Abstract
UK Biobank uses wrist accelerometers which can provide relatively accurate measures of physical activity and are useful for understanding links to mortality, cardiovascular disease and cancer. Only in 1 in 5 participants took part in this aspect of UK Biobank, and the measurements took place 5 years after the initial start of the study, meaning at present analyses are less powerful than they would be had the measures been taken earlier. In contrast, almost all of the 500,000 participants reported their physical activity in a questionnaire at the start of the study, although these data aren't as accurate as the accelerometer data. In this study we tried to improve the accuracy of the questionnaire data by calibration with the more robust accelerometer measures. We then tested the accuracy of our new physical activity estimate in a separate group of people and then used the estimate to investigate how physical activity is related to disease in the main UK Biobank cohort. We found that the group average for the new physical activity estimate compared very well with the group average from the best measure of physical activity available the doubly labelled water method. The correlation of the new physical activity estimate with this gold-standard measure was good considering it came from questionnaire data, but not as strong as device-based measures such as accelerometers. Our results mean that our new questionnaire-based estimate of physical activity is good for putting individuals in a group in rank-order and for telling us how active that group is on average. When we used the new activity estimate to see how it related to disease, we found higher physical activity was associated with lower cardiovascular disease and cancer, and that meeting the current physical activity guidelines resulted in approximately 14% lower risk of mortality.
1 Application
Application ID | Title |
20684 | Understanding the associations between environmental exposures, health behaviours and health outcomes |
1 Return
Return ID | App ID | Description | Archive Date |
3412 | 20684 | Estimating physical activity from self-reported behaviours in large-scale population studies using network harmonisation: findings from UK Biobank and associations with disease outcomes | 14 May 2021 |