Jan 10, 2022
This week's Feature is a Panel Discussion. Circulation invited the Young Investigator Finalists who had a Simultaneous Publication for AHA's 2021 Sessions. Please join authors Amgad Mentias, Matthew Burrage, Shaan Khurshid, Sevedeh Maram Zekavat, and Neel Butala as they discuss their articles.
Dr. Greg Hundley:
Welcome listeners to this very special January 11th issue of Circulation on the Run. And I'm going to tell you why it's special. I'm Dr. Greg Hundley, director of the Pauley Heart Center at VCU Health in Richmond, Virginia and also associate editor at Circulation. Why is this issue special? Because we have the opportunity to speak with finalists for several of the awards that were presented or that these investigators presented for at the American Heart Association sessions. And so we have with us today, five early stage investigators and we are going to hear about each of their research projects and the manuscripts that are coming out and are published in this issue. Want to welcome all five of you and we'll introduce one at a time as we work through their research projects. And the first is Dr. Amgad Mentias from Cleveland Clinic and he was in the session competing for the Elizabeth Barrett-Connor Research Award for early career investigators in training. Welcome Amgad. And we'll start with you. Can you tell us a little bit about the background for your study and what was the hypothesis that you wanted to address?
Dr. Amgad Mentias:
Hi, Dr. Hundley. How are you? Thank you so much for inviting me today. A little background, we know that community economic distress affects outcomes in patients with heart failure. It actually affects both short term and long-term outcomes. What was not studied on a nation level before is how is that impact different or if it's actually different between different races. In White patients and Black patients and Hispanic patients, what would be the differentiated effect of community economic distress on their short and long-term outcomes with heart failure after a heart failure admission?
Dr. Greg Hundley:
Absolutely. And so that hypothesis that you were going to address, what were you hypothesizing here?
Dr. Amgad Mentias:
We were hypothesizing that each race has probably some shared risk factors but also some specific risk factors. We were hypothesizing that the community level economic distress effect on heart failure outcomes is not homogeneous or not exactly the same between all races.
Dr. Greg Hundley:
Very nice. And so what was your study design? And describe a little bit for us the study population.
Dr. Amgad Mentias:
We included the Medicare population. We included patients who were admitted with a primary diagnosis of heart failure from the years 2014 till 2019. We included patients from Black and White and Hispanic races or ethnicities. And we only included the first admission for a patient if the patient was admitted more than once during these years. That was the study population. And we had about 1.6 million White patients and about 205,000 Black patients and around 89,000 Hispanic patients.
Dr. Greg Hundley:
Great. And so was this a cohort design?
Dr. Amgad Mentias:
Yes, it was a cohort design. The study start date was when they were admitted to the hospital for our mortality outcome and when they were discharged from the hospital for readmission outcomes. And then we followed them in time up to one year.
Dr. Greg Hundley:
Very nice. And so tell us, what did you find?
Dr. Amgad Mentias:
Our primary exposure of interest, like we said, was economic distress and race and our primary outcome was three things, mortality, the second thing is readmission burden, which is number of admissions over time the patient is alive and third thing is home time how many days the patient spends at home, out of the hospital and out of skilled nursing facilities and LTACs. And we looked at these outcomes at three different points, at 30 days, at six months and at one year follow up. Initially we did an interaction and that is to see whether our hypothesis, that the effect is different or not between a race and term and the economic distress term. And the interaction was significant, in all three outcomes. Then we went deeper and we started to study each race separately and see how economic distress affects their outcomes. We defined economic distress by distress score called the distressed community index, which is a composite measure of seven things, including education of the people in the zip code, unemployment, poverty rate and income in the zip code compared to the state level and that stuff.
Dr. Amgad Mentias:
We actually found that in White patients, economic distress was actually associated with adverse outcomes in the short and long term. In Black patients, it was affecting the outcomes more robust and more evident in the long term, not in the short term. We also found the geographic location and their approximate location, whether it's urban or rural, residential zip code also affected outcomes. We found that in all races being in a rural distressed community had the highest posterity and the highest admission burden and the worst home time compared to other communities. In fact, people in distressed urban communities had comparable outcomes. People in rural, non-distressed communities had comparable outcomes to urban distressed. We found that the rural location and approximate location near resources affected outcomes in all races but specifically also in Black patients.
Dr. Greg Hundley:
Very nice. And how do you put your results in context with others that are doing research in this area?
Dr. Amgad Mentias:
We show that the interplay between economic distress and societal factors and different things for heart failure is very complex and there is a complex interplay between different factors. I think it's very important for health policies that targeting improvements in community with economic distress and access to care, they are key to improving outcomes and reducing racial disparities among patients with heart failure.
Dr. Greg Hundley:
Beautiful. Well, Amgad, we want to congratulate you on this just excellent work in identifying associations between community level economic distress and risk of adverse outcomes across different race ethnic groups. Congratulations to you.
Dr. Greg Hundley:
Well listeners, next we're going to turn to Matthew Burrage from University of Oxford. And Matt was a finalist for the Melvin Judkins Early Career Clinical Investigator Award. Matt, just like with Amgad, could you tell us a little bit about the background that went into your research project? And what was the hypothesis that you wanted to address?
Dr. Matthew Burrage:
Yeah, certainly. And thank you very much for the invitation to take part in this discussion. Really it's a pleasure to be here. The inspiration for this study was really driven by the difficulties that we've been having in trying phenotype heart failure with preserved ejection fraction or HFpEF, given that until very recently, there were really no therapeutic agents that have significantly improved outcomes for this population. This is despite the fact that around half of all heart failure is classified as HFpEF. And so the thought is that this is a very heterogeneous population but when you dig down into the physiology, there seems to be a central mechanism which involves impairment of myocardial relaxation and in a subsequent rise in intracardiac filling pressures. And this is something that's often unasked by exercise and typically this then results in pulmonary congestion and symptoms of breathlessness. And so some recent translational studies suggests that abnormal cardiac mitochondrial function and energetics may be a unifying feature in the pathogenesis of HFpEF.
Dr. Matthew Burrage:
And given that we know myocardial contraction is dependent on cardiac energy metabolism and that diastolic relaxation is even more energy dependent, we hypothesized that impairment of myocardial energetics may underpin a lot of the physiological changes in the heart that occur during exercise and thus could potentially present a metabolic basis that underlies symptoms in patients with HFpEF, with the hope that this could then lead to new translational drug targets for HFpEF in the future. But then alongside this as well, there's been some really pivotal work on pulmonary congestion during exercise in HFpEF that's been led by Barry Borlaug's group at the Mayo Clinic as the main determinant of patients' symptoms. This has been very well validated against invasive hemodynamics. The second component of our study was to see if we could noninvasively assess pulmonary congestion during exercise and HFpEF. And so to do this, we developed and implemented a new MRI sequence that could quantitatively assess changes in lung water.
Dr Greg Hundley:
Very nice. How did you address the hypothesis in terms of your study design and your methodology?
Dr. Matthew Burrage:
This was a prospective study that followed essentially a basket trial design, where we recruited four distinct groups of participants that were felt to really encompass the spectrum of worsening diastolic dysfunction in HFpEF, which was based on clinical scoring systems, blood biomarkers and echocardiography. We recruited 43 participants split across this group and so we had a cohort of age matched, healthy controls. We had patients with cardiometabolic risk factors for HFpEF like diabetes and obesity who were included essentially if you think of it like a pre-HFpEF group, patients with carefully clinically phenotyped HFpEF and then a cohort of patients with cardiac amyloidosis. And the amyloid group was recruited really as a positive control, that the proof of principle lung imaging sequences, as the presence of restrictive physiology in those patients meant they would be the group that would be far most likely to develop pulmonary congestion during exercise.
Dr. Matthew Burrage:
And so each participant underwent blood sampling, a targeted echocardiogram, they had magnetic resonance spectroscopy to assess myocardial energetics and cardiac metabolism. We do this by measuring the phosphocreatine to ATP ratio and also a cardiopulmonary exercise MRI. The exercise protocol for the study was a fixed low intensity workload of 20 Watts for six minutes with the patient supine within the MRI scanner using an ergometer. And then during exercise, we did whole heart free breathing cine stacks to assess cardiac volumes at rest and exercise, as well as performing our custom proton density weighted lung imaging sequence to look at changes in pulmonary congestion. And if you're interested, the whole rest and stress protocol together can be done within about 15 minutes.
Dr. Greg Hundley:
Wow. Boy, very exciting. Exercise during an MRI scan. Matt, we're very anxious to hear, what did you find?
Dr. Matthew Burrage:
The key findings from this study is that there really is a clear gradient myocardial energetic impairment that exists across the spectrum of diastolic dysfunction and HFpEF phenotypes of increasing clinical severity and worsening diastolic function. And this gradient of impaired myocardial energetics was associated with progressively abnormal exercise responses compared to normal physiology in the age matched controls. And so a greater degree of energetic deficit was linked to impaired left ventricular systolic and diastolic functional reserve. It was also linked to altered right ventricular reserve and abnormal RV-PA coupling and also to exercise induced pulmonary congestion.
Dr. Matthew Burrage:
And we also showed that the pulmonary congestion or changes in lung water could be quantitatively assessed using our new proton density lung imaging sequence and that there is a subgroup of patients with HFpEF who do demonstrate transient pulmonary congestion during exercise that we can assess noninvasively. Overall, the findings suggest a pathway where impaired energetics are linked to patient symptoms and they do this by limiting cardiac reserve during exercise and promoting pulmonary congestion. There seems to be a really important role of resting cardiac energetics in signaling the abnormal ability of the heart to perform high energy consuming processes like active diastolic relaxation and augmentation of contractility and then this leads of course to the downstream effects that we see.
Dr. Greg Hundley:
Very nice. And you were able to even also observe the lung water. It sounds like, help us put this in context for our listeners of how do your results really advance some of the understanding of the pathophysiology of heart failure with preserved ejection fraction?
Dr. Matthew Burrage:
I think one of the key impacts of this study is the fact that the heterogeneity of clinical HFpEF syndromes has been such a major challenge to efforts to develop new therapies to improve symptoms and prognosis in these patients. Pathophysiological phenotyping may represent an important step towards targeting the right therapies to the right patients and specifically targeting myocardial energy metabolism may be a promising therapeutic strategy to improve cardiac reserve and potentially reduce pulmonary congestion in patients with HFpEF. And this really builds on all the translational studies that exist today and have gone before it.
Dr. Matthew Burrage:
Hopefully the mechanistic insights that we get from this could lead to some new translational drug targets, which can be tested against myocardial energetics in patients to see if this metabolic substrate is modifiable. And if this then leads to improvements in symptoms and outcomes. The second aspect very quickly, relates to the evaluation of patients with breathlessness, particularly because invasive hemodynamic assessments may not be possible in all patients who have breathlessness on exertion. The lung water imaging represents a potentially new diagnostic tool that can help to differentiate HFpEF from other causes of dyspnea and I think this is some something that may have a lot of direct clinical applications for patient diagnostics for a wide range of conditions in future.
Dr. Greg Hundley:
Very nice. Well, thanks so much for what outstanding work, identifying this myocardial energetic deficit and then linking that to both cardiac performance, as well as the development of pulmonary congestion.
Dr. Greg Hundley:
Well listeners, we are going to switch to our third author today, Dr. Shaan Khurshid from Mass General and Shaan was a finalist for the Samuel A. Levine Early Career Investigator Award. And so, Shaan, can you give us a little bit of the background information pertaining to your study? And what was the hypothesis that you wanted to address?
Dr. Shaan Khurshid:
Thanks for having me, it's a pleasure to be here. A little bit of background that predicting the risk of atrial fibrillation or AF, may increase the efficiency of AF screening and effectively prioritize individuals for preventive interventions that are designed to reduce the risk of incident AF in the first place. And to that end, risk of AF can be estimated with reasonable accuracy using clinical factors. We already know that. And for example, the CHARGE-AF score is a well validated score that been used in multiple settings. More recently, work suggests that artificial intelligence or AI enabled analysis of the 12 lead electrocardiogram can extract latent information that may be relevant for predicting AF risk. Past models however, have had some limitations. They've utilized very short time intervals. They have not incorporated survival time and censoring with is important for prognostic models. They are kind of a black box and therefore difficult to interpret and they haven't undergone a broad external validation.
Dr. Shaan Khurshid:
Therefore, in this current study, we sought to develop a deep learning model, utilizing the 12-lead ECG to predict risk of incident AF at five years. We call this model ECG-AI quote unquote and compared the performance of ECG-AI directly to the CHARGE-AF clinical risk score that I was mentioning. We also sought to assess a model that combines both ECG-AI and CHARGE-AF to each score alone. We hypothesized that the ECG-AI model utilizing 12-lead ECG could improve the ability to predict five year AF risk as compared to clinical risk factors alone. And we felt that such a model may have practical applications, particularly since wearable devices like smart watches are increasingly able to provide single lead ECGs.
Dr. Greg Hundley:
Really nice. Sounds like a very interesting application of artificial intelligence with electrocardiograms in assessing patients with atrial fibrillation. Can you describe for us your study population and your study design?
Dr. Shaan Khurshid:
Of course. We trained our models utilizing a retrospective cohort. The training population was 45,000 individuals receiving regular primary care at Massachusetts General Hospital or MGH. We then validated our models in three completely independent samples, an MGH internal test set, so individuals from MGH but were not included in training, a separate set of primary care patients at Brigham and Women's Hospital and the UK Biobank Prospective Cohort Study in the UK. The total population in which the models were validated was over 83,000. ECG-AI itself was trained as a convolutional neural network, which was inputted with 10 seconds of the 12-lead ECG and utilized a specialized loss in encoding function that incorporated survival time and censoring in order to produce a five year risk estimate for each individual. We trained models utilizing all ECGs available for each person but evaluated the models utilizing a single ECG alone. We compared each model, ie. ECG-AI, CH-AI and CHARGE-AF by incorporating risk estimates into analogous Cox proportional hazards model so we could compare them apples to apples and calculated traditional epidemiologic metrics of prognostic model performance, including discrimination, calibration and reclassification.
Dr. Greg Hundley:
Very nice. And so what did you find, John?
Dr. Shaan Khurshid:
From our study, we had two major findings. First, the ECG-AI model consistently discriminated five year AF risk comparably to the CHARGE-AF 11 component clinical risk score with C statistics ranging from 0.7 to 0.8. Second, the CH-AI model, which was the combination of ECG-AI and CHARGE-AF, consistently offered greater discrimination than either model alone. Both AI models were very well calibrated across the three test sets with calibration error consistently less than 1%. The ECG-AI and CHARGE-AF scores were moderately correlated, suggesting that the AI model is able to leverage clinical risk factor information extracted from the ECG, yet also adds something further.
Dr. Shaan Khurshid:
Saliency analyses, which are a method of determining which areas of the ECG are most relevant for the model's prediction, highlighted the P-wave and surrounding regions, which provides important evidence of biologic plausibility in our models. Importantly, in sub-analyses assessing the AI models, including only one lead of the 12-lead ECG, we found that model performance was similar, suggesting that AI models utilizing only single lead ECGs may also be effective. We also found that the models performed reasonably well in individuals with prevalent heart failure and stroke, which are populations in whom AF risk destination is particularly relevant.
Dr. Greg Hundley:
Very nice. And so clinically, moving forward, how do we put your results in the context of really where you see this field moving and how we might use it to identify patients at risk of atrial fibrillation?
Dr. Shaan Khurshid:
We're excited. Our work we think provides an important demonstration that ECG-AI based models can utilize a 12-lead ECG to estimate future risk of AF up to five years. And importantly, the AI models were generalizable, providing good discrimination across three large and independent datasets spanning two continents. We're most excited about this finding that models perform well when utilizing single lead ECG data alone, which has important ramifications for wearable devices. In particular, one could imagine a future application of AI in which a wearable device is able to not only be used to screen for atrial fibrillation or AF, but also stratify an individual's risk for AF utilizing ECG based analysis and therefore potentially prioritizing that individual for preventive interventions and also potentially determining how intensely to screen that individual all in a single closed group.
Dr. Greg Hundley:
Excellent. Wow, Shaan, just beautiful presentation, listeners. Really discussing how the artificial intelligence assessment of these EKGs may enable efficient quantification of the future risk of developing atrial fibrillation.
Dr. Greg Hundley:
Well listeners, we're going to turn now to our fourth presenter, Maryam Zekavat from the Bird Institute, Yale University. And Maryam was a finalist for the Genomic and Precision Medicine Council's Early Career Investigator Award. Welcome Maryam. And can you describe for us some of the background pertaining to your study? And what was the hypothesis you wanted to address?
Dr. (Sevedeh)Maryam Zekavat:
Absolutely. And thank you for the invite to be part of this podcast. The title of the work that we presented and that was published in Circulation is Deep Learning of the Retina Enables Phenome and Genome-wide Analyses of the Microvasculature. And so as a little bit of a background, we know that the microvasculature has key roles in maintenance of organ health and that microvascular disease is implicated in conditions across all organ systems. Here, to study the human microvasculature noninvasively, we used data across about a 100,000 retinal fundus photographs. And the purpose of our work was really to address two main things.
Dr. (Sevedeh)Maryam Zekavat:
First, an unbiased assessment of the phenotypes associated with the retinal microvasculature had yet to be performed and that motivated us to ask our first question, namely, what information can the retinal vasculature provide on future ocular and systemic disease risk? And then secondly, therapies such as anti-VEGF, which pharmacologically influence vascular density, are the mainstay of treatment for multiple ocular conditions, including wet AMD, proliferative diabetic retinopathy, as well as many cancers. However, an unbiased screen of genetic targets for other treatments that may influence the microvascular has yet to be performed. And so that motivated us to ask for our second question, namely, what genes influence the retinal vasculature? And so from there, I'll go to our hypothesis, which was that analyses of retinal fundus photos may enable an understanding of the connection between microvascular geometric indices, diseases and genetics.
Dr. Greg Hundley:
Very nice. And so boy, I heard you had almost a 100,000 participants involved in this study. Tell us a little bit about your study design and clarify for us, where did you get all these patients from? What was your study population?
Dr. (Sevedeh)Maryam Zekavat:
Yeah, of course. The study list we utilized the UK Biobank, which is a cohort of half a million individuals, including over a 100,000 fundus photographs from about 50,000 individuals. We first implemented deep learning to remove poor quality images and then to segment out the vasculature fundus photos. And then from there, we went on to quantify two vascular features, branching complexity as measured using fractal dimension and also vascular density. And lastly, we performed phenome and genome-wide association studies to understand how these vascular geometric indices influenced disease risk and what genetic factors influence the vasculature.
Dr. Greg Hundley:
Excellent. And tell us, what did you find?
Dr. (Sevedeh)Maryam Zekavat:
Yeah. First using deep learning, we were able to successfully perform image quality control and vessel segmentation to extract two geometric features of the retinal vasculature. Next through phenome-wide analyses, we identified that lower retinal vascular fractal dimension and density were significantly associated with higher risk for incident mortality, as well as cardiometabolic conditions, including hypertension and type 2 diabetes, heart failure and renal failure among others. And also multiple incident ocular conditions, including future risk of retinal detachment. Thirdly, genome-wide association of these two geometric indices identified seven and 13 novel loci associated with vascular fractal dimension and vascular density respectively. And these were enriched in pathways linked to angiogenesis, such as VEGF, angiopoietin and WNT signaling pathways, as well as inflammation via interleukin and cytokine signaling. And then fourth, through Mendelian randomization for genetic causal inference analysis, we identified that a genetic risk for hypertension and type 2 diabetes is associated with lower microvascular density and that a genetic risk for lower microvascular density is associated with increased risk of retinal detachment.
Dr. Greg Hundley:
Wow. Really interesting. The intersection of this beautiful phenotype characterization of the retina with this genetic information. Where do you see this research moving forward in the future?
Dr. (Sevedeh)Maryam Zekavat:
Yeah, so clinically these findings may support the use of retinal microvascular indices for risk prediction and disease monitoring of systemic and ocular conditions. And of course, further assessment of the identified biological pathways influencing the microvasculature can potentially lead to therapies for not only retinopathies but also other conditions linked to microvascular disease, including oncologic, renal and cardiovascular conditions. And more broadly, our results illustrate the potential for using deep learning on retinal imaging to understand the microvasculature with wide applications across diseases. And of course, more research is needed to evaluate the added benefit, in addition to existing clinical predictors and the feasibility for incorporation into clinical workflows.
Dr. Greg Hundley:
Just beautiful. And thank you so much Maryam and for highlighting for us, the results of your study, indicating that the retinal vasculature may serve as a biomarker for future cardiometabolic and ocular disease and provide insights on genes and biological pathways that influence microvascular indices.
Dr. Greg Hundley:
Well listeners, now we are going to turn to our last speaker today and it's Dr. Neel Butala from Mass General, Beth Israel. And he also was a finalist for the Samuel A. Levine Early Career Investigator Award. Welcome, Neel. And could you describe for us the background pertaining to your study and what hypothesis did you want to address?
Dr. Neel Butala:
I appreciate the opportunity to be here and chat with you. And so there's conflicting evidence on the optimal duration of dual anti-platelet therapy, which is DAPT, after drug eluting stent implantation. Older studies, such as the DAPT study show fewer ischemic events with more bleeding, with longer DAPT duration and our site and the guidelines for DAPT duration after PCI but newer studies show similar ischemic events and actually less bleeding with shorter DAPT, even among those with high ischemic risk at baseline. We wondered whether the DAPT study, which is the only study powered to detect ischemic endpoints and still influences a major cardiovascular guidelines still applies to contemporary practice.
Dr. Neel Butala:
And so we asked two key questions. Now, number one is a US contemporary real world population of patients receiving PCI, different from the DAPT trial population. And again, the DAPT trial enrolled between 2009 and 2011. And here we hypothesized that the populations are probably a little different. And number two, we asked how would trial treatment effects change if a real world population had been enrolled instead? And here we hypothesized that perhaps the ischemic benefit of longer DAPT would actually go away, would be similar to the newer trials that have been done.
Dr. Greg Hundley:
Very nice. And describe for us your study design and your study population.
Dr. Neel Butala:
Yeah. We compared characteristics between DAPT study patients, with those of a more contemporary real world cohort of NCDR cath PCI registry patients. And to do this, we used novel transportability methods to really create a propensity score model to predict an individual's likelihood of trial participation based on patient characteristics. And this type of propensity score model actually gives us inverse probability weights, which we used to reweight the DAPT study patients based on the distribution of characteristics in the real world patients. The intuition here is really to up weight the trial patients with characteristics more common in the real world and down weight trial patients with characteristics less common in the real world. We then compared treatment effects in the DAPT study patients with those of the reweighted DAPT study patients to understand whether DAPT study results would change if a real world population had been enrolled instead.
Dr. Greg Hundley:
Very nice. And what were the results associated with these comparisons?
Dr. Neel Butala:
First, we found that trial and real world populations were different. We found that the contemporary real world population was older, less likely to be White, more likely to have comorbidities and then more likely to present with ACS. Additionally, nearly a 100% of real world patients received a second generation drug eluting stent versus only 58% of trial patients. And then these differences led to differences in the estimated average treatment effects of DAPT, to the results of the DAPT study. And the real world treatment effect in reducing ischemic endpoints is actually no longer present but the increase in bleeding persisted. And so we found that the average top line treatment effect in the DAPT study may not be applicable to contemporary practice. We did find however, the DAPT score in the subgroup did still identify subsets of patients who may benefit from prolonged DAPT beyond one year after PCI in the contemporary population.
Dr. Greg Hundley:
Very nice. And so clinically as we're managing patients, how would we interpret your results and help us with clinical management today?
Dr. Neel Butala:
Yeah, great question. These results really harmonize the DAPT trial with results of newer clinical trials of DAPT duration, which all demonstrate the safety of shorter duration DAPT to reduce bleeding risk and these results more broadly illustrate the importance, the nuance interpretation of clinical trials to guide clinical decision making and really highlight the risk of simply applying the top line trial results to all patients in contemporary practice, even beyond the study. These results emphasize the importance of accounting for patient specific factors and leveraging risk scores when available in deciding how clinical trials results actually apply to a particular patient. And finally, the results actually illustrate the importance of continually evaluating the generalizability of cardiovascular trials to ensure that the guidelines reflect treatment effects in contemporary clinical practice. And the methods that we use in the study can actually be used to do this in RCTs more broadly.
Dr. Greg Hundley:
Very nice, Neel. And thank you for really bringing us this study highlighting the differences between patients and devices used in contemporary clinical practice compared with those in the DAPT study. And how they were associated with attenuation of benefits and greater harms attributable to prolonged DAPT duration.
Dr. Greg Hundley:
Well listeners, what an exciting day. Getting to see these papers in print and also have these early stage investigators from the 2021 sessions that were finalists in many of these competitions, we're so grateful to Amgad Mentias, Matt Burrage, Shaan Khurshid, Maryam Zekavat and Neel Butala for their time today.
Dr. Greg Hundley:
On behalf of Carolyn and myself, we want to wish you a great week and we will catch you next week on the run.
Dr. Greg Hundley:
This program is copyright of the American Heart Association, 2022. The opinions expressed by speakers in this podcast are their own and not necessarily those of the editors or of the American Heart Association. For more, please visit ahajournals.org.