Metabolomic markers of physical activity in the IDATA Study
Principal Investigator
Name
Eleanor Watts
Degrees
D.Phil., M.P.H.
Institution
National Cancer Institute
Position Title
Postdoctoral Fellow
Email
About this CDAS Project
Study
IDATA
(Learn more about this study)
Project ID
IDATA-61
Initial CDAS Request Approval
Jan 13, 2023
Title
Metabolomic markers of physical activity in the IDATA Study
Summary
Physical activity is positively associated with longevity and health. However, the mechanisms through which physical activity conveys health benefits are less established. The development of high throughput metabolomics offers a novel opportunity to investigate these mechanisms. Previous metabolomic studies have identified that higher physical activity levels are associated with amino acids, lipids and triglyceride levels1. However, associations are modest in comparison with obesity, possibly relating to physical activity measurement error and variability over time2 3. More precise physical activity assessment techniques are expensive and time consuming. Consequently, the studies that have utilized these methods have limited sample sizes and generally have not collected metabolomic data.
The IDATA study is the largest study available comprising gold-standard physical activity assessments and a wide range of metabolite data. This provides a unique opportunity to i) generate robust estimates for the associations of physical activity with metabolites and ii) develop metabolomic signatures of these behaviors, which may be useful to other studies without the capacity to collect comprehensive physical activity data.
We anticipate that the proposed analysis will lead to the discovery of novel biomarkers of physical activity and underpin future research efforts focused on understanding the biological mechanisms underpinning physical activity and disease relationships as well as the translation of these biomarkers as modifiable markers of lifestyle interventions and health.
Methods
Using data from participants who have complete:
• ActivPAL, ActiGraph: active and sedentary time (hr/day), steps (n/day)
• Doubly labelled water (DLW): physical activity energy expenditure and physical activity level
• Metabolite data: Metabolon untargeted platform
• Biospecimen collection data
• Urine markers
• Dietary and anthropometric data
This sample will include approximately 700 participants. The physical activity assessment data will be processed using methods described elsewhere2. Where repeat measurements of physical activity and biomarkers are available, we will use the mean value. To understand the temporal variability of these measures, we will also calculate intraclass correlation coefficients for the repeat measurements.
The associations of each of the physical activity measures with biomarkers will be evaluated using linear regression. We will also develop multi-metabolite scores using least absolute shrinkage and selection operator (LASSO) regression to predict physical activity. To cross validate these scores, the data will be partitioned into training and testing sets.
All models will be adjusted for age, racial/ethnic group, study group and body mass index. Smoking information was not directly collected, but current smoking intensity will be accounted for using cotinine and its metabolites4. We will also examine models with and without adjustment for dietary factors (e.g., caloric intake).
DLW and serum metabolomics data are also available from the Health ABC study. To investigate the external validity of the score developed in the IDATA Study, these scores will be cross validated in Health ABC where metabolites overlap.
References
1. Kelly,et al. doi: 10.1016/j.bbadis.2020.165936
2. Matthews,et al. doi: 10.1249/mss.0000000000001428
3. Neilson,et al. doi: 10.1093/ajcn/87.2.279
4. Benowitz NL,et al. doi: 10.1038/clpt.1994.169
The IDATA study is the largest study available comprising gold-standard physical activity assessments and a wide range of metabolite data. This provides a unique opportunity to i) generate robust estimates for the associations of physical activity with metabolites and ii) develop metabolomic signatures of these behaviors, which may be useful to other studies without the capacity to collect comprehensive physical activity data.
We anticipate that the proposed analysis will lead to the discovery of novel biomarkers of physical activity and underpin future research efforts focused on understanding the biological mechanisms underpinning physical activity and disease relationships as well as the translation of these biomarkers as modifiable markers of lifestyle interventions and health.
Methods
Using data from participants who have complete:
• ActivPAL, ActiGraph: active and sedentary time (hr/day), steps (n/day)
• Doubly labelled water (DLW): physical activity energy expenditure and physical activity level
• Metabolite data: Metabolon untargeted platform
• Biospecimen collection data
• Urine markers
• Dietary and anthropometric data
This sample will include approximately 700 participants. The physical activity assessment data will be processed using methods described elsewhere2. Where repeat measurements of physical activity and biomarkers are available, we will use the mean value. To understand the temporal variability of these measures, we will also calculate intraclass correlation coefficients for the repeat measurements.
The associations of each of the physical activity measures with biomarkers will be evaluated using linear regression. We will also develop multi-metabolite scores using least absolute shrinkage and selection operator (LASSO) regression to predict physical activity. To cross validate these scores, the data will be partitioned into training and testing sets.
All models will be adjusted for age, racial/ethnic group, study group and body mass index. Smoking information was not directly collected, but current smoking intensity will be accounted for using cotinine and its metabolites4. We will also examine models with and without adjustment for dietary factors (e.g., caloric intake).
DLW and serum metabolomics data are also available from the Health ABC study. To investigate the external validity of the score developed in the IDATA Study, these scores will be cross validated in Health ABC where metabolites overlap.
References
1. Kelly,et al. doi: 10.1016/j.bbadis.2020.165936
2. Matthews,et al. doi: 10.1249/mss.0000000000001428
3. Neilson,et al. doi: 10.1093/ajcn/87.2.279
4. Benowitz NL,et al. doi: 10.1038/clpt.1994.169
Aims
• Identify the biomarkers associated with physical activity behaviors assessed by doubly labelled water (DLW) (physical activity energy expenditure and physical activity level) and accelerometers (active and sedentary time (hr/day), steps (n/day)) using the IDATA study
• Develop multi-metabolite scores to predict these physical activity behaviors
• Externally validate our DLW metabolite scores using Health ABC data, where metabolites overlap
Collaborators
Eleanor Watts: Division of Cancer Epidemiology and Genetics, NCI, MD, USA
Joshua Freeman: Division of Cancer Epidemiology and Genetics, NCI, MD, USA
Grace Hong: Division of Cancer Epidemiology and Genetics, NCI, MD, USA
Erikka Loftfield: Division of Cancer Epidemiology and Genetics, NCI, MD, USA
Charles Matthews: Division of Cancer Epidemiology and Genetics, NCI, MD, USA
Steven Moore: Division of Cancer Epidemiology and Genetics, NCI, MD, USA
Pedro Saint-Maurice: Division of Cancer Epidemiology and Genetics, NCI, MD, USA