breast cancer screening population database is really big enough and lots of patients are already taking it without any restrictions or without any difficulty. Using that database and using that infrastructure, we are trying to expand the lung cancer screening population as well. So through this trial, we were able to significantly increase this patient population, particularly for women, for increased LCS uptake. An interesting study from US Department of Labor showed that women almost make 80% of healthcare decisions in the United States, whether they are part of the family or whether they are patients themselves. And therefore, even if our study was not focused on men and lung cancer screening, since women make the majority of healthcare decisions in these families through evidence, this intervention has the potential to increase LCS by creating awareness in their families amongst both men and women. This is our acknowledgment, my acknowledgment. These are all my co-authors. A special thanks to all my mentors as well, Dr. Kerry Thompson. And with that, thank you.

lung cancer screening

challenges

uptake

invasive mediastinal staging

geospatial analysis

electronic health record data

interventions

awareness