The Role of Timing in Physical Activity & CVD Mortality Risk

PhD student Sunwoo Emma Cho and Dr. Rahul Ghosal unpack their recent article, “Exploring the association between daily distributional patterns of physical activity and cardiovascular mortality risk among older adults in NHANES 2003-2006” published in the November 2024 (Vol. 99) issue of Annals of Epidemiology. In this study, the researchers quantify the association between daily (temporally varying) distributional patterns of physical activity and cardiovascular mortality risk. 

Read the full article here:
https://www.sciencedirect.com/science/article/abs/pii/S1047279724002436

Episode Credits:

• Executive Producer: Sabrina Debas
• Technical Producer: Paula Burrows
• Annals of Epidemiology is published by Elsevier.
  
  
  

Chapters

• 12:17 Behind the Paper

Transcript

Patrick Sullivan: 0:12

Hello, you're listening to EPITalk: Behind the Paper, a monthly podcast from the Annals of Epidemiology. I'm Patrick Sullivan, Editor-in-Chief of the journal, and in this series we take you behind the scenes of some of the latest epidemiologic research featured in our journal. Today we're talking about the article "Exploring the Association Between Daily Distributional Patterns of Physical Activity and Cardiovascular Mortality Risk Among Older Ad adults in NHANES 2003 to 2006." And we're talking with two of the authors, Dr. Rahul Ghosal and Ms. Emma Cho, about this article. You can find the full article online in the November 2024 issue of the journal at www. annalsofepidemiology. org. I'm going to give you a brief introduction to our guests.

Patrick Sullivan: 1:08

Dr. Rahul Ghosal is an assistant professor of biostatistics in the Department of Epidemiology and Biostatistics University of South Carolina. His research focuses on developing novel statistical methods such as functional data analysis with applications in biosciences and also distributional data analysis for modeling, wearable data with applications in gait aging and Alzheimer's disease. Sunwoo Emma Cho is currently a PhD student at the Darla Moore School of Business University of South Carolina. Previously, she was enrolled in the biostatistics PhD program at the same university. She holds a bachelor's degree in econometrics and operations research from Erasmus University, Rotterdam and a master's degree in information systems from Seoul National University. Rahul and Emma welcome.

Rahul Ghosal: 2:01

Hi Patrick, good morning. It's a pleasure to be here and we're looking forward to this podcast and discussing the papers and the details behind it.

Emma Cho: 2:11

Yeah, thank you for inviting us, yeah.

Patrick Sullivan: 2:14

So, Rahul, we'll start with you. What was the purpose of this study? What research questions were you really trying to get at?

Rahul Ghosal: 2:20

So we know that cardiovascular disease is one of the leading causes of death, both worldwide and also in the US. So previously a lack of physical activity has been shown to be a prominent risk factor for CVD mortality. So we all now monitor our steps in smartphones and get those steps in. But now these traditional measurements of physical activity previously was self-reported and based on various summary level metrics, which had its own drawbacks. But recent advances in wearable device studies provide continuously monitored and objectively measured physical activity data and, making use of more advanced statistical methods, we can have a better understanding of the implications of physical activity in the context of CVD mortality. We can do it by considering both the timing of the physical activity and its daily compositions.

Rahul Ghosal: 3:14

So our research question basically was whether these daily distributional patterns or compositional patterns of physical activity, are associated with cardiovascular mortality risk among the older adults and independent of aging. So that's the research question we are trying to answer here.

Patrick Sullivan: 3:31

Great, so walk us through your study design, then, and the methods that you used to conduct the research that's reported in the article.

Rahul Ghosal: 3:38

Sure.

Rahul Ghosal: 3:39

So for our study, what we used was the accelerated data from the 2003-2006 National Health and Nutrition Examination Survey, also known as NHANES, and this study actually provides minute-by-minute level accelerometer data.

Rahul Ghosal: 3:54

And we focused on the older adults aged more than 50 years and their physical activity data was linked with mortality data coming from National Death Index, and we censored the subjects in December 2019 for collecting their mortality information.

Rahul Ghosal: 4:13

And then we used this recently developed statistical methods in functional and distributional data analysis, which is very close to my research area, and we obtained, based on this minute-by-minute physical activity data, some temporarily varying this minute by minute physical activity data, some temporally varying distributional patterns of physical activity, and we call these as L moments, and this can be essentially thought as kind of robust analogs of traditional moments like mean or standard deviation. But now, because we have multiple days of data, we can calculate them in a way such that they are varying over the time of the day, so we can collect various information, such as the daily average, the variability, the skewness and their kurtosis pattern throughout the various times of the day. So now they kind of provide a lot more information beyond the daily average of physical activity, and these L moments were then used as our main exposure variables in a Cox model to estimate the hazard of cardiovascular mortality, while adjusting for other known confounders such as gender, age, smoking status and coronary heart disease. So that was basically the method we used to answer our research question.

Patrick Sullivan: 5:24

Great, and I'm going to turn to you for the next question, and I wonder if you could tell us about what the main findings from the study was, and how did they align with the hypotheses that you had going into doing the research?

Emma Cho: 5:36

So in most of like association study, when we explore relationship between physical activity and CVD mortality, those will usually go with mean. What we found from our study is that it's not just the average intensity of physical activity during the day that matters, but also the variability of physical activity, especially in the afternoon, could also play a very crucial role to varying the CVD mortality risk Also, interestingly, we also saw that the association between daily mean, which is L1 moment, and daily variability L2 moment of physical activity with CVD mortality were much more pronounced on the weekdays and it was not really significant on weekend. So it was like, aligned with our research hypothesis, yeah, which was surprising in a good way, yeah. So we think our study demonstrated both circadian rhythm of physical activity plus daily decomposition can be useful to design time of day and intensity specific physical activity interventions to protect against the CVD mortality risk.

Patrick Sullivan: 6:47

Thank you. So have others studied this? It seems like a very interesting and a very specific hypothesis, and a testable one, which is great. But have others studied this same kind of question, and how do your findings compare with previous literature on this subject?

Rahul Ghosal: 7:02

Yeah, that's a very good question. So our findings actually match with the previous literature in terms of finding a protective effect of physical activity against the risk of cardiovascular mortality.

Rahul Ghosal: 7:13

But, as Emma mentioned, most of these previous studies were based on various summary-level measures of physical activity, such as total activity count or moderate to vigorous physical activity, which is definitely useful. However, our results now kind of complements those results and also provide a deeper understanding into the relationship between physical activity and cardiovascular mortality risk. By zooming into this physical activity pattern, because we have continuous physical activity minute by minute level we can, instead of using a single summary level metric, zoom it into much more and obtain these patterns of physical activity varying over the time of day, and so, by considering its timing as well as composition, we one of the key takeaways is, even if your average physical activity is same throughout the day, a person with lower variability of physical activity would have a lower reserve of physical activity and therefore would be an increased risk for cardiovascular mortality.

Rahul Ghosal: 8:18

So I guess the takeaway message, like our research would be like it's not only the average physical activity that matters. It's also good to have different ranges of physical activity and particularly different times of the day.

Patrick Sullivan: 8:30

Right. It seems like, just as a matter of you know, thinking about some of the epi aspects of this, that one, just having the objective device reduces some information bias, right, because especially when we ask people about physical activity, there is a social desirability component. But then you have this kind of huge quantity of data that really reflects much more. And I wonder if you feel like, could you have gotten to the issue of the variability without that resolution of data? If you're just asking people like how many minutes do you exercise a day, would you have even had the right data to ask the question about the variability?

Rahul Ghosal: 9:09

Right. That's a very interesting question. I think what helps in capturing this variability, and especially at various time points of the day, because we have such multiple days of continuously measured physical activity. I think- so therefore our variability estimates are much more less uncertain. We know in statistics it's not only the point estimates that matter, also the uncertainty of the estimate. And because we have such hugely collected, intensively longitudinal data, we can capture patterns like variability or skewness with much less uncertainty. So I think that's also one of the definite benefit of such objectively measured data. And another aspect is there could be subjective bias introduced. People are trying to describe the variability in physical activity, so I think definitely that is also taken care of in such objective assessments.

Patrick Sullivan: 9:58

Yeah. Great. Can you talk a little bit about some of the strengths and any important limitations to your study?

Emma Cho: 10:05

Yeah, sure, I think already Dr. Ghosal mentioned a lot of trends and key insights from our study. Yeah, as he mentioned earlier, studies on the association between physical activity and CVD mortality mainly relied on summary scalar metrics such as total activity count, or usually we use survey data and we are asked by threshold-based measure for specific activity like light intensity, LIPA, moderate to vigorous physical activity, and VPA. But in our study we took it a step further by applying a novel method, a partially functional distribution approach, to dive deeper into the structural PA in a more detailed, granular way, a temporally varying distribution patterns beyond the mean. So, yeah, key insight I think from our research is the importance of variability in physical activity, particularly in the afternoon. Yeah, we mentioned which can be leveraged to design more effective PA interventions that aim at reducing CVD mortality, which is a very significant problem and public health concern in the US and worldwide.

Patrick Sullivan: 11:16

Thanks for that information. I wonder if you could also talk a little bit about the limitations of your study.

Emma Cho: 11:21

Yeah, so, likewise any study, our research also has a few limitations. First of all, the associations we found between daily physical activity patterns and the CBD mortality risk don't imply the causal relationship. Yeah, still, there could be a possible residual confounding bias. Second of all, since we use a cause-specific Cox model, participants who passed away from causes other than CBD were treated as right-sensor, but some of those individuals might have underlying cardiovascular conditions that we didn't account for, which could introduce some kind of bias in our estimates. And lastly, yeah, there could be a further methodological work that can more directly incorporate survey-weighted causal survival model to explore as a further study.

Patrick Sullivan: 12:19

Great. So now we're going to move on to a section of the podcast that we call Behind the Paper, and we really think it's important to recognize that our methods are analytical and our process is scientific, but this work is ultimately done by people, and especially the relationships between co-authors and how we work together. So I just want to ask each of you a couple questions to understand more about how you work on this project together and how that sort of fit into your mentoring relationship. So, Emma, I'll start with you and just ask what's your favorite experience been in your PhD program so far?

Emma Cho: 12:55

Uh, so far- actually, my initial goal to join the biostat program was like developing, apply the computational method that effectively bridge medicine or probably carries over statistical or data science methodologies. Yes, through this program I've been able to explore new methodologies, especially for this study. Yeah, we can see more deep of the new data structure, new data frame with new methodological studies. So, yeah, it was most fascinating part. I was in the PhD program and then I worked as a hospital researcher for a while, so I saw all the moments that come together and then, yeah, I realized I can develop further study based on the program I learned.

Patrick Sullivan: 13:50

Yeah, those moments when all the pieces come together are like the days that we work for for sure. Is there anything that you wish you knew before you started the PhD program?

Emma Cho: 14:00

Yeah, actually I was interested in a lot of different sources and data structure. Also, like because usually if we when we like dealing with the table data, it does not give all information. So I was interested in other objective measures such as like variables or so image and text and yeah, those all new methodology curve yeah, that was I think it was. I want to dig into, know more about. Like to improve better medicine research.

Patrick Sullivan: 14:34

Great. Okay, and Rahul, we're going to turn back to you now for another question, so I'm going to call you Dr. Ghosal for this one, because I'm asking you like, in your role working with Emma as her advisor, what was that experience like and what advice do you have for first time mentors who may be working with students about how to enter into that mentoring relationship?

Rahul Ghosal: 14:56

So actually Emma was kind of my first student when she started working with me and I frankly had a great time working with Emma.

Rahul Ghosal: 15:04

She's like an excellent listener and also has a natural kind of scientific curiosity as a researcher, apart from her excellent skills in coding.

Rahul Ghosal: 15:15

So I really enjoyed working with her, like our meetings, and the project eventually turned out to be a lot of fun and actually we are also continuing our collaboration and exploring some interesting new directions based on the findings we've seen in this study. So definitely I had a very good experience working with her and regarding my advice to first-time mentors, it would be to have a good balance between being hands-on and also giving the students their own space for their own independent research and exploration. And we can all probably kind of use our personal experience because it was not so long ago that we were students ourselves, so we can identify that, what aspects we needed most from our mentors, and we can apply those in our day-to-day dealings or the meetings with the students. Also, it helps to understand the motivation, the goals of the student and connect with them non-academically too, and when you are a mentor and I have seen that worked a great bit in my experience working with the students.

Patrick Sullivan: 16:21

Great. So thanks for sharing that from both your perspectives about just how it was to work together. I'm also really always interested in how the work is done on these papers, so I wondered if you could say a little bit about what roles each of you played in the like the conception, the analysis, the writing, the submitting. How did you divvy up that work?

Rahul Ghosal: 16:43

So actually we had a great team behind this one. Emma was obviously the driving force behind the study and she contributed in the research in terms of formal investigation, doing all the analysis, validation, visualization and software and also writing. But also behind this we had some amazing colleagues of mine working. I should mention the name of D r. Marcos Matabuena. So he is currently a postdoc in Harvard and expert in this functional data analysis area and he has a lot of experience working with this enhanced data. So he was kind of our go-to guy, particularly in terms of the data aspects of it.

Rahul Ghosal: 17:22

Dr. Jingkai Wei, who is an assistant professor currently at the University of Texas Health Science Center at Houston. So he's an expert in cardiovascular disease epidemiology, cognitive aging and dementia. So he was kind of the domain expert and the epidemiology expert in this study and he kind of definitely helped us in the conceptualization and the writing aspect of the study. And finally, Dr. Enakshi Saha is also currently assistant professor at the University of South Carolina. She is an expert in Bayesian statistical methods and functional data analysis and machine learning. So she also helped in the conceptualization, the writing part and the supervision of the project. And whenever I needed my co-authors they were there for me and, like, helped as the best to their capacity. So I think we had, as I mentioned, a great team on this one, and I think that is how good science should be done working with a great team. So definitely I'm grateful to all of them.

Patrick Sullivan: 18:23

It's definitely the best science and I think it's more fun and just sort of professionally fulfilling, like in two ways. One, to interact with colleagues, but also that I think the best research rarely sits in one person's domain of expertise, and so when you talk about what each person brings and sort of crossing over, sometimes you get into an analysis and then you realize like, oh, there's a type of data here that I am not quite sure how to handle, and so then finding that colleague whose whole thing is handling that kind of data.

Patrick Sullivan: 18:48

I do think that, as much as we make an analysis plan, those things have to evolve, and so sometimes the authorship list may also either grow or you know.

Rahul Ghosal: 18:58

Definitely, definitely. Yeah, I agree. And sometimes, yeah, it definitely helps to have like experts at their respective fields working with you. And definitely yeah, while working on this project, I also began to appreciate like being a part of a team and working on to solve a common research question. I think that's really powerful. So I really enjoy when working this type of projects now where I can bring some of my methodological work to solve like interesting scientific answer, scientific questions, and definitely looking forward to doing much more of it in future.

Patrick Sullivan: 19:30

Great. So Dr. Ghosal, Ms. Cho, do either of you have have anything else you'd like to share with our listeners?

Emma Cho: 19:38

Yeah, I think I forgot like one of important, my enjoyable moment from my PhD. Actually, it was Dr. Ghosal, yeah, yeah, because yeah, he was always encouraging like to give more curiosity and yeah, how I find to derive this research experience and research moment and if they helping for interpretation and also like try to think together. And yeah, it was like, yeah, most of my good memory of, yeah, biostat PhD was from him actually, yeah.

Rahul Ghosal: 20:23

hanks Emma for saying that, and I would like to also like thank here the Annals of Epidemiology, the editors, the reviewers, the team and also the EPITalk team that we are here with today for doing this kind of outreach program, and I think it's really important for science to reach to people also and to bring out the takeaway message of our research that we are doing to a common person. I think such programs definitely help. So I thank Patrick and Sabrina for organizing this and all the team behind the EPITalk.

Patrick Sullivan: 20:54

Well, thank you. Thank you both, and that brings us to the end of this episode. Thank you again to Dr. Ghosal and to Ms. Cho for joining us today. It was such a pleasure to have you on the podcast. Likewise, thank you, thank you.

Patrick Sullivan: 21:08

I'm your host, Patrick Sullivan. Thanks for tuning in to this episode and see you next time on EPITalk, brought to you by Annals of Epidemiology, the official journal of the American College of Epidemiology. For a transcript of this podcast or to read the article featured on this episode and more from the journal, you can visit us online at www. annalsofepidemiology. org.\\