Thank you, and thank you to the organizers for inviting me to speak on environmental factors today. I'm actually going to focus specifically on ambient particular air pollution matter, because it ranks second only to smoking in terms of estimates of lung cancer deaths from these selected risk factors, and at far outnumbers are traditional risk factors such as occupational exposure to asbestos, secondhand smoke, and residential radon. There is increasing evidence that air pollution is indeed a major cause of lung cancer, especially in those who have never used tobacco. There are many large epidemiological cohort studies proving this. And in 2013, IARC actually classified PM2.5, so it's that particulate matter that measures 2.5 microns, as a group 1 carcinogen. And if you look at the annual mean exposure globally, you'll see that very few countries indeed are now meeting the World Health Organization air quality guidelines of less than 5 micrograms per cubic meter. And in fact, many exceed this by between 35 to 100 and exceed it excessively. And this matters. It matters because there is a concentration response function for mortality associated with particulate matter. This is some work done in Canada looking specifically at low air pollution exposure between 5 and 10 micrograms per meter cubed. And you can see that there is still a curve associating mortality with this particulate matter. This is a group of 2 million Canadians that were followed over 20 years for non-accidental death. And at the same time, this was done by the Canadian group. Other groups were doing similar studies. And you can see that the average PM2.5 was 7, 11, and 15 in USA, Canada, and Europe, meaning that no one is meeting these WHO criteria guidelines. There are large epidemiologic studies proving this, but what we wanted to do was see if we could do an actual personalized PM2.5 risk assessment for individuals. In Vancouver, we have about 40% of our lung cancer patients who are never smokers. And so we took this cohort and did a residential history from the time of birth to the time of their cancer diagnosis. We did this with time-weighted mean exposure of PM2.5 for each individual residence, dating at least back to 1996, where we had accurate satellite data. We used satellite data observation, chemical transport models, and ground measurements combined to give us a resolution of about 10 by 10. And this required recording postal codes from patients from time of birth to time of cancer diagnosis. And what we found was that the mean 20-year exposure of greater than 10 micrograms per cubic meters of PM2.5 significantly increased their lung cancer risk in never-versus-ever smokers, particularly in Asian females, as you can see in the top blue line. I don't think we can really talk about air pollution and cancer without mentioning Charlie Swanton's paper that was published in Nature, which for the first time is looking at the mechanisms by which these environmental carcinogenic exposures promote cancer formation. And what they concluded was that they found oncogenic EGFR driver mutations in about 18% of normal tissue, greater in females than males. And this oncogenic mutation accumulates with the natural aging process of the lung. And PM2.5 actually acts on these cells harboring this pre-existing oncogenic mutation to promote cancer. To the right of the screen, this is some work that we participated in in this publication by showing, again, these personalized PM2.5 calculations for a 3-year cumulative risk or a 20-year cumulative risk in never-smoking lung cancer patients. And what we found was a statistically significant increase in EGFR in never-smoking lung cancer in a short 3-year cumulative exposure, indicating that maybe just 3 years of a high PM2.5 exposure may be sufficient for EGFR mutant lung cancer. I don't think this answers all the questions, as we have to figure out how we explain all the long-term exposures. And we know that from previous work done in Taiwan, that the change in air pollution exposure in lung cancer incidents can take 15 to 20 years. So once you see that uptake in air pollution, it takes about 10 to 15 years to see the incidence in lung cancer rise. We also don't fully understand all of the aspects of early life exposure to air pollution in a developing lung. We know that there's some data showing that exposure to air traffic or diesel air pollutants during infancy can have long-term lasting impact on lung function that lasts up until adolescence. And more work is currently underway looking at the effect of PM2.5 causing premature lung aging. Our group took a look at hypothesizing that perhaps it's never-smokers who develop lung cancer metabolize carcinogens that we ingest or inhale differently. And so we did this by using a volatile organic compound and using breath collection for a metabolic washout of a peppermint tablet, where we knew there's well-documented washout curves for each of the components of a peppermint tablet. And you can see the cancer in the lung cancer patients. These are all never-smokers with previous lung cancer versus age- and sex-matched controls. You can see that the washout, on average, the elimination rate is much slower for all of these compounds. And this also makes us think that perhaps this could be a way to identify those patients who are higher risk for lung cancer. This is some other work that we are doing in collaboration with Owlstone Medical, where we're doing high intense exposures to air pollution and then looking at the inflammatory response of the lung over time using breath biomarkers. To the left is the exposure, the APEL exposure lab, which is the air pollution exposure lab in Vancouver, B.C., where we pump in 300 micrograms per meter cubed of diesel fuel into a small plexiglass box. All the participant exercises within there. This is ethics approved. And then post-exposure, the participant does breath samples for us at 1, 3, 6, and 24 hours. So I just want to show you some. We've done a targeted analysis, and this is very prelim data. But I'll show you some of the data just looking at using targeted analysis on known volatile organic compounds that indicate lung inflammation. And what we see is the baseline, this is in the red, is the pre-exposure breath. And the post-exposure breath is easily separated by these inflammatory markers. And this is one hour post-diesel exposure. This effect lasts up to three and even six in the green. And it's not until 24 hours where you start to see, the 24 hours is the blue stars or the navy stars, where this shift is starting to come back. So we know that after this very short exposure to air pollution, we have lasting inflammatory reaction for at least 24 hours in the lung. We also did corresponding blood biomarkers, and we're just working on putting all of that data together. We just lived through the hottest month on Earth's record. And because of that, had unprecedented wildfires around the world. And I don't think this is going away anytime soon. And wildfires in certain parts of the country, such as Canada, are a big problem. We know that wildfires release human carcinogens, such as polycyclic aromatic hydrocarbons, benzene, formaldehyde, heavy metals, and people living within that area have dense exposure. This is a community about four hours east of Vancouver that was destroyed by a forest fire this summer. And this is the subsequent skies in Vancouver that we lived through for weeks, when the air pollution, depending on which way the wind blows. And at this particular, this is a picture from my office window. It looks apocalyptic, but the PM 2.5 at this time was 350. And this lasted for several weeks. There were fires on the east coast of Canada that drifted down, covering the United States in smog. And people lived through this for weeks as well. As well as the fires in Maui. This is Lahaina. And this is the forest fires in Turkey. And this picture just shows the dark, dark, smoggy skies and people out exercising and enjoying the beach. And it just speaks volumes to the amount of awareness and work that we have ahead of us. Where people need to know that you can't be exercising in these conditions. We're just starting to now get some data showing linking wildfire and cancer incidents to lung cancer. And this was a study published this year that looked at the exposure of patients, or participants, who lived either 20 or 50 kilometers near a large burn area. So they excluded people living in urban centers. These were rural centers of patients who lived either 20 or 50 kilometers near large burned areas. And what they found was, this was 2 million Canadians followed from 1996 to 2015. That compared to the cohort members who were never exposed to wildfires, exposed populations displayed consistent elevations in incidents of lung cancer and brain cancer. So as the work lays ahead of us, ISLAC has now put forward a position statement on air pollution and lung cancer, which I think is an important step forward. This was led by Christine Berg and Stephen Lamb. Chris Berg could not be here today. And it's a call to action and a call for awareness that this is indeed a risk factor. We recently, our same group, published this paper that just came out in JTO. And again, it's air pollution and lung cancer, a review. And we do have a poster. So come visit me standing by my poster tomorrow, tomorrow night to have further discussions about this. So the future directions, can we integrate this air pollution exposure information on a personal level into risk prediction models that can help identify these high risk never smokers? And ultimately, can we actually do things to prevent the adverse effects of exposure to PM2.5? Because the earth is on fire. These things aren't slowing down anytime soon. So we need awareness for personal protection. And we need ways to prevent people from being exposed. That's all I have for you. Thank you. Thank you, Stephen. So I'm going to talk about the tumor biology-based biomarker for lung cancer. I was going to talk about why do we need it. But after the session this morning, I think I'm going to distort it to three words. One is we need targeted, Stephen, screening. Second, our tools are too blunt, Raymond. And we need objective measurement, Hillary. So that's why we need a biomarker. I just also want to mention that because the proportion of the lung cancer now arriving in a contemporary population really started to decrease in terms of those that fit into the classifications, if we don't take any action to sharpening our tools, the percentage of people who can benefit from CT screening will be decreasing in the next three decades. So we need some action to do some work here. And biomarkers is one of the possible tools that we can use to really optimize the screening pathways from eligibility assessment to determine the intervals and to nodule malignancy prediction after screening. So how do we do that? So we don't need to start from scratch. We do know something about lung tumor. We know lung tumor exhibits substantial differences in many different aspects in terms of the somatic mutation spectrum, epigenetic signatures, alterations, their differential protein profiles and macronase, and more. And they tend to map really nicely to the cancer hallmarks that has been described. And there are 14 hallmarks that are showing on the right-hand side there, just published last year. In particular, the blood-based biomarkers has that potential to leverage to identify those differences, to identify the people we need to screen and target population, know what to do with them after screening without the need of biopsy. So it is a very active area. So there have been quite a few commercial panels that have been put out there. Some of those that you might have heard of, the early CDT lung and measuring the autoantibodies, notify XL2, that's the ratio of the two proteins. However, there are also limitations that have been reported for those commercial panels. The sensitivity is just way over what is going to be clinically useful, and some of the panels still need more clinical validation before it can really be useful. At the same time, there are active clinical trials and ongoing prospective cohorts that are taking place. For example, BioMild taking place in Italy looking at microRNAs intervals. And SOSUS is actually a Singapore lung cancer one-arm trials that I believe just started last year. And it's a new trial looking at different biomarkers and many more. So this is not an exhaustive list, but just to give you a sense that there are many different classes of biomarkers, and they are in different stages. And to decide what biomarkers to use in what situation will depend on what is a clinical need. So I'm going to talk about a couple of the projects we've been working on in this field. One is the cell-free methadone, and in which case I'm going to bring up the issue of multi-cancer specificity. The second project I want to talk about is circulating proteome, and in which case I'm going to emphasize the need of multi-stage validation based on pre-diagnostic samples. So cell-free DNA methadone. The circulating cell-free DNA is considered to be an ideal alternative of the biopsy because it has already been shown to be released into bloodstream through tumor, apoptosis, and necrosis, among other things, obviously. And it was already demonstrated to reflect somatic mutations. Instead of the somatic mutation, the epigenetic alterations, they do have several advantages. For example, it has a higher sensitivity. It's across the whole chromosomes, sorry, whole genomes. And it has a unique property of carrying what we call an ID card, because you can actually infer the epigenetic changes back to the tissue of origin, which will help you to determine the tissue specificity. So this is a project we did a few years ago when we published in Nature. We started with the emphasis on lung and pancreatic cancer, but we ended up doing seven different cancer sites with a total of about 380 plasma and controls. And what we were trying to do is to see whether or not we can capture tumor signals in a cell-free DNA compared to what is the primary tumor in the TCGA data. And the TCGA data stands for the Cancer Genome Atlas Data that's publicly available. So the graph on the right here, it shows you that there is a much higher than by chance concordance between the cell-free DNA and a tumor, a primary tumor organ in terms of epigenetic alterations. So we are capturing the tumor signals from the primary organ. If you visualize the epigenetic signatures in a space, what we see is that each tumor has a very distinct methylation pattern in a cell-free DNA. And on the top panel is the data from our study samples. We take the same methylation regions and go into the TCGA data, and we use 4,000 samples, and again, we see very distinct patterns. So we are pretty sure that, confident that methylation can provide a very tissue-specific result. And we actually were also able to map it back to the tissue region based on those motif analysis and active transcription factor binding sites. Now this project is now moved to the prospective samples with large cohorts. I'm working with PENCAN studies as well as PLCO. How does that compare to the other cell-free assays? So this is the result that's published by GRAIL just last year. And so what they did is they did a systematic comparison of different assays from cell-free DNA, and they will look at mutations, they will look at copy number changes, they look at epigenetics. They used the circulating cell-free genome atlas study. And what they found is basically the same conclusion as what we had is that methylation is optimal if you are interested in multi-cancer specificity. There are remaining issues with this, though. First of all, we don't see data represented with a systematic comparison with a tumor profile from this GRAIL study, which might come out later, but currently is lacking. and the other issue is it needs a very large amount of cell-free DNA. So that puts the issue into a feasibility when you're trying to do a screening in an asymptomatic population. So that's for the cell-free DNA methadone, and the second project I want to talk about is circulating proteome, and there is a longstanding interest with the circulating protein markers and cancer, and actually some of the classical cancer markers are the circulating proteins. It is thought to be an ideal biomarkers that fill in the gap between genetics and phenotype, but the problem is the previous study tend to be either small or they use targeted proteins, so they tend not to be very reproducible. That thing's changed when this recent high-throughput technology is put on the market and for circulating proteomics, and that really facilitates a more comprehensive investigation. As a result, quite a few high-profile papers has popped up in the last two years in Nature and Science looking at plasma proteome and various different type of human diseases, including cancer. So we have a project that's funded by NIH, which is called INTEGRAL. It stands for Integrative Analysis for Lung Cancer Etiology and Risk, and quite a few of you actually in the audience are part of the projects. In INTEGRAL, we are focusing on two parallel lines of research queries. One is a prescreening, looking at lung cancer risk, and that's led by IR, Hillary is one of the key investigators on that front, and the other one is a post-screening, looking at pulmonary nodule malignancy prediction, and the design elements is shown here, and you can see that we are really focusing on the multistage validation to ensure the robustness of the biomarker. And this is published, actually, last year in Annals of Epidemiology, so I can provide a reference if people are interested. One thing about looking at the protein profiles is also that you can look at time-to-diagnosis. You can see the dynamics of the protein, and you can see how that would change based on the time-to-diagnosis. So with the multistage discovery and validation, we included 14 population cohorts for the risk biomarkers and six CT-screening studies for the nodule malignancies, and all the analysis was done based on pre-diagnostic samples. So for both results, it has been published this year in Nature Communication and JNCI, so I won't go through the details, and I'm happy to provide a reference. But I do want to mention that because the title of the talk is tumor biology, we do have that interest in what does that represent in terms of tumor biology. So on the lung cancer risk side, we found that the biomarkers that were identified has a really good sort of matchup with the cancer hallmarks in terms of the nodule malignancies, and what we found is a very tightly linked networks, including the apoptosis, the immune system regulations, et cetera, that are really dominating that part of the biomarker. Based on the two results, we have now come together and really configured, we call it a dual-purpose custom panel, and the idea is with one panel, we are going to be able to sort of take the patient throughout the journey of the screening from screening eligibility to nodule malignancy process prediction. So if you look at our design elements, again, we're currently at the validation stage, using the absolute quantification. If people are interested in some more results on this, so I believe a fellow working with Hilary Hanna is actually going to present this on Sunday at 4 p.m., so feel free to drop by for the poster. So I just want to make a few points. Circulating biomarker with predictive ability, sometimes it reflects the somatic tumor properties directly from the target organ, which is the case with the cell-free DNA that we see, but sometimes it doesn't. It might represent system dysregulation or the microenvironments that eventually link to tumor developments. It could be inflammation or a response to it, but if the biomarker can provide optimal sensitivity and specificity, whether or not it reflects directly the tumor properties from the target organ, it could still be a good biomarker. And so how do we make sure they have optimal sensitivity and specificity? The multi-stage validation is really the key, and I do want to mention that the optimal class of biomarkers, as I mentioned, there are many different types, but it can really depend on the clinical need. Finally, the discovery and the validation of biomarkers is not the endgame. Our endgame is really to establish the risk model. So to have a risk model to aid the clinical discussions, which means a model that's well-calibrated is just as important as a model that has a high AUC and validated. Finally, to have a successful implementation, it really needs, of the biomarker, it really needs a multifaceted consideration. You know, the cost-effectiveness analysis that some speaker talked about, how do we communicate the results to the participants? Does it affect equity? Does it affect the screening intake? And all that needs to be considered under the benefit and harm balance. That was also highlighted in the previous session. So a quick acknowledgment slides for the people who have done the work, and many of you who are also part of that collaboration. Thank you. My title is Next Generation Lung Cancer Risk Prediction Model, and I will focus on the absolute risk prediction models, which provides the probability that the individual develops lung cancer in an upcoming period. It also has the clinical utility of risk stratification in the lung cancer screening. And the Early Detection and Screening Committee gave a report published in JTO in 2021, and they reviewed four manuscripts for the never-smokers. And in fact, in this review, we can see that the best for the validation data, as you see, is around 0.714. And Tomasas also gave a review, reviewed about around 50 papers. And through the review, they mentioned that the future directions include model-specific thresholds to define the high-risk population for different settings, and also the evaluation of effectiveness and cost-effectiveness of risk-based screening programs to see if it averts more lung cancer deaths and minimize unnecessary invasive exams. The dynamic nature of lung cancer risk and personalized screening schedule for different groups should also be paid attention. And let us see that within these two years, there are several important papers coming out. The first is the NCCLC model conducted in China, and the sample size is around 1.4 million, and the prediction horizon is three years. And the good thing is they developed the thresholds for never-smokers and ever-smokers. And the second one is the CAN predict model conducted in England, and the overall sample size is around 19 millions, and the prediction horizon can up to 10 years. But however, they didn't give the models for never-smokers. And Regent Hong and her team used the case control from the ECO and the validation from the U.K. Biobank, and they tried to see that the polygenic risk score, adding this with or without, with, can improve the performance. Here in Taiwan, we used the case control because we didn't have enough large sample size to have the cohort to follow up. So therefore, we combined these with the synthesized multiple data sources, and the performance is not too bad. The AUC for never-smokers is about 0.714. And how about the risk factors? And for the China paper, they only contained about four or five risk factors. And for the England paper, because they used electronic health records, so they'll be able to put the comorbidity, like history of pneumonia, asthma, into the model. And of course, in the Regent's paper and also ours, we did put some SNP panel data or PRS. And let us see that about the NCC model conducted in China. For the three years lung cancer thresholds, they recommended 0.47 percent for never-smokers. And they'll be able to screening 11 percent of the Chinese never-smokers aged 40 to 74 years. And it would detect 27 percent of the lung cancers. And however, we expect more findings for longer prediction horizons, like five, six, or 10 years. And about the U.K. can predict the model. And the good thing is that they provide the net benefit, which is defined to be the true positive minus the true negative weights by the odds at the threshold probability. And it shows that according to the decision curve here, based on the net benefit, the can predict model had the highest net benefit for women and also for men. What about if a Regent's group, they added the polygenic risk score? And we can see that the red line here is for the top 10 percent of the polygenic risk score decile. And you can see that clearly that the polygenic risk score would affect individuals up to the risk with increasing age. So therefore, from this, we can see that individuals' genetic background may inform the optimal timing for the low-dose CT screening. However, it is based on the U.K. biobank data and that assuming 1.5 percent of five-year absolute risk lung cancer as a threshold, this is the never smokers. And we can see that never smokers did not reach the recommended for the low-dose CT. The highest here is .3 percent, however, the threshold is 1.5 percent. And regardless, they're PRS deciles. So therefore, we do need some models for Asian people, especially in this region, not many prospective cohort data are available. So here we develop some absolute risk model without the prospective cohort data. As long as you have the case control study and we synthesize the multiple data sources, and we'll be able to develop the models for the never smoking women and also recalibrate the PLCO model in Taiwan. So for the never smoking model, and if we use the same threshold, so 1.51, then we're screening only 5 percent of never smoking females. It would detect 27 percent of cases. And for the PLCO, for the ever smokers, the AUC is around .78. So as good as in the original paper by Tamagotchi. And you can get this from the website. You just Google lung cancer risk calculator for never smoking women. You'll get all these app. And here we also model, do the model improvement by adding more environmental factors such as indoor, outdoor air pollution, as the environmental tobacco smoking, PM 2.5. And also we add some new SNPs, newly found for non-smoking females for the yeast patient to create the proteogenic risk scores. But we'll not be able to present here, and the paper is still in preparation. And you might be interested to know how do we synthesize the multiple data sources. We use the cancer registry data, National Health Insurance Research Database, and also the Adult Smoking Behavior Survey, and also the Taiwan Biobank. We will be able to get the distribution of the risk factors in the population. So with all these different sources of data, we'll be able to synthesize the data set resembling the population of cancer-free ever smokers in 2010, and then follow them for the next few years. So the take-home message here are, first, region-specific or setting-specific models and thresholds are needed due to different genetic and environmental backgrounds. And better models for never smokers are urgently needed for screening policy recommendation. And effectiveness needs consideration. And especially in the region of Asian, some regions do not have the prospective cohort studies. Then you can try to use the synthesized data sets from multiple sources. And still, you can get performance prediction models. And when we talk about the risk factors, of course, there are more appropriate risk factors to be incorporated in the model for better improvements, such as secondhand smoking, such as air pollution. And we talk about over-diagnosis. Again, the further research is required to elucidate the impact on lung cancer mortality over-diagnosis and the cost. And finally, I'd like to thank our group. Dr. Yang Panchi, he led our study, GELAC, and also the collaborators from the NCI NIH in U.S. and from NHRI. Especially, I'd like to thank the Data Science Center of Ministry of Health and Welfare in Taiwan to get all those data. Thank you very much. Thank you very much. So, yeah, I'm going to talk about ways we can potentially modify lung cancer risk. So, first off, starting to look at smoking, which, as we all know, is the leading cause of lung cancer. So, this systematic review, which was published by O'Keefe and colleagues, looked at 99 cohort studies containing more than 7 million individuals and over 50,000 instant lung cancer cases. And you can see from here that women actually have a very similar risk profile to men, whereas we typically think of it being kind of a disease of males. That's probably because the female stages of the lung, the smoking epidemic, haven't been fully reached across the world. What you can also see from here is it doesn't matter what we adjust for, be it age or where people live or any other background. The risk remains largely the same. And there is also a dose-response relationship. so the more people smoke, the higher their risk of lung cancer is. Evidence here is from two different studies which looked at the risk of smoking and lung cancer and found that if you continued smoking, your risk was far higher than quitting smoking at any age. Obviously, if you're quit by the time you're kind of below 35, as shown in the study by Jarr on the left, the risk reduces to pretty much the same as a lifetime never smoker. If we look to think about secondhand smoke exposure, which is also quite a big problem, so looking at the relationship between secondhand smoke exposure and lung cancer, predominantly by histological type, this study used the International Lung Cancer Consortium data, which included 18 case control studies and more than 12,000 cases of lung cancer. And you can see at the top that ever exposure to lung cancer, ever exposure to secondhand smoke exposure compared to never exposure has about a 27 percent higher risk of developing lung cancer, and that risk is particularly high in small cell lung cancer at a hazard ratio of 1.44. So if we compare the smoking prevalence around the world, which is the figure on the left, with the age-standardized incident rates of lung cancer on the right, you can see that there's a very similar pattern across both. However, they're not identical, which suggests that there are other things that are at play that are having quite a large impact on lung cancer rates around the world, one of which being air pollution, and I won't spend much time on this because Ronell's talked with far greater expertise than I ever could on this topic. But just to reiterate, PM10 and PM10.5 are both linked to lung cancer causally, as defined by IARC, responsible for around one in 10 cases, and sources include the combustion of gasoline, diesel, oil, and wood. If we look at exposure around the world, again, this has already been covered, but the WHO has air quality guidelines, which recommend an average of 10 micrograms per meter cubed for PM2.5 concentration, and around 95% of the world's population exceed this limit, so it's something that we do need to think about tackling to reduce lung cancer risk. These data were taken from the ESCAPE trial, compared both PM10 and PM2.5. They looked at 17 cohort studies from nine European countries and 2,095 lung cancer cases. So for PM10, we can see there's a significant increased hazard ratio at 1.22 with air pollution exposure, similar increased risk with PM2.5 at 1.18, but that wasn't significantly raised. Again, this has already been covered. This is the study that was published from Charles Swanton's group earlier this year. They used data from over 400,000 people and found that air pollution wakes up cells which have got cancer-causing mutations, which can lead to tumor production. Again, I'll mention it here because it will be of relevance when we talk about modifying risk. So going more locally, we look at household air pollution, and nearly 2.5 billion people cook using polluting open fires or simple stoves using kerosene, biomass, or coal fuel, particularly in low-income countries. Combustion from biomass group products is a 2A carcinogen, which means they're likely to cause lung cancer, and we can know that there were about 3.2 million deaths in 2020 from illnesses attributable to household air pollution. Six percent of these were from lung cancer, and 11 percent of all lung cancer deaths are attributable to household air pollution. Now because women and children, particularly in developing countries, seem to bear the brunt of kind of indoor cooking, fuel collection, et cetera, the impact is far greater in women than it is in men. If we look at occupational other exposures, so radon, radon's a naturally occurring radioactive gas, which has been defined as a group one carcinogen by IARC. It's a leading cause of lung cancer, contributing to between 3 and 14 percent of all lung cancer deaths, depending on the country of origin and their smoking prevalence. And we also know that radon's kind of a double whammy, so there's a synergistic action with cigarette smoking, so if you're exposed to radon and you smoke cigarettes, your risk of lung cancer is about 25 times higher than radon exposure alone. Similarly, we also have exposure to asbestos, another group one carcinogen, and we know that 125 million people have been exposed to asbestos at work. It's the cause of around 50 percent of occupational cancer deaths, and again, there is a synergistic action with cigarette smoking. And I think this is one of the biggest questions, certainly I'm from a tobacco control background, so does vaping pose a lung cancer risk? Now, vaping is promoted as being 95 percent safer than smoking. It's predominantly pushed as a harm reduction aid for people to stop smoking, but we don't yet know the long-term safety effects. We know that vapor contains some carcinogens, particularly nitrosamines, though they are in significantly lower quantities than found in combustible tobacco products. We don't have much research here yet, but there has been some evidence published. These are just two examples that are suggesting that e-cigarette smoke may induce lung cancer in mice. So talking about preventive action, then, first of all, so for tobacco, we need to support smoking cessation, which I've already covered on the earlier slide to see the impact that that can have. We need to reduce uptake and protect never smokers, so we've got the WHO report on the global tobacco epidemic. We have the Framework Convention on Tobacco Control, which is the first world public health treaty which supports signatories to implement effective tobacco control measures. And this just on the right-hand side, this is from data from the UK, which is one of the world leaders in smoking cessation and tobacco control, and it just shows the impact that you can have to reduce smoking rates with actions, be that legislation, mass media campaigns, public service interventions, et cetera. If we look at air pollution, so the WHO has guidelines for air quality, which we've already mentioned. Not much you can do on an individual level to reduce your air pollution risk, but as a population and systemic level, we can burn less fossil fuels. We need to look to develop cleaner fuels, initiatives such as clean air zones, active travel networks, reduce vehicle emissions, and so on. And if we go back to the work of Charles Swinton, if we can look at further research for how we can stop these cells from developing into tumour cells upon air pollution exposure, then we can reduce lung cancer prevalence as a result of air pollution, and also protect never smokers. Radon, as I've said, it's naturally occurring. The WHO recommends reference levels of 100 becquerel per cubic metre, which can increase up to 300, depending on the country, and there have been efforts to look to try and reduce this. We did a Radon survey in 2019. Of the respondents, which was pretty low, only 44% had national Radon action plans, 39% included protected measures in code for new buildings, and there was only 12% having education for building professionals and 15% financial support to fix existing buildings. So this is very much a structural issue. Reducing Radon in existing buildings is very much around ventilation, and so there needs to be an investment into that to reduce Radon exposure. Asbestos, I think everyone knows how to reduce the risk, so replacing with safer substitutes, prevent exposure to asbestos in situ and during removal, and have improved surveillance for those with a history of exposure, and I think this links to the debate that was going on this morning. If we know people have got long-standing exposure to asbestos through their work or environmental exposure, should they be considered for lung cancer screening, so that if they are to develop lung cancer, it can be detected earlier and therefore more effective treatments? Vaping, one very close to my heart, I think very much has a place as a harm reduction measure for people who cannot or will not quit combustible tobacco, but we do need to really be getting the messaging right to reduce the likelihood of uptake by never-smokers, and particularly young people. We know that with smoking, the age of initiation is an important factor for development of lung cancer, be that through impact on the lung development or through a dose-response relationship. We really, really need to get that right, and I don't think we've got that yet. They've caused for taxation to reduce the appeal of disposable vapes, which are particularly appealing to youth, which feeds into the same problem I've just discussed. We need better regulation and monitoring of e-liquid content, and I think one really nice example of that was the Avali outbreak, which was largely linked to vitamin E acetate through poor quality vaping liquids, and I think we absolutely need closer monitoring of long-term health effects. We do not want to be creating a new time bomb that we had with combustible tobacco through vaping. So, I think the take-home message for me is the vast majority of contributors to lung cancer can be reduced or even completely eliminated. It needs concerted action by policymakers, by individuals like ourselves in the room, but we need to have a system-level approach. This is pretty much none of the things I've talked about are individual-level actions. We need system-level actions to be most effective. Thank you very much. Well, thank you to the organizers for giving me the privilege of closing this wonderful session. Thank you to the speakers and to the people who asked questions, which I think got to a lot of the points I wanted to make. I know I'm the thing between you and lunch, so I'll try to be brave. So, I mean, my perspective as I'm hearing the first session, of course, and then this one is that, of course, boy, we had it, in a way, easy when smoking was the leading factor and cause of lung cancer, and when it was easy to use smoking as a way to predict lung cancer risk. And the good news is that, unfortunately, that's at the same time is, of course, but the great news from public health is that, for the most part, smoking is coming down, and now we have to deal with all the things that we happily ignored in the past as smoking was driving most of the risk. So as we're getting to that point that now really we need to address not only risk assessment, but then also, for instance, how do we prevent lung cancer among this population or groups of individuals who didn't smoke or didn't smoke much, but now also have lung cancer risk, and how do we assess that risk is that sort of I see like two sort of maybe three ways that people are trying to address it, which is take all models or existing models, and I just want to take this opportunity also to acknowledge the wonderful benefit that we've had of having validated, effective, predictive lung cancer risk prediction models, PLCO, LLP, many others, and that build on those at some of the markers or the factors that we didn't have before and expand them, then also maybe do different type of prediction based on more on biological outcomes and using biomarkers to measure risk, and then maybe the combination of both, right, because certainly we can expand risk models by putting them or combining them some way with biomarker data as was shown here, and it was great examples of what we have today. So I think it's really sort of where the field is going, and I think a very important point, for instance, to talk about in terms of the integration is that we really need to understand how lung cancers occur in people who didn't smoke. We need to understand the mechanisms. It's very important, then, that those mechanisms inform not only the assessment or development of biomarkers, but also the development and understanding of risk, per se, and the development of risk prediction models. So I see, in a way, a lot of excitement. We're going to have, of course, a much more difficult time than what others had in the past in measuring risk, but, of course, we have challenges, but that brings lots of opportunities. So one way that was highlighted with the last question and with the work of Dr. Myers is that we need innovative ways of measuring exposures. In the case of air pollution, I mean, it was shown, right, you can use addresses to get a measure of air pollution assessment with all its limitations. Yeah, it's going to be limited. It's not going to be perfect, but let's not forget the smoking assessment is also imperfect. We're just asking people how much did you smoke when you were 20 years old, when they're 70 or 80. That's not going to give you a precise measure. So we need to keep in mind that we don't have this. We need to understand what is the level of precision that we're trying to find based on what we want to do with those models. Certainly there's a challenge in terms of the relationship between exposures, biomarkers, and risk. One thing that I didn't see, for instance, is how does those exposures, air pollution, asbestos smoking, relate to the biomarkers themselves? Not only, of course, because there's a lung cancer tumor there, but also because there might be differences in level based on exposure history. And I don't know that we spent too much time looking at that. The other thing is the interactions. We had an easy time with smoking, one major risk factor, the other sort of at a little bit. So we still need to look at risk factors that don't have a lot of predictiveness by themselves, but it's how they interact. And how they interact and how they result in risk is going to differ between never smokers, ever smokers, and not risk. And that's also a very important thing. I really like the point made that the risk models and the thresholds should vary by population and purpose. If you're going to use a model to predict someone in the clinic, do you have lung cancer? It's a very different type of question and precision that you need than is the question is do you have risk high enough for me to recommend you to get screened? It's a very different question, right? The risk of sending someone to screening is not necessary. You don't need so much precision, right? You just need to know if they're high enough so that screening might be beneficial for them. So certainly populations and countries, it's going to be very important to keep that perspective that we need specific risk models and we need specific risk thresholds attached to each specific risk model based on what these models are going to be used for. There are challenges of implementation. Of course, cost was also brought up. The issues of risk-based screening, the good news is now we have practical implementations of lung cancer screening based on risk in the U.K., Canada, Australia, and others. Once you have that in principle and then you improve your risk model and now you add air pollution, maybe you already have the framework to bring screening to others that didn't meet that risk threshold just with smoking, but now when you account for other things they actually do. So the fact that we have implementations of risk-based screening for lung that we don't have for any other cancer is something really powerful and that you should take advantage to really study and learn a lot from those implementations. And in terms of other challenges, I think Dr. Murray pointed out, for instance, emergence of new risk factors. Right now it's vaping, and certainly vaping is going to be an important one that we're going to need to care. And just going back to the challenges, the effect of vaping is probably going to be different if it's a never smoker who then vaped for 30 years versus it's a former smoker that then switched to vaping and maybe there was already some lesions there for vaping to act on versus if there weren't before. So I think coming back to the first point, or one of the first points I made, that it is really we need to come back and let mechanisms drive epidemiological analysis and insights and then risk assessment. And then really come back to also the point of we have the advantage of being able to have really good risk assessment methods and these all new developments that we heard today and many others that we'll hear during the conference to really use risk assessment to drive population and individual lung cancer prevention. So with that, I want to thank the speakers and just join me in giving them a hand for the wonderful presentation session. And this concludes the session. Thank you.

lung cancer

risk prediction

prevention strategies

smoking

air pollution

occupational exposures

vaping

biomarkers

genetic data