Dataset for the Criterion Validation of Four Activity Sensors against Doubly Labelled Water for Assessing Habitual 24-Hour Energy Expenditure
Documentation of the data | The dataset is provided in tabular format, with each row representing one participant and each column representing one variable. The participant identifier (ID_Number) is used to distinguish records and is not intended to provide direct information about participant identity. Continuous variables are reported using the units specified in the variable names. These include years for age, centimetres for body height, kilograms for body mass and fat-free mass, kg·m⁻² for body mass index, percentage for body fat, MET-min·week⁻¹ for physical activity, ml·kg⁻¹·min⁻¹ for estimated VO₂max, hours and minutes for valid sensor wear time, and kcal·day⁻¹ for total energy expenditure. Categorical variables include biological sex, physical activity level according to the IPAQ-SF, hand dominance, and Fitzpatrick skin-tone type. Hairiness scores for the wrist and upper arm are recorded on a 1–4 scale. The meaning of categorical and ordinal variables should be interpreted according to the variable names and the corresponding study documentation. TEEValid_DLW represents total energy expenditure determined using doubly labelled water and serves as the criterion measure. TEE variables associated with Polar Ignite 3, Polar Verity Sense, Garmin vívoactive 4, and GENEActiv represent sensor-derived estimates of habitual 24-hour energy expenditure. Where both LOG and SW variables are available, LOG refers to processing based on self-reported wear time, whereas SW refers to manufacturer software-derived wear time. Wear-time variables indicate the amount of valid data available for the respective device and processing approach. Differences between log-based and software-based wear-time and TEE variables reflect differences in data processing and validity criteria. Missing values indicate that a measurement or valid estimate was not available for the respective participant and variable. Users should not interpret missing values as zero. Before statistical analysis, users should review the relevant validity and wear-time variables and apply inclusion criteria appropriate to their research question. The dataset should be analysed at the participant level. Variables should be interpreted in conjunction with the study manuscript and accompanying metadata, particularly when comparing TEE estimates across devices or processing approaches. | |
Additional geographical or spatial references | Saxony | |
Countries to which the data refer | GERMANY | |
Description of the data | This dataset contains individual-level data collected as part of a study investigating the criterion validity of four activity sensors for assessing habitual 24-hour total energy expenditure (TEE), using doubly labelled water (DLW) as the criterion method. The dataset includes demographic and anthropometric characteristics (biological sex, age, body height, body mass, body mass index, body fat percentage, and fat-free mass), as well as characteristics potentially relevant to sensor-based measurements, including skin tone, hairiness at the wrist and upper arm, and hand dominance. Information on habitual physical activity is provided using both a MET-based measure and the International Physical Activity Questionnaire – Short Form (IPAQ-SF). Estimated maximal oxygen uptake (VO₂max) values obtained from selected activity sensors and their associated fitness tests are also included. For each measurement method, valid wear-time information is provided where applicable. The primary outcome is habitual 24-hour total energy expenditure measured using doubly labelled water (DLW). In addition, TEE estimates obtained from four activity sensors are included: Polar Ignite 3, Polar Verity Sense, Garmin vívoactive 4, and GENEActiv. For selected devices, estimates based on different data-processing approaches are provided, including log-file-based processing (LOG) and manufacturer software (SW). The dataset was created to support the assessment of agreement and criterion validity of activity-sensor-derived estimates of habitual 24-hour energy expenditure compared with DLW. Variable names indicate the measured construct, device or assessment method, processing approach where applicable, and unit of measurement. The dataset contains pseudonymised participant identifiers and does not contain directly identifying personal information. | |
Type of the data | Dataset | |
Total size of the dataset | 35048 | |
Author | Heydenreich, Juliane | |
Author | Lutz, Helena | |
Author | Manzke, Christopher | |
Author | Oeschger, Regina | |
Author | Gilgen-Ammann, Rahel | |
Upload date | 2026-09-24T06:42:08Z | |
Publication date | 2026-09-24T06:42:08Z | |
Data of data creation | 2026 | |
Publication date | 2026-09-24 | |
Abstract of the dataset | This dataset contains data from a study evaluating the criterion validity of four activity sensors for assessing habitual 24-hour energy expenditure (EE) against doubly labelled water (DLW), used as the reference method. The dataset includes measurements of habitual energy expenditure obtained using DLW and estimates derived from four activity sensors, together with relevant participant characteristics and study-related variables. The data were collected to enable the comparison of sensor-derived estimates with DLW-based measurements of habitual 24-hour energy expenditure and to support analyses of agreement, validity, and measurement error across the four devices. | |
Public reference to this page | https://opara.zih.tu-dresden.de/handle/123456789/2923 | |
Public reference to this page | https://doi.org/10.25532/OPARA-1565 | |
Publisher | Universität Leipzig | |
Licence | Attribution-NonCommercial 4.0 International | en |
URI of the licence text | http://creativecommons.org/licenses/by-nc/4.0/ | |
Specification of the discipline(s) | 2 | |
Title of the dataset | Dataset for the Criterion Validation of Four Activity Sensors against Doubly Labelled Water for Assessing Habitual 24-Hour Energy Expenditure | |
Research instruments | GENEActiv (Activinsights, Kimbolton, UK) | |
Research instruments | Polar Ignite 3 (Polar Electro Oy, Kempele, Finland) | |
Research instruments | Garmin vívoactive 4 (Garmin International, Inc., Kansas City, USA) | |
Research instruments | Polar Verity Sense (Polar Electro Oy, Kempele, Finland) | |
Software | Garmin Connect (Garmin Ltd., Olathe, KS, USA) | |
Software | Polar Flow (Polar Electro Oy, Kempele, Finland) | |
Software | IBM SPSS Statistics for Windows (29.0.1.0; IBM Corp., Armonk, NY, USA) | |
Project abstract | Background: Accurate assessment of total energy expenditure (TEE) is important in sport science, health research, and clinical practice. Body-worn activity sensors offer a practical approach for estimating TEE under free-living conditions; however, their validity against the doubly labelled water (DLW) gold standard remains insufficiently examined. This study validated four body-worn activity sensors, including consumer wearables and a research-grade accelerometer, against DLW for habitual TEE assessment. Methods: Fifty-six healthy adults (66% female; 29.4±7.2 years; BMI: 22.9±2.6 kg·m⁻²; estimated VO₂max: 49.5±9.8 ml·kg⁻¹·min⁻¹) wore four activity sensors simultaneously at predefined body locations (Polar Ignite 3 [IG], Polar Verity Sense [VS], Garmin vívoactive 4 [VA], GENEActiv [GA]) for 10 consecutive days under free-living conditions. Participants with ≥8 valid days, defined as ≥12 h·day⁻¹ of sensor wear time, were included. Valid periods were determined using wear time logs (LOG) or sensor software (SW). IG and VA provided TEE directly from manufacturers’ algorithms, whereas VS and GA provided wear-time energy expenditure, to which estimated non-wear energy expenditure was added to derive 24-h TEE. DLW-derived TEE (TEEDLW) was calculated as mean daily TEE across the measurement period and served as the criterion measure. Agreement and prediction accuracy were assessed using MAPE, Wilcoxon signed-rank tests, Bland–Altman analyses, and linear regression (α=.05). Results: TEEDLW was 2930±697 kcal·d⁻¹. In the LOG scenario, mean sensor-derived TEE was 3080±592 kcal·d⁻¹ for IG, 3039±634 kcal·d⁻¹ for VS, 2528±457 kcal·d⁻¹ for VA, and 2705±470 kcal·d⁻¹ for GA. VA and GA significantly underestimated TEEDLW, whereas IG and VS did not differ from DLW. In the SW scenario, IG and VS again did not differ from DLW, while GA significantly underestimated TEE. Across all sensors and inclusion scenarios, MAPE ranged from 13.6% to 16.6%, with 30.3–56.9% of participants achieving a MAPE ≤10%. Associations with TEEDLW were moderate to high (R²=0.39–0.65), but prediction errors remained substantial (SEE: 448–587 kcal·d⁻¹). Bland–Altman analyses showed wide limits of agreement across all sensors. Conclusions: Wearable activity sensors demonstrated moderate to high agreement with DLW-derived TEE at the group level. Nevertheless, substantial inter-individual variability and wide limits of agreement limit their applicability for individual-level TEE assessment. | |
Funding Acknowledgement | This project was funded by research grants from the Federal Institute of Sport Science (Bundesinstitut für Sportwissenschaft; BISp) based on a resolution of the German Bundestag (grant number: 2524BI0108). | |
Project title | Criterion Validity of Four Activity Sensors against Doubly Labelled Water for Assessing Habitual 24-Hour Energy Expenditure |
Collections
