S1 File -

Introduction
The ActiGraph (AG) accelerometer is widely used to assess physical activity (PA) in heart failure (HF) patients. However, the validity of the AG in this population remains unexplored.
Objective
Therefore, this study examined the criterion validity of the AG-GT9X for measuring step counts (SC) and energy expenditure (EE) among HF patients.
Methods
16 patients with HF with preserved ejection fraction (mean age = 60.3±12.1yrs) completed a total of 41 symptom-limited cardiopulmonary exercise tests on a treadmill across multiple time points (median (IQR) = 2.5 (1.5–3.5)). All participants wore the AG (model: GT9X) on both the right ankle and waist locations during the test. Manually counted steps and indirect calorimetry-derived EE served as criterion measures. AG-derived EE was estimated using six different prediction equations previously developed for waist-worn AG. AG-derived measurements were compared with criterion measurements by calculating correlation coefficients, equivalence tests with two one-sided tests, mean absolute percentage error (MAPE), percentage bias, and Bland-Altman plots using mixed models to account for the nested nature of repeated measures within subjects.
Results
Ankle-worn AG-SC was significantly equivalent to the criterion (p < .05) and had lower MAPE (<10%) compared to the waist location, regardless of PA intensity level. Sasaki-EE was significantly equivalent to the criterion (p < .05), with the lowest percentage bias overall (0.7%).
Conclusions
The ankle-worn AG-SC and Sasaki-EE showed better accuracy among HF patients in laboratory settings. Further research is warranted to cross-validate the results in different settings.