The dose-response associations between MVPA and cardiovascular disease varied markedly across three accelerometer-derived MVPA metrics, suggesting that research using a single accelerometer metric may caution about the interpretation of the association.
The study included 90,237 cardiovascular disease-free participants at baseline, with 1,298 incident strokes and 2,031 myocardial infarctions observed during follow-up.
For stroke, a linear decrease in hazard ratio was observed with the machine-learning metric, but not with LFENMO and activity counts metrics.
For myocardial infarction, machine-learning and LFENMO showed a curvilinear decrease in hazard ratios, whereas activity counts showed a linear decrease.
The dose-response associations between MVPA and cardiovascular disease varied markedly across the three accelerometer-derived MVPA metrics.
The study used restricted cubic splines within Cox regression models to characterize the shape of dose-response relationships for each metric.
This research examined whether the way physical activity is measured by wrist-worn devices (accelerometers) affects what we conclude about the relationship between exercise and heart disease risk. Using data from over 90,000 UK Biobank participants, the researchers compared three different mathematical methods for converting raw accelerometer movement data into estimates of moderate-to-vigorous physical activity (MVPA): a signal-filtering approach (LFENMO), a machine-learning algorithm, and a traditional activity counts method. They then tracked who developed strokes or heart attacks over roughly a decade. The study found that the three methods told noticeably different stories. For stroke, only the machine-learning method showed a straightforward linear reduction in risk as physical activity increased, while the other two methods showed different patterns. For heart attacks, the machine-learning and signal-filtering methods suggested the protective benefits of activity level off at higher amounts (a curved relationship), while the activity counts method suggested risk keeps dropping steadily with more activity. These differences in curve shape were meaningful and not trivial. This research suggests that conclusions drawn from studies on physical activity and heart disease may depend heavily on which technical method researchers choose to process accelerometer data, even when the underlying device data are identical. This has important implications for how public health guidelines are developed and how individual studies should be interpreted, since a finding from one study using one processing method may not be directly comparable to findings from another study using a different method.
Upload your bloodwork. We'll cross-reference your results against this study and 4,700 others.
Upload Your LabsLapointe Y, Kapur A, Sharma A, Panter J, Fuller D, Mamiya H. (2026). Investigating the comparability of wearable accelerometer methods in the association between physical activity and cardiovascular disease: a cohort study using UK Biobank.. Preventive medicine. https://doi.org/10.1016/j.ypmed.2026.108603