Objectively and Subjectively Measured Physical Activity and Their Associations With Cardiometabolic Risk in the UK Biobank: Retrospective Cohort Study

Abstract

Background: The association between physical activity (PA) behavior and cardiometabolic risk factors has depended largely on questionnaire-based reporting. More studies are turning to mobile health (mHealth) device solutions to measure PA. While there are differences between self-reported activity levels and objectively measured accelerometer-based activity, how these differences manifest in disease risk is unknown.

Objective: Here, we sought to evaluate these differences between self-reported and mHealth-based PA and to model the impact on their association with cardiometabolic factors. Our study provides a framework to assess the quality of relationships measured by mHealth technologies, which is generalizable to other sensors or activity-measuring devices.

Methods: We assessed PA using both wrist-worn accelerometer data and self-reported questionnaires in 16,000 participants of the UK Biobank (UKB) between 2013 and 2015, focusing on walking, sleeping, sedentary, and moderate-to-vigorous physical activity (MVPA). We compared the concordance between self-reported and objective measures of PA. We also compared the association between objectively measured or self-reported PA and future clinical biomarker levels (eg, BMI, pulse rate, glucose control, and cholesterol).

Results: Participants underestimated their weekly sedentary duration on average of 2.86 hours, and the coefficient of correlation (r) between subjective and objective activity was 0.12 for sedentary time, 0.16 for MVPA, 0.18 for walking, and 0.13 for sleeping. We found an inverse association between objectively measured MVPA and cardiometabolic biomarkers such as BMI and pulse rate, but found no association between subjectively reported activity and cardiometabolic biomarkers. We estimated that there is a 6% larger association between subjectively measured MVPA and BMI in healthy adults (vs the objective counterpart). We also estimated a 2%-3% difference on a healthy adult heartbeat (healthy range: 60-100 bpm) if relying on subjectively reported observations instead of measured PA.

Conclusions: These findings suggest that the association based on self-reported activity is likely overestimated and biased compared with objectively measured PA. Therefore, care should be taken when assessing the effects of self-reported PA on key cardiometabolic factors, such as BMI and pulse rate. We emphasize that while the associations are biased when comparing PA modalities, we cannot conclude which method more closely reflects the daily activity load.

Keywords: accelerometer; biobank; biomarkers; body mass index; cardiometabolic; cardiometabolic risk; cholesterol; clinical biomarker; exercise; glucose; observational study; physical activity; pulse rate; risk factor; sedentary.

Figure 1.. Flowchart of the study: in blue the main study population; in green the subpopulation that wore the accelerometer for an uninterrupted period of 7 days.

Figure 2.. (A) Three individuals’ average measured hourly activity over the course of a day overlaid with their scaled averaged recorded triaxial acceleration (black line) separated by weekday/weekend (black line discontinuities correspond to values in excess of the figure window), (B) the averaged (mean) classified measured hourly activity of the study cohort over a 24-hour period separated by weekday/weekend, and (C) the cohort’s weekly measured and self-reported physical activity, stratified by activity type.

Figure 3.. Scatter plots of the study population comparing the reported subjective minutes to the objectively measured minutes in each type of physical activity. The correlation coefficient (rho) is reported in each grid. The black line is the identity line (y=x), corresponding to a perfect one-to-one match. Hexagons are color-coded by their density: (A) weekly minutes of sedentary time, (B) weekly minutes of walking, (C) weekly minutes of sleeping, and (D) weekly minutes of moderate to vigorous activity.

Figure 4.. (A) Standardized effect sizes of objective physical activity adjusted for dietary intake, adapted CCI, and subjective physical activity on selected biomarkers. The subjective physical activity coefficients are plotted as a comparison; (B-E) effect sizes of PA on selected biomarkers issued from two separate models of objective and subjective physical activity, stratified by age group: (B) moderate-to-vigorous physical activity BMI, (C) moderate-to-vigorous physical activity pulse rate, (D) sedentary BMI, (E) and sedentary pulse rate.