Introduction. If we could all start to take our seats, we really need to get moving with the next session. Well, before we begin, I'd like to acknowledge some of our sponsors. Medtronic and Intuitive Surgical as educational supporters of the 2018 meeting, Abbot, Edwards Lifesciences, Medtronic, LivaNova, Intuitive Surgical, and Zimmer Biomed as supporters of breakfast, lunch, and this evening's symposium taking place on Thursday and Friday of the meeting. Don't miss the lunchtime symposium sponsored by Medtronic on meaningful innovations from 12:30-1:45 pm in Salon 6, followed by a residents' symposium on transitioning to practice from 1:45-3:00 pm in the same room. After all the afternoon learning, please join Zimmer Biomed for a reception symposium, “Improving Outcomes and Reducing Complications in Cardiac Surgery” from 3:15-5:00 pm in Salon 4.
        A couple of social notes. The Spanish Nights theme dinner will be this evening and the attire is cocktail. All attendees must have a badge to enter the event, so please make sure that you and your registered family and guests wear your badges, and please be sure to sign up for a table for Saturday night's banquet. The seating chart and signup sheets are located in the registration area. As moderator to this session, I want to remind presenting authors that their disclosure slide should precede their presentation and can be displayed for 3 seconds.
        Now I'd like to invite our first presenter for the second session to the podium. This is Dr Melanie Subramanian and she will be presenting “Imaging Surveillance for Surgically Resected, Stage I Nonsmall Cell Lung Cancer: Are More Scans Better?”
        Dr Subramanian. I'd like to thank the Western for allowing us to present our study. There are more than 400,000 estimated lung cancer survivors in the United States, and that number is likely to increase with improvements in therapy as well as lung cancer screening. Patients with early stage, surgically resected, non–small cell lung cancer (NSCLC) remain at risk for development of disease recurrence as well as new primary lung cancer. This makes postoperative surveillance, and especially surveillance imaging, an important component of care for survivorship. However, there are limited high-quality longitudinal data to best inform evidence-based regimens that are comprehensive of both imaging modality as well as intensity. Additionally, previous literature and existing guidelines have traditionally clumped together multiple stages of disease, and it is unclear that these patients should all be managed the same.
        When we look at recommendations from national practice organizations like the American College of Chest Physicians, the National Comprehensive Cancer Network, and the International Association for the Study of Lung Cancer, we can see that imaging at intervals varied quite a bit with intervals ranging from every 3 months to annual visits. This heterogeneity is likely reflective of the lack of high quality longitudinal data, and this is particularly true for patients with stage I disease where surgeons are likely to play a lead role in surveillance.
        We proposed a study to compare computed tomography (CT) imaging surveillance intensities in patients with surgically resected, pathologic stage I NSCLC. Our primary outcome of interest was 5-year overall survival and we hypothesized that surveillance intensity was not associated with 5-year overall survival. With regard to imaging modality, we chose to focus on CT images given previous level I evidence from the National Lung Screening Trial that suggested the superiority of CT images compared to chest radiograph in lung cancer screening. In addition to our primary objective, we examined trends in locoregional as well as distant disease recurrence and examined trends in the development of new primary lung cancer. We partnered with the American College of Surgeons Commission on Cancer to perform a special study, which allowed for the abstraction of enhanced patient information that is not otherwise obtained in the National Cancer Database. All Commission on Cancer sites that contribute to the National Cancer Database were required to participate. In this special study, the Commission on Cancer collected information on detailed comorbidities, 5-year overall survival, disease recurrence, development of new primary lung cancer, and all postoperative images as well as an indication for each image.
        Using the special study data, we performed a retrospective cohort study with 3 surveillance intensity cohorts. These cohorts were based off of time from surgery to the first CT surveillance image and were informed from previously published guidelines. These cohorts were based off of visit windows that mirrored 3-month, 6-month, and 12-month time points from surgery and they corresponded to high-intensity, moderate-intensity, and low-intensity cohorts, respectively.
        This is a schematic of the special study mechanism that was used. We started with all commission on cancer sites that contributed to the National Cancer Database and up to 10 patients from each site were randomly selected for data abstraction. These patients were required to have a diagnosis of NSCLC and have undergone curative intent therapy between 2006 and 2007. Additionally, they were required to have complete 5-year follow-up information. From there, we imposed additional inclusion and exclusion criteria, identifying patients with pathologic stage I disease and having their first surveillance CT image fall within 1 of our predesignated cohorts. We excluded patients who were symptomatic at the time of their first CT where the indication of the first CT was unknown. From here, we obtained 2565 patients of whom 845, 1282, and 438 patients fell within the high-intensity, moderate-intensity, and low-intensity cohorts, respectively.
        For statistical analysis, we performed a mixture of Cox-proportional hazard modeling as well as Kaplan-Meier analysis to model survival. We performed competing risk analysis to examine locoregional as well as distant disease recurrence. Of note, patients who were found to have positive surgical margins on initial analysis were excluded from competing risk analysis.
        This table represents demographic information across cohorts. We observed similar age, gender, race representation, and comorbidity profiles across cohorts. However, we should note that high-intensity and moderate-intensity cohorts had a higher prevalence of receiving radiation.
        When looking at tumor-related variables, we identified similar surgical resection type across cohorts with lobectomy being the most common procedure. Additionally, tumor size, histology type, grade, and positive surgical margin rate were similar across cohorts.
        On Cox proportional hazard modeling, we noted that surveillance intensity was not associated with 5-year survival. Variables, including age, gender, comorbidities like chronic obstructive pulmonary disease, congestive heart failure, and psychiatric disease, positive margin status, receipt of radiation therapy, histologic grade, and nonlobectomy resection type were all associated with 5-year survival.
        On Kaplan-Meier analysis, we observed similar 5-year overall survival probabilities across cohorts, averaging around 70%. The log rank statistic was 0.37, suggesting that surveillance intensity was not associated with 5-year overall survival.
        When examining trends and locoregional disease recurrence, we identified a locoregional recurrence rate of 13%. Approximately 50% of locoregional recurrences occurred in the same lung only, whereas 20% of locoregional recurrences were noted in both the same lung as well as regional lymph nodes. On competing risk analysis, we noted only tumor size to be associated with risk of locoregional recurrence. Surveillance intensity was not significantly associated.
        In examining trends of distant disease recurrence, we identified a metastatic rate of 13%. The most common sites of metastases were bone, contralateral lung, and brain in descending order. On competing risk analysis, significant covariates for distant disease recurrence included tumor size; histologic grade; and interestingly, surveillance intensity.
        In examining trends of new primary lung cancer, we identified a new primary lung cancer development rate of approximately 7.2%. New primary lung cancer was determined by the treating physician as reported in the medical record. Looking at information regarding new primary lung cancer, approximately 56% had a different histology than the initial tumor.
        There are some important limitations to note to our study. One is the timeliness of the data. Our study included patients who were treated between 2006 and 2007. However, this was to allow for complete 5-year follow-up information to be obtained as well as additional time to allow for the special study to abstract data and subsequent data processing. Additionally, the special study was limited in the fact that there was no collection of subsequent staging information for recurrences in new primary lung cancers.
        There are some important strengths to our study to note. First is the fact that we are able to obtain not only the postoperative imaging information, but also the indication for each study. Thus, we had confirmation that these studies were obtained for true surveillance purposes. Additionally, we decided to focus on only stage I NSCLC because previous studies have combined stage I through III disease and thus we eliminated some of the heterogeneity in the interpretation of our results. An additional strength is that we had relatively complete data of 5-year follow-up information, and finally, because the National Cancer Database captures 70% of incident lung cancers. The results from our study were fairly generalizable and have real-world clinical application.
        So in summary, surveillance intensity is not associated with 5-year overall survival in patients with pathologic stage I, surgically resected disease and our study shows that surveillance intensity should not fit a 1-size-fits all approach. Rather, imaging surveillance should be guided by knowledge of tumor biology, patient history, patient preference, and potential candidacy for subsequent treatment.
        I would like to thank the Wash U Thoracic Surgery Clinical Research Group and our funding sources from the Patient Centered Outcomes Research Institute and National Institutes of Health T32 training grants. Thank you.
        Moderator. The discussion will be open by David Cooke.
        Dr David Cooke (Sacramento, Calif). Thank you. I would like to thank the Association for inviting me and providing me the opportunity to discuss this very interesting manuscript. I would like to thank Dr Subramanian for a wonderful study and also sending the manuscript to me so far in advance I forgot that you sent it. In fact, I was wondering, well, we're getting close to the meeting, where is this manuscript and I said oh, oh wait a minute, let me just search my old e-mails.
        Your study highlights that lung cancer survivorship and surveillance is painted with a broad brush, and it's neither patient centered or targeted. It adds to the literature showing that lung cancer survivorship to be improved. In fact, Leah Backus' group showed 4 years ago that in the community-based practices, lung cancer survivorship surveillance rarely fits published guidelines and is very heterogenic in nature.
        You also point out, so I have basically 3 questions, I have basically 3.5 questions. The first, you point out that the American College of Chest Physicians and National Comprehensive Cancer Network guidelines are relatively descriptive where they start off with moderate or medium intensity for the first 2 years, and then move on to high intensity for the last 3 years for 5 years total. In your study, you look at just the first scan and make a definition of cohorts as low, medium, and high, but the reality is, most patients in those cohorts probably cross cohorts and start off at medium intensity and they move on to high intensity. Did you, how did you come, how did you make those definitions and cohorts and do you think that makes your data problematic?
        Dr Subramanian. So thank you for that question and those comments, and you bring up a very important point. There is previous literature to suggest that adherence to established guidelines, especially interval-based surveillance within the first 2 years is poor and even being cited as low as 60%, so thus, trying to capture different cohorts and different intensities as it happens in the real world remains a significant challenge. So we used this approach expecting for the fact that a lot of these patients would not adhere to interval-based images especially as time from surgery increases. However, after we established our cohort design, we calculated median number of scans per patient year, patient year in this study being not only time into the study, time contributed study or death, but also time until a new abnormal imaging finding was found because in the real world, if a recurrence of primary lung cancer is detected, the frequency of imaging may change afterward. And based on our calculations, we did find significant differences across cohorts that corresponded to our predesignated cohort intensity definitions. So although our cohorts don't follow exactly every 3 months, every 6 months, or annually, we did note a significant pattern toward that the cohorts were actually truly different.
        Dr Cooke. Okay, thank you. The second question. Can you just help me understand the role of radiation in the outcomes of your study? Radiation, from my understanding, is an independent predictor for 5-year mortality. Are you controlling radiation for positive surgical margin? Who are these patients, are they, are these patients that receive radiation for a positive margin or are they patients who were initially treated with either stereotactic body radiation therapy or definitive radiation and this is a salvage resection for in-field recurrence?
        Dr Subramanian. So within our data, within the National Cancer Database, it was difficult to interpret the true intent for radiation therapy. We thought initially it might be due to surgical positive margins. However, we observed an average about 3% positive margin rate with about 7% receiving radiation therapy, so that can only account for a portion of those patients. So we speculated, perhaps it might be due to proximity to the margin or near margins, but we're limited in our data in that we can't speculate, we can't obtain the actual data to determine why these people had radiation treatment.
        Dr Cooke. Okay. And then the final question I have, your Kaplan-Meier curves were very compelling that shows no survival difference between low-, median-, and high-intensity screening. What I don't find is a good subanalysis description. So for instance, you find that histologic grade and extent of resection are predictors for mortality, for 5-year mortality. So presumably there's a population of patients who would benefit from the high-intensity surveillance. Did you take, say, your medium high-grade and undifferentiated grade patients, see them as an individual cohort, and then look at low-, medium-, and high-intensity surveillance in that population to see if you see a survival difference. And then a follow-up of this, obviously not designed for this study but for future studies, have you thought about taking those identified patients, so specifically grade, resection, extension, and even margin, and creating a model or a tool to use as a predictor of who would need high-intensity screening and then validate that tool either by going back to the National Cancer Database or using your own robust patient population retrospectively at Wash U?
        Dr Subramanian. With regard to your first question of whether or not we performed a subgroup analysis for the Kaplan-Meier analysis, we did not, I think that would be a good future direction to take with this work or potentially in revisions. However, that's why we also looked at Cox proportional hazard modeling to adjust for factors like histology and grade, and even after adjustment we found that surveillance intensity was not associated with 5-year overall survival but your point is well taken, and we should definitely do that subgroup analysis and with regard to the future directions of potentially developing a risk prediction model, I think that's a great idea, especially because the dataset provided from this special study is so unique, we should take advantage of it.
        Dr Cooke. Great. Thank you for a wonderful study.
        Dr Subramanian. Thank you.
        Dr Leah Backhus (Stanford, Calif). Thank you for your presentation, I think it's really timely. Obviously, it's a subject very near and dear to my heart and you bring up the point that we kind of have this 1-size-fits-all approach, which is probably not best. Recently, I think it was last year, the French had their biggest prospective randomized trial looking at this very thing and they concluded the same thing that you concluded here, which is that, you know, high intensity is really of no benefit. If there is any benefit, it really doesn't show up until after 2 years. Did you see any stratification like that in terms of when there might be some benefit, or how it is that you can take your conclusions and go from there toward actually making better recommendations. My second question is with regard to the downsides of high intensity screening. If anything, your survival curves show that those guys didn't do as well although it wasn't significantly so. So did you do any analyses to look at the unintended consequences of doing more aggressive screening, for instance, radiation oncologists tend to do very aggressive surveillance where they do positron emission tomography/computed tomography every 3 months almost which is, in most people's opinion, excessive. Thank you.
        Dr Subramanian. With regard to your first question, did we examine by time if there were any changes in survival or when higher intensity surveillance might be useful. We didn't divide into specific time points like pre- or post-2 years, for example, and most guidelines suggest changing surveillance intensity. However, I think that would be a good future analysis to do. And with regard to the second question, which I believe was potential harms of higher surveillance intensity. We were able to capture information on whether or not patients underwent treatment or had a biopsy that was performed, and I think probably one of the most useful pieces of information would be to see whether or not there were high false positive rates associated with more frequent imaging. However, we weren't able to, with the information captured by the special study, to calculate false positive rates but I think for any future study, I think that would be among the most useful pieces of information to obtain.
        Dr Backhus. Thank you.
        Speaker. That was a great presentation and great study. I just have 1 comment. It's a really big challenge trying to establish a relationship between a diagnostic workup and long-term outcomes in the observational setting. I have great empathy for those challenges because our team is facing those same challenges, studying incidentally detected lung nodules in the observational setting. That said, I think you have a superb research team and you guys have done the best you can. I also want to point out that there was a trial that was presented at the European Society for Medical Oncology meeting in 2017, it's a French trial and they randomized patients to either examination and a radiograph versus examination and computed tomography in a heterogeneous group of lung cancer patients and there were no differences in 3-year disease-free survival and there were no differences in 8-year overall survival. That paper hasn't been published as far as I can tell, I think it still has to be critiqued and we have to look into the details of the study, but it may be that there really is no relationship between surveillance and long-term outcomes although that's contrary to our belief. So anyway, I just wanted to point that out and congratulations on a great study.
        Dr Subramanian. Thank you.