Good afternoon, everybody. It's the end of the day, and this is the last session of Tuesday, one to the last day of CHEST. And I'm very, very excited to present the speakers of today. This session has been once presented on Zoom last year, and it was well-received. And so this year, we are lucky to have it in person. We have three wonderful speakers. I look up to each one of them and admire them. They're all fantastic, so I'm really looking forward to hearing these talks. Our first speaker is Dr. Horyana Grosu. She's an associate professor of medicine at MD Anderson Cancer Center. She's a director of pleural disease over there, and she's done many of the really landmark studies in the area of pleural disease. And she's a dear friend and a role model, and I'm looking forward to hearing from her. Thank you, Samira, for the very nice introduction. So I have nothing to disclose for this talk. And my talk is socioeconomic disparities and lung cancer survival. And the learning objectives for today are to review socioeconomic disparities in lung cancer survival, and also review system and patient level barriers that may influence lung cancer survival. But before we go into the disparities, I just wanted to go over a few definitions, because many definitions are used differently in the literature. And this is the NIH definition of what makes somebody socioeconomically disadvantaged. And you have to have two or more criteria to be in this category. So if you were or currently are homeless, if you were or currently are in the foster care system, were eligible for free lunch or reduced lunch program for two or more years, have or had no parents or legal guardians who completed a bachelor's degree, or received support from special supplemental nutrition program, or grew up in one of the following areas, US rural area or center for Medicare and Medicaid. There is also a validated social deprivation index. And this index uses 17 census-based social indicators, indicators of education, occupation, wealth, income distribution, unemployment rate, poverty rate, and housing quality. And you can actually see here, this is the United States map of using the socioeconomic deprivation index to show in red areas with lowest socioeconomic index, or the most deprived population. And in blue, we can see the highest socioeconomic population, or least deprived. A very important term is also social determinants of health. These are things from the environment, where we are born, where we worship, where we work, age. And it's very well known that these social determinants do affect quality of life, and also risks of disease. So before going to lung cancer, I wanted to show you this study. So this is from a WHO initiative. It's called 25 by 25, because the initiative was planning to reduce non-communicable diseases by 25% by 2025. And it's a large study. It's a multi-cohort study. And the conclusion was that low socioeconomic status was associated with a 2.1 year reduction in life expectancy. And you can actually see corresponding years loss for comorbidities, also physical inactivity, and current smoking. One big problem with studies, when we look at socioeconomic status, is that they assume a unidirectional causality and association. So in this case, low socioeconomic status is associated with chronic disease, hence the reduction of life expectancy. But you can see how it can go the other way around. You can just be chronically sick. And because of that, you cannot really get a good employment. And you fall in this low socioeconomic status group. But clearly, how disparities revolve around income. And I want to show you this graph again, not related to lung cancer. But this is looking at average life expectancy at age 50. In red, we have females. In blue, we have males. And it's very interesting to see how this gap changes over time. So we can see here, bottom 10% of males born in 1920 compare with top 10% born in 1920. The gap is about five years. But 20 years later, the gap actually increases. So 1940, 76 years average to 88, so approximately 12 years. So the gap does increase over time. And Americans making more money are living longer. Going to lung cancer, we know that stage is the most important predictor of survival. And of course, patient diagnosed with stage IV have poorest survival. And this is, again, a study looking at deaths per 100,000. And the authors looked by gender, race, and education. Now, socioeconomic status education is actually seen as a surrogate for socioeconomic status. So if you are a college educated 16, 17 years, you are in this bracket down here. And you can see that black men have lower or equivalent mortality risk compared to those white men at the two lowest socioeconomic status levels. But they actually have higher lung cancer mortality rates at higher levels of socioeconomic status. And for women, black women have markedly lower mortality risk than white counterparts at low socioeconomic status, equivalent risk at middle, and elevated risk at the two highest levels of socioeconomic status. So you can actually see that the issue is very complex and multi-layers. And there are factors underlying these racial, socioeconomic, gender disparities. And even looking at comparability between socioeconomic status among the race, it's very possible. And actually, there is data to show that compared to white people, black people tend to have lower levels of income at every level of education, less wealth at every level of income, higher exposure to occupational risk with the same occupation categories, and less purchasing power. There are also other mediating factors that could include exposure to distinct environmental conditions, which can be linked to residential segregation, genetic differences, which are not studied very much, resilience factors, nativity, migration, and cultural practices and beliefs. And I'm going to go in my next slides over a few of these variables. But the question is how to study socioeconomic status and non-cancer survival. Because you cannot just look at one term and say it's gender, it's socioeconomic status. All these factors, they actually do overlap. So one idea is to look at an intersectional approach and look at them as intersecting factors. Because we are not equal. Another idea is that our life changes. Each of our life changes over time. So from birth to death, we actually have many changes in life. So looking at the multi-level life course perspective is essential given the interactive nature of exposure across the lifespan and the dynamic and heterogeneous patterning of risk. So this is actually a DAG just to show you how complex this becomes. And you cannot just look at sex and ignore age, a diagnosis, and ignore everything else. And again, it's a dynamic change over time that has to be taken in consideration. Now going back to the factors that I mentioned before, one of them is access to care. So does geographic location matter? And absolutely it does. Prevalence of cigarette smoking is higher in rural counties. Adolescence in rural communities initiates smoking at an earlier age. And actually, the incidence for individual living in rural areas is estimated to be 20% higher for lung cancer, while the access to quality treatment is very limited. Also, access to care insurance, a very important factor. Medicaid or the uninsured have poorer lung cancer survival than those with private health insurances. And low-income individuals who often depend on Medicaid coverage are less likely to receive appropriate treatment. So this is actually a study from a national cancer database. And these were all N0 patients, clinically N0, who underwent surgery. And those patients who were pathologically diagnosed with N1 disease were candidates for receiving adjuvant chemotherapy as a standard of care. Now, this study showed that patients living in rural areas who are uninsured or on Medicaid are at increased risk of not receiving adjuvant chemo as a standard of care. And this study actually shows it in the same state. This is the state of Nevada. And you have the northwest Nevada, and then you have the southern Nevada and rural Nevada. And the differences in survival are striking. So there are considerable survival disparities by geographic region, with southern Nevada's disproportionately affected. So geography does matter. Now, going again to insurance and governmental insurance, these are two studies that show that. The study on the left, this is for locally advanced non-small lung cancer. Again, it's from a national cancer database. And again, the study showed that patients who have no insurance or governmental insurance or lower education are at risk for not receiving guideline-concordant care. The study down here, looking at EGFR testing and treatment, it's CR database study, and it shows the same thing. Patients with no insurance, governmental insurance, or lower education do not get EGFR testing and are at risk of not being treated with targeted therapy. Now, going to another factor, environmental and occupational exposure. Again, it's important because underrepresented minority population and those with lower socioeconomic status are actually affected the most. Secondhand smoke, asbestos, chromium, arsenic, and air pollution also play a role in lung cancer risk. Genetic differences and gene-environmental interaction due to racial residential segregation. Also, socioeconomic and racial ethnic groups are often located in different residential and occupational environments. Gender groups often occupy different occupations. This is an area that is not as much studied, but biology, including genetics, is adaptive to environmental exposure. Migration, cultural beliefs, behavior, again, very hard to study. But individuals with low socioeconomic status demonstrate greater beliefs of fatalism, nihilism, and the futility of medical intervention. There is also stigma regarding smoking that perpetuates this fatalism and feeling of hesitancy in seeking medical care. And men are often socialized to demonstrate strength, autonomy, dominance, and physical aggression and to avoid the expression of emotion and vulnerability. And actually, this was something that was used for years by tobacco companies to basically promote their products. Interestingly, Hispanic women are less likely than men to smoke cigarettes. And the socioeconomic status gradient in smoking among Hispanic people is much less marked than that for black and white people. Also, Hispanic, white, Asian, and black immigrants all have lower current rates of smoking than their native-born counterparts. And socioeconomic status is less strongly related to health status and health risks among recently-arrived immigrants. Immigrants of all major racial groups also have markedly lower lung cancer mortality rates than their native-born counterparts. So this actually makes the research even more complicated. So this is a study that looked at influence of sociodemographic factors on decision in non-small lung cancer. This is stage four disease, and radiation chemotherapy was offered for palliation, but treatment refusal was associated with low neighborhood income, no insurance, Medicaid, Medicare, and other governmental insurances. Also, this is another study that looked at immunotherapy, and the same thing was found. Patients with Medicaid or without insurance were less likely to receive immunotherapy and more likely to refuse it. Actually, over 10% refused treatment. Resilience factor, I actually like these factors because, again, they're not very much studies, but they are actually exposure to protective resources and patterns of response that are mobilized to deal with potential threats that can minimize the negative effects. So for example, black women have lower rates of smoking. Immigrants have lower rates of smoking. People who attend religious gathering have lower rates of smoking. So that's why various community institutions, like churches, can be agents of change to seek solutions to local problems. So in conclusion, poverty, culture, social inequality affect underlying comorbidities and health, and lack of health insurance and limited access to care among residents of rural and most disadvantaged areas may account for their higher rates of late-stage cancer diagnosis. And clearly, America has a problem. The world has a problem. The most important thing, I think, to find a solution is actually to focus on how to do the research. And we need to look at the data and maybe re-evaluate how we're gonna define it because there is so much heterogeneity. Somebody like me, white, female, immigrant, may be very different or have a very different lung cancer risk than somebody who is native-born, white, American, even if we are both in a low socioeconomic status. So there is a need for adequate attention to be given to analytic methods needed for the quantitative study of intersectionality. And also, we have to look for the life course approach. And I think Dr. Olson is later going to touch upon table two fallacy and how to look at certain variables when we interpret the data. Thank you. We have a little bit of time. Let's do some questions. Any questions from the audience? I have one question, Horiana. So you pointed out that one of the first things that we need to do is more research in this area. Absolutely, yes. And you started with a definition of what socioeconomic disparity is. In your study of the data that you showed, is that definition routinely used to define different groups? Yes, yes. It is routinely used. And different studies, and that's why I actually showed, different studies will use the EDNAGS. Some studies will just use the NIH definition. Some studies will use a combination of the two. But yes, in general, these definitions are used in studies. And I think the problem is that if you just look at one variable, and you don't, just like with a DAG that I showed, I think there is so much overlapping that the research itself has to be, I don't want to say redone, but maybe create indicator variables for each group of people. And just look at from, you know, let's say four black men, immigrant, so. Yeah, really define and develop as homogeneous a population as possible within different groups to the comparative trials. Okay, and the gold standard, the reference standard for most studies is Caucasian? Or is, are? Right, yes. Yes, okay. Yes. And we need to figure out if that's a, because one problem with using a very homogeneous group, in my opinion, as we do a study, is that then the comparative, the comparing group is just one other group, as opposed to visualizing differences across different race, gender, education, socioeconomic status, et cetera, which makes the research very complex. I think that probably just leads into our next discussion as we talk about race and ethnicity. Any other questions? Any questions from the audience? Yes. My name's Max. I'm one of the co-members of the BFC. So I'm not very knowledgeable, so I'm including myself now. But are there socioeconomic criteria that you think could factor in for a need for lung cancer screening outside of the current smoking-induced parameters that are used in the United States? For lung, oh, absolutely. For lung cancer screening, like are there patients, like non-smokers, do you suspect or do you know that they have criteria that should qualify them for lung cancer screening? If they're not non-smoker but exposed to something else, you're thinking of? There are non-smokers that don't fit the criteria. Well, I really don't know the answer to that question. I would, you know, probably, I think just because you are on a lower socioeconomic status, I mean, if you were to know the exposure of an individual, like there is data on radon, right, as a risk factor for lung cancer, that we don't really have so much data on, you know, maybe you can make an argument. But right now, I would just probably stick with the smoking or the screening. I don't know if any of you. There was a session earlier today from Peter, actually kind of addresses, he did a super great job on saying, hey, here are all the different guidelines and what they recommend. And the US, I like the US Public Health Task Force, which said those prediction models are not suitable for individual patient-level use. I think there's a difference between predicting in populations and predicting in individuals. Even most of the socioeconomic data, the deprivation index, we actually don't measure individuals. We say you are a person who lives in a poor neighborhood or a middle-class neighborhood. You could be a rich person living in a poor neighborhood or a rich person living in a rich neighborhood. So they're models, they give us insights. But I like her answer. Keep it simple based on evidence rather than, hey, we've got to extrapolate beyond the evidence. Better to say, I don't know. All right, well, it's my pleasure to introduce the next speaker. Dr. Melinda Aldridge is an Associate Professor of Medicine, Thoracic Surgery, and Biomedical Informatics at Vanderbilt University. She's a thought leader in the area of disparities and lung cancer screening, disparities and genetic differences, as well as genetic and non-genetic risk factors in lung cancer. And she has actually never been to CHESS, which is why I'm so honored that she has agreed to do this talk for us, which is really, really fantastic. Thank you so much. Well, thank you very much for the introduction and really my pleasure to be here. So thank you. So I will be discussing racial ethnic disparities in lung cancer survival. As was just mentioned, I'm an Associate Professor here at Vanderbilt University Medical Center. Hopefully you got a chance to have a good time in our fair and fun city of Nashville and maybe experience the honky talks, which are quite unusual experience. I have no disclosures. So today I'll be discussing health equity and some of the complexities around lung cancer disparities. I'll review racial ethnic disparities in lung cancer survival, and then touch on clinical trials and screening as ways that we can improve health outcomes for lung cancer patients. So before we get started, I wanna make sure we're all on the same page and define what is health equity. So this is when everyone has a fair and just opportunity to be as healthy as possible, to really achieve their highest level of health. But it's important to keep in mind that not all health differences are health disparities. You can have, say, two equal populations and you apply some external force and you have a health difference. So let's say, for example, you have soccer players in the general population. Soccer players may have more injuries than the general population, but that is a health difference, not a health disparity. But once you apply some inequity to that, then you have a health disparity. So Healthy People 2030 defines health disparity as a health difference between population groups that's linked to social, economic, or environmental disadvantage. This often occurs across a variety of dimensions, such as race, ethnicity, sex, sexual identity, age, disability, socioeconomic status, and geography. So it's well established that there are racial disparities in smoking, lung cancer incidence, and mortality. A couple of years ago, we pulled together some data and published in the annals on smoking prevalence, incidence, and lung cancer mortality across different racial ethnic groups, white, black, Hispanic, American Indian, Alaska Native, and Asian Pacific Islanders. And you can clearly see from the, let's see if I can get my pointer, that the prevalence of smoking is lowest among Asian Pacific Islanders and highest among American Indian, Alaska Natives. We see that racial disparities occur with lung cancer incidence, with Hispanics having the lowest incidence, and white and black populations having the highest. And that same pattern is mirrored with mortality. So well established disparities across smoking and lung cancer. When we break that down and look at it by stage, we see those same disparities, where in fact, black Americans have lower survival and are more likely to be diagnosed at distance lung cancer stage. So on the left, you have five-year survival, and you can clearly see that black populations, as shown in the blue, even though they're diagnosed at localized stage, still have a lower five-year survival. We see that then as well with stage of diagnosis, where black populations are less likely to be diagnosed at localized stage and more likely to be diagnosed at distance stage. We see similar patterns occur in Hispanic populations. Here on the left, we have stage of disease with Hispanics shown in purple, and less likely to be diagnosed at localized stage compared to non-Hispanic whites. And we see that reflected in the survival numbers on the right, where in particular, Hispanic males have a lower five-year survival compared to non-Hispanic whites. It looks like Hispanic women and non-Hispanic white women have essentially equal lung cancer survival. When we look at incidence, we see a different pattern, where it's really inconsistent with tobacco patterns. These are some eloquent data published by Dan Stram and colleagues, where they've identified that the risk of lung cancer is higher in black Americans than white Americans. But in particular, this elevated risk occurs for populations that have low smoking and smoking less than pack a day. In particular, considered to smoke 10 or fewer cigarettes per day, is where you see this elevated racial disparity. And this is perplexing because black Americans tend to smoke fewer cigarettes per day. They also begin smoking later in life, and yet they have an earlier age of diagnosis and take in higher levels of nicotine. And it's this particular complexity that we are trying to understand in our current work. So as I said, racial disparities in lung cancer are complex, can occur across a number of factors that are social, environmental, biological, whether this is health literacy, access to care, belief systems, air pollution, geographic location, metabolism of nicotine, et cetera. But underlying all of this, it's really racism and discrimination that lead to structural inequities and societal injustices that lead to these racial disparities. So really a multifactorial system that leads to these complexities. So we sought to understand what are the factors that are really driving racial disparities in lung cancer survival. So we undertook this in a prospective observational cohort called the Southern Community Cohort Study. This is a population that was recruited from community health centers across the Southeast, such as this location here in Orangeburg, South Carolina. And as expected, based on the national data, we find that black individuals were diagnosed more often with distant-stage lung cancer as shown in the black, so the black individuals as shown in the purple. And this was reflected in all-cause mortality where we observed significant racial disparities in mortality among lung cancer cases, where the median survival time in white persons was longer than the median survival among black persons. So over the last couple of years, there have been many discussions around race, ethnicity, and ancestry in research and clinical medicine. There have been a number of opinion pieces that have been published, recommendations published by NIH from NHLBI, as well as the National Academies that has weighed in really providing recommended guidelines for how we consider race, ethnicity, and ancestry, both in our research and our clinical medicine. And yet we know that there is importance in population diversity. This is just an example with the EGFR mutation, which has a wide variation across the globe where you have low frequencies among populations in Europe and in Africa, and much higher frequencies in South American populations. So clearly demonstrating the importance of population diversity and the potential clinical impact that this can have. So we know that diverse populations are really a rich resource to study for complex disease and really understanding the architecture of lung cancer. We know historically that populations migrated out of Africa across the Bering Strait. We know that there was also the forced transport of African slaves from Africa to the Americas. So you have these populations that were previously isolated and have now begun mixing and really forming the diversity that we know today. So with this, we can genetically infer ancestry and use this as a potential biomarker to help us disentangle racial disparities. So using genotyping alone, you can see that populations tend to cluster based on their continental origin. The bar plot shown here represents over 1,000 individuals, each vertical bar is an individual. You can see that populations cluster together based on where they live. And so we can infer this genetic ancestry, and we did this in the Southern Community Cohort within our self-reported black and white lung cancer cases. And shown on the left are the self-reported black individuals, on the right are the self-reported whites and you can see there's a variation in their African and European ancestry among these individuals. And so we investigated the relationship between genetic ancestry and lung cancer survival and found it is not due to ancestral background. That really, it's stage of diagnosis and treatment that are predictors of lung cancer survival. So let me just walk you through this model here and the figure where we looked at the area under the curve and put all of our factors into our Cox proportional hazard and find an area under the curve of nearly 0.8. We then remove from that model our ancestry, our genetic ancestry, and see no change in that AUC. But when we remove stage and treatment, as shown in the blue and the green lines, we get a marked shift in that area under the curve, suggesting that stage of diagnosis and treatment are really what are driving these disparities in lung cancer survival. And this is where we can then intervene and really begin to make an impact on outcomes. And then just to show you that, what that looks like in terms of the hazard ratio, we see that genetic ancestry is independent of lung cancer survival with wide confidence intervals around the null value of one. And so we all know that precision oncology trials are an opportunity to really have improved outcomes and offer improved survival, but there's really limited diversity within these trials. This is a figure from the recent AACR Cancer Disparities Report, showing an overwhelming representation of white individuals in these clinical trials. An under-representation of Hispanic, black, and American Indian Alaska natives. We also see that Asian populations are over-represented in clinical trials. When we break that down and look at tumor type, we see those same patterns emerge with non-Hispanic white and Asian populations being over-represented in lung trials, and black and Hispanic populations under-represented in lung trials. This last year, we looked at small cell lung cancer trials in particular, and found the same patterns emerge as well, with Asian and white individuals over-represented in the clinical trials. And unless we begin to include these diverse populations in our research and our trials, we will not be able to offer the same precision medicine opportunities. The good news is that lung cancer survival is improving with screening. The first USPSTF guidelines that were introduced in 2013 led to the shift of the change in stage of disease as expected. So with a reduction in distant stage diagnoses, and an increase in localized or early stage disease. And along with that, we're beginning to see improvements in lung cancer survival, which is the good news. The challenges is that trial that the guidelines are based on had really low racial diversity, almost overwhelmingly white, with nearly 91% white individuals. As a result of this trial, we now have screening with low-dose CT. And so we sought to look at whether or not, how these guidelines play out in African-American adult smoking individuals, given what we know about some of the tobacco behaviors that vary across populations. And we found that there were racial differences in the age of diagnosis in cigarette smoking behaviors. The two things that are included for the lung cancer screening guidelines. So we find that African-Americans were diagnosed at age 59 versus in whites, it was age 64. And African-Americans tend to smoke half a pack a day, but whites tend to smoke a pack a day. So striking racial differences here. And so this plays out where African-Americans are diagnosed with lung cancer, or African-Americans diagnosed with lung cancer are less likely to be eligible for lung cancer screening. Nearly 70% of African-Americans diagnosed with lung cancer were not eligible for screening, and really a missed opportunity. So our solution was to redefine the lung screening eligibility criteria to be more equitable. We proposed that the guidelines be shifted to 20-pack years of smoking, as well as reducing to age 50. And we actually proposed it was only to be done for African-Americans and not for white individuals. And so the criteria, what exists today as shown here, probably you're aware, age 50 to 80, a 20-pack year history, and if a former smoker having smoked in the prior 15 years. However, the guidelines were implemented across all race ethnicities. And so we still see that there are racial disparities that remain under the current guidelines, no matter how you look at those outcomes, whether it's life years gained or number needed to screen, et cetera. So definitely racial disparities that continue to exist. So one way that we can begin to also make improvements in lung cancer survival is really through educating communities. This is an infographic my group designed in partnership with a community advisory board to really raise awareness about the availability of lung cancer screening. So what are some key takeaways to address racial ethnic disparities in lung cancer survival? We need to increase the ancestral diversity in our research and our clinical trials. We also need to make sure we report race ethnicity. It is often missing from clinical trials data. It's really important to be able to understand how we are beginning to shift and make improvements. We need to treat the underlying causes and address structural inequities, such as access to healthcare and treatment, including also eligibility and access to lung cancer screening. And finally, importantly, we need to engage with communities and really help to inform our research, but also raise awareness about availability of screening and treatments. And with that, I'll thank the team and the funding and be happy to take any questions. Thank you. That was such a fantastic and enlightening talk. Yes. Hi, welcome to CHESS. We're so happy to have you. Perhaps you're the reason that CHESS is in Nashville this year because we had to come to you if you weren't gonna come to us. I don't know if you've heard, but it's gonna be in Hawaii next year. So you should just start now and keep going. My name is Abby Vigneault. I'm at the University of Minnesota. Thank you for all of your great work. I was wondering if you would go back and look at maybe that first table that you had with the racial disparities. Sorry, all the way back to the beginning. Oh, that's okay. Because this is something you are probably aware of. Yeah, that one. So I practice in Minnesota. I do work with indigenous populations there. One of the things that a lot of these national data sources, because we all know whites and blacks nationally, there's a big difference there. And then Hispanics and some of the other ones, it kind of seems like maybe they're about even or there's not huge disparities. But that the national numbers really muddy the water of what's happening in pockets and places, right? So like in Minnesota, we have two to three times the death of indigenous people compared to white Minnesotans. There's a lot of tobacco use in the indigenous populations there, which of course is due to cultural and historical reasons. You probably know this, David, and I think in Texas, for example, not all Hispanics are the same, right? So I think Mexicans have much higher rates of tobacco-related disease and lung cancer than others. So the thing that sometimes really bothers me about these is that they're national sources of data and we just can't understand these different groups that fit under these huge umbrellas, but it's like an untold story, it seems, in all the disparities discussions. So I just wanted to unearth that for everyone here. I appreciate your comment, and I actually debated about adding that in. So I appreciate your, I absolutely 100% agree with you. Hispanics are a very diverse population. Puerto Ricans, for example, have much higher African ancestry and just very different populations. Same with Asian Pacific Islanders. There is a lot, a lot of diversity there. So I kind of agree, we need to unpack these in much more detail. Black Americans versus maybe black immigrants. So a lot of complexities that I think are just really missing in a lot of our national data, and we really need to be able to do a better job at, so I fully agree with you, so. I have a question. Dr. Aldrich, it was very interesting to see your research in terms of looking at people who get included into trials. Do you have any solution for that? Because when, if a clinical trialist goes out to recruit patients, there are very few trials that, in their protocol, say no African American patient to be included. But then why is it, and I can think of many reasons, one of it would probably be systematic racism and the history of that. But what are the things that would be the cause of such small number of African American population in biomarker studies, clinical trials for different actionable targeted mutations, et cetera, from the patient as well as medical professional team? Yeah, it's an area I probably know a little less well, but I can certainly speak to some of the, and we actually have a grant right now where we're starting to try to understand that for lung cancer. Certainly, if people are asked and invited to join trials, then they will participate. I don't think we're doing as good of a job as asking and reaching out to communities where we need to be. So I think that's one barrier. I think it's having trust, and we trust in the provider that's doing the asking and really building those relationships. And so I think there are a number of barriers that we can overcome just by going to communities and really engaging with communities early on. A lot of these communities are willing to participate. We just aren't doing a very good job at asking them to participate. So, yeah. Thank you. Any other questions? All right, thank you very much for a wonderful talk. And it's my pleasure to introduce Dr. David Oust. David is a professor of medicine at MD Anderson Cancer Center. He has published many landmark studies in the area of lung cancer, both management, diagnosis, staging, and he's editor-in-chief of the Journal of Oncology and International Hormonology. He's one of my personal mentors and the reason I decided to get my MPH, he was like, you should get yourself trained. And I can't wait to hear him talk about healthcare disparities and social determinants of cancer care. Thank you, Samira, and thank you to the audience. I know on the last presentation of the last session of the day, I appreciate your patience. I'm gonna try to go fast so we can get as much time for discussion. These are my disclosures. Most of them really don't apply, other than to say these robots are expensive. All the things here are expensive. And how does cost interact with social determinants of health, however you define it, to impact outcome? I wanna try to unpack some of that. I've learned a lot from Samira inviting me and talking to her. And my conclusion is it's even more complex and difficult to solve than I thought. And I'll try to share some of the things I've learned along the way and to have interaction with our community so we can move forward. So my first goal will be to clarify definitions so that we can communicate as a community of scholars to move things forward. One of my goals is to distinguish how social determinants of health impact population level mortality through both two separate aspects which are difficult to dissect, but vitally important, because I'm going to assert that a lot of people in the health disparities realm confound these things and get it wrong. Now that's a pretty aggressive assertion. So I'm not saying that they're wrong because they haven't thought about it, but because it's such a hard problem. And specifically, I wanna look at how do social determinants of health interact with mortality through both incidents and healthcare delivery. It's not always easy to dissect those two elements out, but it would be easy to make a mistake of inference because our data sets aren't that great. And we heard some of it, just, hey, you know, not every Hispanic person's the same. You know, not every African American is the same. So our markers are crude. How can we get better? And then I wanna look at how increasing cost of healthcare really modifies the relationship between social determinants of health and mortality and where we as physicians fit in. Dr. Grosser covered this very well. These are the social determinants of health as defined by the US Department of Health and Human Services but I think it's a good jumping off point because as Dr. Shoji said, you know, we need some standard definitions and it would be useful when comparing different papers to pay attention to those details to make sure we at least are comparing apples to apples. So we're looking at education, economic stability, social and community aspects. There was a very nice graphic of how people cluster. You know, I love that thing. Healthcare access and quality go along with this and then the neighbor and built environment. Now, of course, we all know this, healthcare costs are going up even after you adjust for 2019 dollars, it's going up and up. We're spending upwards of $12,000 per person and it's going up. We'd like to be more efficient. We have limited resources. We've got to spend it wisely. And Dr. Mazzone had a very nice slide of, for healthcare screening, for those who are interested, it's about arguably net benefit, you know, while still maintaining an aspect of justice, however you wanna define it. Not everyone's definition of justice will be concordant with everyone else's, but however you wanna define it, think about how you quantify it. Now, healthcare effectiveness, cost, social determinants of health, all interact in a complex relationship. Like Dr. Gross who said that, she had that beautiful DAG. A DAG is a directed acyclic graph for studying causality. I didn't know that when I was a fellow. Frankly, I didn't know when I was attending. But super important if we wanna dissect this. And when we think about ourselves as physicians and social determinants of health, the majority of healthcare costs are for treatment. Secondarily, they're for what I do, diagnosis and staging. Some is for prevention, but that's the lowest dollar amount. But the social determinants of health impact also incidents of disease as well as diagnosis and treatment in a very complex way. And Dr. Gross who hinted at it, it begins way, way back in childhood, in early adulthood, where you grew up. How can I know that it was at this stage of your life when that social determinant of health impacted your outcome? Was it when you were 12 or when you were 50, when you were diagnosed? I can't always dissect that out. But the majority of healthcare costs is in our end, treatment, secondarily diagnosis, last is prevention. So let's think about how this interacts. And why does it matter? Well, let's think about healthcare delivery, meaning once you have the disease, how do social determinants of health impact it? Well, first and foremost, you have to have an effective treatment. And the social determinants of health impact that. So in the early HIV era, when I was a resident or student actually, there was no treatment. Now, rich or poor, you still suffered bad outcomes. But if you had higher socioeconomic status, you did better. But notice that there are other factors like concurrent comorbidities in the population, such as TB, which would impact that. So you can see how complex that is. And it's not just social determinants of health, but it's a combination of the cost, which is really access to care in a sense, as well as the effectiveness of the treatments. Because in the early HIV era, as I'm gonna show you, well, when there are no good effective treatments, we all suffer. But of course, you're better off rich than poor. You didn't need to go to Harvard to figure that one out, although you are poorer after you go because you have to pay. So it's both cost and effectiveness. Here's the HIV data. I'm gonna walk you through it. On the left is the population, age-adjusted death rate per 100,000 people. And the red line is the lowest socioeconomic quintile. The middle gray is the middle three quintiles. And the black line is your poorest quintile. Sorry, is your highest, is your richest line. And if you look at the left-hand graph, you see that in the pre-treatment era, prior to HAART, which is that dotted line, indeed, if you had better sociodeterminants of health, you did better. But there was no real great treatment back then. So why did you do better? Well, it was your other comorbidities, your access to supportive care back then. And notice when we're looking at the left graph, after you introduce that, all the lines get better. And if you look at the difference between the graphs, that's the risk difference at the population level, they're getting closer. Okay, that seems pretty much straightforward. On the right-hand graph, we have the relative risk. By sociodeterminants of health. The black line, which is your highest quintile, is your reference. Everything else is relative to that. Notice, prior to the development of highly effective antiretroviral therapy, the relative risk of the poorest quintile is relatively fixed. It's two. Now, the risk difference, as you can see from the left, was actually increasing. It was getting worse and worse. After you introduced highly effective antiretroviral therapy, if you looked at the relative risk, you might conclude that, wow, it got worse. The risk ratios are increasing. But you look at the left-hand graph and you say, but the risk differences are getting smaller. So all I want you to encourage you to think about is think about ratios and risk difference and pay attention to the risk difference. A lot of times, we don't think about that. But the left-hand graph clearly shows risk differences are improving, but sociodeterminants of health are still there. But arguably, this is one of the great triumphs of modern medicine in the last millennium. You know, like, hey, this was a huge problem and it's not gone, but we're getting better. And the risk difference for the sociodeterminants of health have gone down with effective treatment. So if you even thought about that, you say, are inequalities across socioeconomic status increasing, decreasing, or staying the same? It depends on your measure. I would say in that example, the relative one is the risk differences. But they interact both on incidence of HIV as well as treatment. So we're gonna do lung cancer next. I want you to think, hey, sociodeterminants of health impact both the incidence of cancer as well as how well we treat it. But those are two different things and they are tough to dissect as well as cost. So mortality changes due to incidence and our changes in treatment. And I'm gonna use non-small cell lung cancer, not as good as HIV, not even remotely close as the example of progress. And I'm gonna use small cell lung cancer as absence of progress, where we'll see no effect. You know, you're pretty much going to do badly no matter what. But it will be useful illustratively. This is from the New England Journal. The blue is the observed incidence and then the red is the incidence-based mortality from each type of lung cancer. On the left are men, on the right are women. And women, as usual, do much better than the men. But notice that the incidence of lung cancer is dropping because of changes in the population in smoking and presumably other factors. Now, those changes may vary by sociodeterminants of health, but our healthcare system, when we look at cost and access to care, really becomes involved only for those who get the disease. Which, you know, if you're unfortunate, but if you look at the population level, deaths per 100,000 of that socioeconomic status, the incidence is a big factor. But once you have it, if you're looking at the two-year survival, that's a different number. So we have to really think hard about that. Was it the social determinants of health for the population level? Could have had huge impacts in the past that are very tough to dissect out. Now, if we break it down by race and ethnicity, this is the same New England Journal. We see here that the, you know, arguably the best group here in terms of two-year survival is Asian men and women, and African, sorry, non-Hispanic blacks are the lowest. There's, you know, some degree of statistical difference here, but it really depends on the population you're looking at, and these are relatively crude markers. But it does say maybe there are other social determinants of health which are not necessarily well captured in this particular study. Not that it's bad, it's just different. Let's contrast that with small cell lung cancer. Here we can see on the bottom panel the two-year survival line, just flat, right? We haven't made that much progress there. Even though the incidence is going down because smoking changes. So you might see different changes in incidence because within certain socioeconomic groups, smoking cessation may improve. It was a very interesting thing that, and there are differences within genders and within cultures. So you could see a population-level change even without any change in the efficacy of treatment or substantive change. So population-level mortality from non-small cell lung cancer did decrease with EGFR inhibitors. Survival after diagnosis improved substantially, but you see it was already decreasing due to changes in incidence. So dissecting out those and the impact of social determinants of health is very tough. Now, where does cost come in? Well, if you're a small cell, not as big a deal, but once you have effective treatment, analogous to HIV, then it does start to make a difference. Now, here's data over time, looking at all cancers. I'll show you the lung cancer data. And each line is a quintile. The one at the top is the highest socioeconomic status quintile. And you can see from 1993 to 2006, it's improving, but all the lines are improving. But the gap between the highest line and the lowest line is actually the spread has improved, even though all the lines are improving. So the difference in survival has changed over time, has increased, even though all ships are indeed rising in this graphic over those years. And this is 2006, when we didn't really have all those EGFR interventions. So social determinants of health here are changing. And you might say, hey, this is access to care, American access to care, you know, Medicaid, poor, non-insured, bad, except that this is Canadian data. So even then, but it could be anything in that long life cycle, right? It's population level effects are subtle. You can't get rid of them. Meaning, hey, you know, you smoke more, but you also have high blood pressure, you have heart disease, you're overweight. Those things all trended impact outcome. Here it is by different cancers. And of course, different cancers have different rates of progress in terms of their effectiveness. Now, over the range of years that they looked here, there wasn't that much progress in lung cancer. You see the top quintile there with that little itty bit of difference, but in the risk difference scale, we're talking about tiny fractions, you know, not a huge effect on the risk difference scale. Now, of course, there are limitations there. That's only for patients with cancer. So it eliminates the effect of social determinants of health on incidents, which drive the population level dynamic. Maybe, I don't know the Canadian system, maybe it's really not equal access to care. But it's just the concept that social determinants of effect have many impacts on cancer care, through the comorbidities, right? So all of that, but if we fail to control for that, it makes measurement and quantification difficult. So what about small cell? This is SEER data we're going to look at over three decades from 1983 to 2012, socioeconomic status by county. Remember, it's not individual level data, it's a surrogate. But you don't need to look at long at this. It's like, okay, this is by race, not a whole lot of absolute difference. This is by low, medium or high poverty, looking not at individuals, but at the county marker of where you live. And yeah, you can see that bottom one. Oh, there's a 0.0001, but that's because of the high sample size. The absolute magnitude of the difference is modest because there really hasn't been revolutionary, good treatment changes between the groups. So risk difference, not just looking at some P value and saying, yeah, P value is significant, but you really look at clinical significance. And last, I would like to close up with this JAMA article. It highlights a lot of the things that Dr. Grosu mentioned. This is the association of race, socioeconomic factors and treatment characteristics with overall survival, again, in limited stage small cell lung cancer. This is stage one, small cell, which is pretty rare. You look at the sample sizes, it's pretty rare. African-Americans are the blue line. The orange line is whites. And the other groups are kind of sparse. Sample sizes for Asians and Native Americans, Hispanics are really tiny in this cohort. But that's stage one, small cell, which is rare. This is stage two. You see the group here, no significant difference. And this is stage three where there is a difference. And it turns out in the Cox model, after adjustment, African-Americans in this particular study actually do better. But this is with multivariable model, looking at sex, lower average county income. It's not individual income. Education measured at the county level, not at the individual level. Insurance, histology, lymph node sample and cancer stage. And this is your nice, complicated model. That's for myself to make sure I stay on time. And this is my closing message. So I say, wow, I got my study. I'm interested in this variable. I see that I want to be well-educated and rich. That's why you go to college to learn those things. In this case, in this data set, I would like to be either Asian or African-American. But the fallacy is the mutual adjustment fallacy, is that I can interpret these coefficients as being true for every variable on that table. That is not true. Now, let's just think about how many tables we look at with a multivariate model. And we look at the coefficients. And we believe that that coefficient reflects the association of that variable with the outcome of interest after adjustment for all the other variables in the model. But it's actually not true. This is the table 2 fallacy, the mutual adjustment fallacy. I'll stay after it for anyone who wants to talk about it. And I learned it in school. Horiana tried to tell me more about it. And I'm like, oh, I don't need that. And then when I started looking at this, I'm like, oh, I don't really need to learn that because there are mediators. So depending on what your exposure variable of interest is, for the gentleman who wanted to study screening, that coefficient is different. But prediction is different than causality. If you're doing an observational study for inference of causation, you often will require more than one equation. So you have to draw your directed acyclic graph, like she said, say, this is my exposure variable. This is my equation. For this other variable, it may be that this coefficient does not mean what you think it means. And that table 2 fallacy is rampant in medicine. Think about your journal clubs. It's very true for observational data. And this is what the fallacy is. It's that C is a confounder. My exposure is E. O is the outcome. It's that when I get those coefficients, I put them all on the table, that the coefficient of C, the confounder, is, I believe, incorrectly, that it estimates the total effect of C on the outcome. That is not true. It does can represent the direct effect of C on the outcome. So I'd encourage you to look that up. In conclusion, cost interacts with social determinants of health. It's cost and effectiveness which are effect modifiers here. Access to care is related but different. We saw that rural care is different. Look at the absolute risk difference, not just the ratios. Risk difference, we know that as clinicians. But we get confounded by these complicated ratios and P values, too much attention to P values, too little attention to absolute risk differences. The social determinants of health impact total population mortality, both via incidence as well as survival amongst those with the disease, which is the health care delivery. Making inferences, as Dr. Grossi said, super hard because some of these things, they begin so long ago. How do I know where it was? And if you're going to go down that road, beware of the mutual adjustment fallacy. And then we can all move forward as a community to be more like the HIV model. Well, I may not know all the social determinants of health, but I know that risk difference has shrunk. It's gotten better. That's where we need to get. The other extreme is the small cell, where we made no progress and things are terrible for everyone. And that's what I'll leave you with. Thank you very much.

socioeconomic factors

social determinants of health

lung cancer survival

socioeconomic disparities

access to care

racial disparities

ethnic disparities

lung cancer screening guidelines

healthcare delivery

HIV treatment