- Email: [email protected]
Validation and reliability of two activity monitors for energy expenditure assessment
G Model JSAMS-1104; No. of Pages 5
ARTICLE IN PRESS Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Journal of Science and Medicine in Sport journal homepage: www.elsevier.com/locate/jsams
Original research
Validation and reliability of two activity monitor for energy expenditure assessment Anne-Sophie Brazeau a , Nadia Beaudoin a,b , Virginie Bélisle a , Virginie Messier a , Antony D. Karelis c , Rémi Rabasa-Lhoret a,b,d,e,∗ a
Montreal Clinical Research Institute (IRCM), Montreal, Quebec, Canada Department of Medicine, Université de Montréal, Montreal, Quebec, Canada c Department of Kinanthropology, University of Quebec in Montreal, Montreal, Quebec, Canada d Department of Nutrition, Université de Montréal, Montreal, Quebec, Canada e Montreal Diabetes Research Center (MDRC) of the Université de Montréal Hospital Research Center (CR-CHUM), Montreal, Quebec, Canada b
a r t i c l e
i n f o
Article history: Received 11 July 2014 Received in revised form 10 October 2014 Accepted 1 November 2014 Available online xxx Keywords: Test–retest Activity monitoring Accelerometer Motion sensors Doubly labeled water
a b s t r a c t Objectives: This study explores the reliability and validity of the SenseWear Armband (SWA) and Actical (ACT) for free-living total energy expenditure, and energy expenditure during rest and light-to-moderate exercises (walking, ergocycling). Design: Participants wore the 2 devices during 7 days (free-living) and then participated to 3 days of testing in our laboratory. Methods: SWA and ACT estimates of total energy expenditure was compared to the doubly labeled water technique (7 days), and energy expenditure during rest (60 min), treadmill (45 min; intensities ∼22% to ∼41% VO2peak ) and ergocycling (45 min; ∼50% VO2peak ) were compared to indirect calorimetry over the following 3 days. Paired T-tests and intra-class correlation coefficient (ICC) with 95% confidence interval (CI95 ) were computed. Results: Twenty adults were recruited (BMI 23.1 ± 2.3 kg/m2 ). Compared to doubly labelled water, SWA overestimated energy expenditure by 94 kcal/d (±319; P = 0.2) and ACT underestimated by −244 kcal/d (±258; P = 0.001). Energy expenditure during rest (SWA 210 ± 116, ACT 124 ± 133 kcal/d; p < 0.05) and treadmill (according on intensity: SWA 54 ± 46 to 67 ± 38, ACT 68 ± 25 to 84 ± 40 kcal; p < 0.05) were overestimated and underestimated during ergocycling (SWA −93 ± 65, ACT −269 ± 111 kcal; p < 0.05) compared to indirect calorimetry. High ICC were observed at rest (SWA 0.994 CI95 0.987–0.997; ACT 0.998 CI95 0.996–0.999) and during ergocycling (SWA 0.941 CI95 0.873–0.975; ACT 0.854 CI95 0.687–0.939). Conclusion: Acceptable estimation of total energy expenditure was observed with the SWA. Both devices were reliable but not accurate for energy expenditure’s estimations during rest and for specific exercises. © 2014 Sports Medicine Australia. Published by Elsevier Ltd. All rights reserved.
1. Introduction It is important in clinical, nutritional and exercise sciences as well as in epidemiological research to develop simple and accurate devices for the evaluation of energy expenditure (EE). There are many methods available to estimate EE depending on financial and human resources as well as on research objectives. On one side, there are self-report measures such as activity diaries and questionnaires which are relatively inexpensive but also less
∗ Corresponding author. E-mail address: [email protected] (R. Rabasa-Lhoret).
accurate mainly due to recall bias.1 On the opposite, objective methods, such as doubly labeled water and calorimetry chambers, are more accurate but also more expensive. In between, there are portable activity monitors, such as accelerometers and motion sensors, which objectively estimate total energy expenditure (TEE)2 and provide detailed physical activity (PA) information (duration and intensity).3,4 They offer a unique opportunity to capture freeliving EE. Many devices are commercially available, including the Actigraph, Actical (ACT) and SenseWear Armband (SWA). The ACT, an omnidirectional accelerometer, is used, for example, for population surveillance.5 This device produces «count» value as its output. To allow comparison, counts are processed using predictions
http://dx.doi.org/10.1016/j.jsams.2014.11.001 1440-2440/© 2014 Sports Medicine Australia. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
G Model JSAMS-1104; No. of Pages 5
ARTICLE IN PRESS A.-S. Brazeau et al. / Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
2
equations into METS (i.e. metabolic equivalent) or calories. Validation studies among healthy adults, using published equations, have provided inconsistent results.6,7 The ACT provided valid resting energy expenditure (REE) estimation compared to indirect calorimetry for normal to overweight adults,8 but discrepancies were observed for EE during short PA periods.7,9 In a mechanical set-up, the ACT presented a good intra-instrument reliability (coefficient of variation (CV) = 0.5%).10 Despite its wide use, the ability of ACT to evaluate with accuracy the daily TEE has not been established. To our knowledge, this device and its software has not been validated against doubly labeled water (DLW), the gold standard to evaluate TEE.11 The SWA is another device commercially available. This device includes a 2-axis accelerometer but also considers physiological measures (e.g. skin temperature) to estimate EE. It showed good reliabilities over a large range of activities, from rest2,4 to cycling12,13 but underestimated EE during low speed walking on treadmill14 and during high intensity exercises.15,16 Its validity to estimate TEE has been assessed using the DLW technique; improved precision was found with newer versions of the sensor (Pro3) and its software (Innerview 6.1)2,3,17 compared to previous (Pro 2; Innerview 4.1),4 but SWA still underestimates TEE compared to DLW (−1122 and −1174 kcal/day). Comparisons between the two software versions have been previously published.3,17 Technologies improve rapidly and additional independent validation studies are needed to confirm the accuracy of these newer generation devices and software. Moreover, to compare EE estimated by these devices (i.e., comparing one only measuring accelerations to one measuring accelerations but also capturing physiological measures), head to head validation is required. Freeliving studies allow evaluating accuracy in real-life condition where intensities and type of activities varies, which is more representative of an actual lifestyle. Laboratory studies allow for controlling additional parameters, thus extending our understanding of the various limits of the devices (e.g., such as bicycling). We aimed at investigating the accuracy of the ACT and the SWA (i) for free-living TEE (main objective), (ii) for EE during different intensities of PA, (iii) for REE. A sub-objective was to examine the reproducibility of these two devices for EE during ergocycling (iv) and rest (v). We hypothesized that both devices would be accurate for TEE and REE.
2. Methods The project was designed to evaluate the SWA and the ACT in out-clinic (i.e., real life) and in-clinic (i.e., in the laboratory) settings (see supplementary figure). Participants wore the 2 devices for a 7-day period of free-living (out-clinic) and then participated to 3 days of testing in our laboratory (in-clinic). During the outclinic period, we evaluated TEE against DLW. During the last 3 visits, we evaluated the accuracy of these devices to estimate REE and EE at different intensities on the treadmill and during ergocycling against indirect calorimetry. We also examined the SWA and ACT’s reliability for EE during rest and ergocycling. The Institut de recherches cliniques de Montreal ethic review board approved the study. Written informed consent was obtained prior to data collection. Twenty healthy adults (85% Caucasian) participated in the study. They were recruited through word of mouth. Inclusion criteria were: aged between 18 and 45 years old, normal or overweight, non-smoker, without major disease, illness or medications use that would impact metabolism. During the first visit, the participants underwent a medical exam. Body weight and height were measured (light clothing, shoes removed). Then participants performed a graded exercise test on an ergocycle Ergoline 900 (Bitz, Germany) until voluntary exhaustion to measure their VO2 peak . During the
test, power output was increased by 25 W every 3 min. The highest oxygen consumption obtained was considered the VO2 peak (mL/kg/min). Breath-by-breath gas samples were analyzed using an Ergocard (software version 6, MediSoft, Dinant, Belgium) cardiopulmonary exercise test station. Participants were then trained on how to wear the SWA and the ACT according to the manufacturers’ instructions. Participants were asked to wear them for the complete study period (i.e., 10 days), to only remove them when showering and to restrain from water activities. The SWA Pro 3 (Bodymedia, Pittsburgh, PA) was worn on the upper right arm (on the triceps at the mid-humerus point). The SWA captures data through a 2-axis accelerometer and heat flux, galvanic skin response, skin temperature, and near-body ambient temperature sensors. The data as well as gender, age, body weight, height, handedness and smoking status is used to calculate EE using the Innerview Research Software version 6.1 developed by the manufacturer. The Actical device (Mini Mitter, Bend, OR) is a small electronic omni-directional monitor capturing movements in multiple planes, although it is most sensitive to vertical accelerations when worn on the hip.18 This sensor uses a piezoelectric technology to register accelerations called activity counts. Participants wore this device on the hip, positioned laterally above the iliac crest and secured with an elastic belt. The device was initialized using 60 s epochs. Data was analyzed using the manufacturer software (version 2.1) and counts were transformed to EE measures using a 2-regression equation.19 During this first visit, urine samples were collected and isotope solutions for DLW measurement were administered to assess TEE over a 7-d period. Each participant ingested 0.3 g/kg body weight of 2 H18 O. Complete procedures are described elsewhere.4 Briefly, urine sample obtained on visit 1 were used to assess the amount of tracers ingested. The four other samples (2 on visit 2 and 2 on visit 3) were used to evaluate the rate of disappearance of 2 H and 18 O. All samples were measured in triplicate. An Isoprime Stable Isotope Ratio Mass Spectrometer connected to a Multiflow-Bio module for Isoprime and a Bilson 222XL Autosampler (GV Instruments, Manchester, United Kingdom) were used for the measurement of TEE. Data were processed with the MassLynx 3.6 software (Waters Corp, Milford, MA). Stability tests were performed each day, which yield a standard deviation of 0.026% for 2 H and 0.004% for 18 O. The next day (visit 2), participants came to provide urine samples. A week after the first visit, following the 7 days out-clinic period, participants participated to the 3-day (10-h/day) in-clinic reliability trial. Each day, they followed, under supervision and with a strict time schedule, the same program of activities. They arrived, fasting, at the laboratory at 07h30 and participated to 60 min of lying awake and two periods of 45-min standardized physical activities (treadmill: intensities from ∼22% to ∼41% VO2peak ; ergocycle: ∼50% VO2peak ). The 3 days were identical except for the walking speed on the treadmill which increased between each visit to induce higher EE allowing validation over a wider range of speeds. Over the residual 7.5 h, which included the standardized breakfast, lunch and two snacks, participants remained seated. On the first day, they completed a chart of their sedentary activities (e.g. reading, watching TV) in order to reproduce them on the subsequent days. REE was measured during a fasting state by indirect calorimetry during one of the 3 mornings of the in-clinic trial. Carbon dioxide and oxygen were measured with a SensorMedics Delta Track II with the ventilated-hood technique. Measurement of gas concentrations were used to determine the 24-h REE with the Weir equation.20 Participants were lying in a supine position, without speaking or sleeping for 60 min. The first 10-min was an acclimatization period and was discarded. The last 50 min were used in the analysis. The calorimeter gas analyzers were calibrated prior each measurement.
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
G Model
ARTICLE IN PRESS
JSAMS-1104; No. of Pages 5
A.-S. Brazeau et al. / Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
During the standardized physical activities, EE was evaluated, at 1-min intervals, by indirect calorimetry with the same ergocard used to assess VO2peak . Data was analyzed with the Statistical Package for Social Sciences 20.0 (SPSS, Inc., Chicago, IL) and the MedCalc Software version 9 (Mariakerke, Belgium). All study variables were examined for normality of distribution prior to analysis (Kolmogorov–Smirnov). Data are presented as means ± standard deviation. The extent of agreement between measures of EE was evaluated with Pearson correlations, ICC and Bland and Altman plots. Mean differences between methods were also compared with paired t-tests. To evaluate the reliability ICC and CV were computed with 95% confidence interval (CI95 ). 3. Results Twenty healthy adults (85% Caucasian) participated in the study (age: 26.2 ± 3.6 years; body mass index was 23.1 ± 2.3 kg/m2 (range 20.0–30.3 kg/m2 ; 15% >25 kg/m2 ) and VO2peak 41.0 ± 11.6 mL/kg/min (range 22.6–65.7). The SWA was worn 97.7% of the time during the 7-day out-clinic period and 100% of the time during the three 10-h periods (in-clinic). The ACT was also worn 100% of the time during the in-clinic period (direct observation). However, given that the ACT does not capture information on wear time this information is not available for the 7-day out-clinic period. As per instructions, it should be similar to the SWA. No discomfort or negative effects were reported for the two devices. Table 1 showed comparisons of EE estimations. Mean TEE over 7 days estimated by both the SWA and ACT were significantly correlated with TEE measured with DLW. Bland–Altman analysis (Fig. 1) shows that there was no tendency of over or underestimation with SWA across the wide range of TEE. Meanwhile ACT showed a
3
tendency to increased underestimation with higher TEE. Accordingly, ACT significantly underestimated TEE by a mean of 244 kcal compared to DLW. Although a significant relationship between REE estimated by both devices with REE measured by indirect calorimetry was shown, SWA and ACT overestimated REE by 210 kcal (12.4%) and 124 kcal (7.4%), respectively. Most of the EE estimated by the 2 devices during the standardized exercises were significantly correlated with the indirect calorimetry measures (r = 0.80–0.90; p < 0.05) except for SWA during the lowest speed walk (mean 3.1 km/h) and ACT during ergocycling. For every intensity that was tested both SWA and ACT overestimated EE during the treadmill exercise (see supplementary online figures). Fig. 2 shows the repartition of EE during the 10-h periods spent in the research facility and illustrates that REE and EE during ergocycling and sedentary activities were reproducible over the three visits. Both devices showed high ICC at rest (SWA 0.994, CI95 0.987, 0.997; ACT 0.998, CI95 0.996, 0.999) and during ergocycling (SWA 0.941, CI95 0.873, 0.975; ACT 0.854, CI95 0.687, 0.939). Low CV were observed for both the SWA and the ACT at rest (1.8% CI95 1.3, 2.3 and 1.1% CI95 0.8, 1.4 respectively). CV was smaller for the SWA (12.1% CI95 8.8, 15.4) compared with ACT (20.5% CI95 14.0, 26.9) during the ergocycling exercise. Increases in EE over the 10-h period from visit 3 to visit 4 to visit 5 followed the increased intensities on the treadmill, as designed. 4. Discussion This study offers further validation on the accuracy of two monitors for estimating TEE and EE during rest and specific activities in healthy adults under free-living and control conditions. The design included out-clinic and in-clinic assessments and concomitant estimations allowing direct comparisons.
Table 1 Comparisons of energy expenditure estimated by activity monitors (SWA and ACT) to a reference method (DLW or indirect calorimetry). Energy expenditure
Mean difference
Pearson’s correlation
Intra-class correlation
r
p
R
CI95%
2.8 ± 12.4 −9.6 ± 9.9
0.804 0.877
0.00003 0.00003
0.892 0.906
0.705–0.956 0.757–0.964
12.4 ± 6.8 7.5 ± 7.7
0.891 0.856
0.717 0.674
0.267–0.891 0.154–0.874
−37.0 ± 25.8 −374.2 ± 206.6
0.796 0.356
0.00005 0.135
0.886 0.388
0.705–0.956 −0.588 to 0.764
kcal
kcal
TEE (kcal/day) Reference method: DLWa SWA ACT
2711 ± 507 2806 ± 514 2468 ± 361*
94 ± 319 −244 ± 258
REE (kcal/day) Reference method: Indirect calorimetry SWA ACT
1487 ± 246 1697 ± 250* 1611 ± 249*
210 ± 116 124 ± 133
Ergocycle at 50% of VO2peak (kcal)a,b Reference method: Indirect calorimetry SWA ACT
358 ± 116 266 ± 84* 89 ± 53*
Treadmill at 3.1 km/ha (kcal) Reference method: Indirect calorimetry SWA ACT
136 ± 40 191 ± 52* 220 ± 64*
55 ± 50 84 ± 40
23.3 ± 36.3 36.5 ± 12.4
0.443 0.796
0.07 0.00008
0.598 0.835
−0.074 to 0.850 0.559–0.938
Treadmill at 4.3 km/ha (kcal) Reference method: Indirect calorimetry SWA ACT
173 ± 57 240 ± 61* 241 ± 48*
67 ± 38 68 ± 25
28.0 ± 12.8 29.3 ± 12.9
0.794 0.897
0.00008 <0.0001
0.885 0.940
0.692–0.957 0.838–0.977
Treadmill at 6.4 km/ha (kcal) Reference method: Indirect calorimetry SWA ACT
255 ± 99 309 ± 73* 327 ± 100*
54 ± 46 71 ± 47
19.2 ± 13.8 21.8 ± 16.9
0.895 0.89
<0.0001 <0.0001
0.923 0.941
0.800–0.970 0.848–0.977
−93 ± 65 −269 ± 111
%
<0.0001 <0.0001
Data are means ± standard deviation. * Significantly different from DLW or indirect calorimetry (p < 0.05). a Missing data for 1 participant. b Data obtained during visit 3. ACT = Actical, DLW = doubly labeled water, REE = resting energy expenditure, SWA = SenseWear Armband, TEE = total energy expenditure, CI95% = 95% confidence interval.
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
G Model
ARTICLE IN PRESS
JSAMS-1104; No. of Pages 5
A.-S. Brazeau et al. / Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
4
A)
TEE by SWA - TEE by DLW
1000 +1.96 SD 720,3 500 Mean 0
94,2
-500
-1.96 SD -531,9
-1000 2000
2500
3000
3500
4000
4500
Mean of TEE by SWA and DLW
B) 600
TEE by ACT - TEE by DLW
400 +1.96 SD 261,3
200 0
Mean
-200
-243,5 -400 -600 -1.96 SD -748,4
-800 -1000 1500
2000
2500
3000
3500
4000
Mean of TEE by ACT and DLW Fig. 1. Bland–Altman plots. (A) TEE estimated by the SenseWear Armband and measured by doubly labeled water. (B) TEE estimated by the ACT and measured by doubly labeled water. ACT = Actical, DLW = doubly labeled water, REE = resting energy expenditure, SWA = SenseWear Armband, TEE = total energy expenditure.
Energy expenditure (kcal)
Similar to previous studies, the SWA presented a good estimate of TEE when compared to DLW with less than 100 kcal mean daily difference. A 7-day period offered TEE’s estimate as accurate as 10–14 days2,4 and, as previously observed,2 the SWA’s software version 6.1 showed an improved accuracy compared to the version 4.2 in the study by St-Onge et al.4 Even if REE was adequately REE Sedentary activities Ergocycle Treadmill
1400 1200 1000 800 600 400 200 0
V3
V4
SWA
V5
V3
V4
V5
ACT
Fig. 2. Repartition of energy expenditure during the 10-h period (V refers to visit number). ACT = Actical, DLW = doubly labeled water, REE = resting energy expenditure, SWA = SenseWear Armband.
estimated by the SWA in studies using a previous software’s version,21 similar to our findings, more recent studies support an overestimation of REE by approximately 12% in healthy older people.22,23 This study is the first, to our knowledge, to propose a validation of TEE estimate by the ACT using DLW. In this group of young adults, ACT significantly underestimated by 244 kcal the daily TEE. Oppositely, REE was significantly overestimated by the ACT compared to indirect calorimetry. This overestimation was greater than previously observed in adults (<32 kcal).8 The large standard deviations observed for all the differences between activity monitors and objective EE assessment indicates large inter-individual variations. When comparing to other validation studies performed among Caucasian individuals, reasons for difference in EE estimations may result in the variety of ethnicity of the participants.13 Since REE estimations relied on prediction equations,24 which have shown discrepancies between different ethnicities,25,26 it may explain part of the differences between our results and others. For specific exercises, both devices presented similar accuracy on the treadmill, except the SWA at the lowest speed. Both devices were more accurate at a higher speed. This was previously reported by Spierer et al.7 who observed that the ACT was more accurate at jogging speeds compared to walking. The increase accuracy in EE estimations associated with increased speeds make it difficult to detect accurate changes in EE associated with increased speed walk. For example, in this study, the treadmill speed increased from a mean of 3.1 (visit 3) to 4.3 (visit 4) to 6.4 km/h (visit 5). Accordingly, ACT was more accurate between visit 4 and 5 as compared to indirect calorimetry (ACT: 80 ± 59 vs. indirect calorimetry: 81 ± 59 kcal) than from visit 3 and 4 (ACT: 18 ± 39 vs. indirect calorimetry: 35 ± 27 kcal). SWA was more accurate in EE estimation as speed increased. One limitation of accelerometers is their inability to measure EE when there is no acceleration as confirmed by the incapacity of the ACT to show increased EE during ergocycling, a static activity. However, the SWA, without great accuracy, indicated increased EE during static activities as shown previously in other studies.15,27 Both monitors were reliable to estimate REE, EE during ergocycling and for the sedentary tasks performed during the remaining 7.5-h period. By design, we observed significant differences during the treadmill exercises. The SWA has been previously shown to offer reproducible values over a wide range of activities in humans.3,12 The ACT has been shown to provide reproducible measures in a mechanical setting.10 This study offers an additional validation of the ACT by showing adequate reliability in day-to-day activities. The acceptability to wear any of these devices was adequate as no participants reported any burden during the 10-day trial. Both devices had acceptable practicability28 as they require minimal time for staff training and it is easy to upload patient information and to download data. However, the SWA presents the advantage that data can be downloaded directly to the computer using a USB cable compared to the ACT which needs an intermediate port. Moreover, SWA’s software proposes a more user-friendly report of the results. The present study has several limitations. Our study design would have been improved by additional measures of REE. Each subject was only measured once, thereby not allowing estimating the intra-individual variation in REE. This validation study only tested two standardized activities at four different intensities. Additional studies using other activities may be needed to extend the validation. During the free-living period, we did not ask the participants to report their water activities, although we asked them to restrain from these activities. We did not impute for the missing
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
G Model JSAMS-1104; No. of Pages 5
ARTICLE IN PRESS A.-S. Brazeau et al. / Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
2.3% of the time over the 7 days that the two PA monitors were not worn. Also, although each participant used the same device for the entire protocol, multiple ACT and SWA were used throughout the study. However, it should be noted that multiple devices are often used in research studies. Larger sample size would allow sub-group analysis (e.g. impact of sex, age, body mass index or ethnicity). Accuracy and reliability of activity monitors in obese people should also be assessed as they would greatly benefit from the changes in EE (e.g. overweight or obese patients with cardio metabolic complications). As reported previously, REE was quite variable among body mass index categories.24 This direct comparison of two activity monitors showed differences in accuracy between monitors. The SWA seems to better predict TEE than the ACT. Consequently, the ACT should be limited to expressing PA duration and intensity using counts cut-offs.29,30 The selection of one monitor over the other depends on the research objectives. For example, in the context of population surveillance, accuracy and reliability of an activity monitor has to be addressed over days and not only for specific physical activities. Additionally, in a research context, both activity monitors may benefit from a direct measure of REE by indirect calorimetry to eliminate some bias in the EE estimations introduced by predictions equations. 5. Conclusion The SWA provided acceptable estimations of TEE under freeliving conditions but not the ACT. Both devices were reliable but not accurate for REE and for specific physical activities. Since accuracy varies according to the types and/or intensities of exercises, it may be difficult to evaluate with exactness the size of the changes in EE associated with the changes in the type of exercise or intensity. Practical implications • This study provides an additional validation of the ACT and SWA. • The differences between EE estimation by the two devices indicate that their data cannot be pooled. • ACT should expressed physical activity duration and intensity using counts cut-offs. Acknowledgments We thank Diane Mignault for DLW measurements. This work was supported by the J-A De Sève research chair awarded to RRL. At the time of the study, ASB held a scholarship from the Canadian Institute for Health Research (CIHR) and RRL was awarded by the Fonds de Recherches du Québec en Santé (FRQ-S). The authors declare that they have no conflict of interest. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jsams.2014.11.001. References 1. Craig CL, Marshall AL, Sjostrom M et al. International physical activity questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003; 35(8):1381–1395.
5
2. Johannsen DL, Calabro MA, Stewart J et al. Accuracy of armband monitors for measuring daily energy expenditure in healthy adults. Med Sci Sports Exerc 2010; 42(11):2134–2140. 3. Mackey DC, Manini TM, Schoeller DA et al. Validation of an armband to measure daily energy expenditure in older adults. J Gerontol A Biol Sci Med Sci 2011; 66(10):1108–1113. 4. St-Onge M, Mignault D, Allison DB et al. Evaluation of a portable device to measure daily energy expenditure in free-living adults. Am J Clin Nutr 2007; 85(3):742–749. 5. Tremblay M, Langlois R, Bryan S et al. Canadian Health Measures Survey pre-test: design, methods, results. Health Rep 2007; 18(Suppl.):21–30. 6. Lyden K, Kozey SL, Staudenmeyer JW et al. A comprehensive evaluation of commonly used accelerometer energy expenditure and MET prediction equations. Eur J Appl Physiol 2011; 111(2):187–201. 7. Spierer DK, Hagins M, Rundle A et al. A comparison of energy expenditure estimates from the Actiheart and Actical physical activity monitors during low intensity activities, walking, and jogging. Eur J Appl Physiol 2011; 111(4):659–667. 8. Dellava JE, Hoffman DJ. Validity of resting energy expenditure estimated by an activity monitor compared to indirect calorimetry. Br J Nutr 2009; 102(1):155–159. 9. Crouter SE, Churilla JR, Bassett Jr DR. Estimating energy expenditure using accelerometers. Eur J Appl Physiol 2006; 98(6):601–612. 10. Esliger DW, Tremblay MS. Technical reliability assessment of three accelerometer models in a mechanical setup. Med Sci Sports Exerc 2006; 38(12):2173–2181. 11. Melanson Jr EL, Freedson PS. Physical activity assessment: a review of methods. Crit Rev Food Sci Nutr 1996; 36(5):385–396. 12. Brazeau AS, Karelis AD, Mignault D et al. Test-retest reliability of a portable monitor to assess energy expenditure. Appl Physiol Nutr Metab 2011; 36(3):339–343. 13. Fruin ML, Rankin JW. Validity of a multi-sensor armband in estimating rest and exercise energy expenditure. Med Sci Sports Exerc 2004; 36(6): 1063–1069. 14. Harrison SL, Horton EJ, Smith R et al. Physical activity monitoring: addressing the difficulties of accurately detecting slow walking speeds. Heart Lung 2013; 42(5):361–364, e361. 15. Benito PJ, Neiva C, Gonzalez-Quijano PS et al. Validation of the SenseWear armband in circuit resistance training with different loads. Eur J Appl Physiol 2012; 112(8):3155–3159. 16. Drenowatz C, Eisenmann JC. Validation of the SenseWear Armband at high intensity exercise. Eur J Appl Physiol 2011; 111(5):883–887. 17. Farooqi N, Slinde F, Haglin L et al. Validation of SenseWear Armband and ActiHeart monitors for assessments of daily energy expenditure in freeliving women with chronic obstructive pulmonary disease. Physiol Rep 2013; 1(6):e00150. 18. John D, Freedson P. ActiGraph and Actical physical activity monitors: a peek under the hood. Med Sci Sports Exerc 2012; 44(1 Suppl. 1):S86–S89. 19. Crouter SE, Bassett Jr DR. A new 2-regression model for the Actical accelerometer. Br J Sports Med 2008; 42(3):217–224. 20. Weir JB. New methods for calculating metabolic rate with special reference to protein metabolism. J Physiol 1949; 109(1–2):1–9. 21. Malavolti M, Pietrobelli A, Dugoni M et al. A new device for measuring resting energy expenditure (REE) in healthy subjects. Nutr Metab Cardiovasc Dis 2007; 17(5):338–343. 22. Heiermann S, Khalaj Hedayati K, Muller S et al. Accuracy of a portable multisensor body monitor for predicting resting energy expenditure in older people: a comparison with indirect calorimetry. Gerontology 2011; 57(5):473–479. 23. Casiraghi F, Lertwattanarak R, Luzi L et al. Energy expenditure evaluation in humans and non-human primates by sensewear armband. Validation of energy expenditure evaluation by sensewear armband by direct comparison with indirect calorimetry. PLOS ONE 2013; 8(9):e73651. 24. Hasson RE, Howe CA, Jones BL et al. Accuracy of four resting metabolic rate prediction equations: effects of sex, body mass index, age, and race/ethnicity. J Sci Med Sport 2011; 14(4):344–351. 25. Frankenfield D, Roth-Yousey L, Compher C. Comparison of predictive equations for resting metabolic rate in healthy nonobese and obese adults: a systematic review. J Am Diet Assoc 2005; 105(5):775–789. 26. Brazeau AS, Suppere C, Strychar I et al. Accuracy of energy expenditure estimation by activity monitors differs with ethnicity. Int J Sports Med 2014. 27. Brazeau AS, Karelis AD, Mignault D et al. Accuracy of the SenseWear Armband during ergocycling. Int J Sports Med 2011; 32(10):761–764. 28. Fitzpatrick R, Davey C, Buxton MJ et al. Evaluating patient-based outcome measures for use in clinical trials. Health Technol Assess 1998; 2(14):i–iv, 1–74. 29. Wong SL, Colley R, Connor Gorber S et al. Actical accelerometer sedentary activity thresholds for adults. J Phys Act Health 2011; 8(4):587–591. 30. Colley RC, Tremblay MS. Moderate and vigorous physical activity intensity cutpoints for the Actical accelerometer. J Sports Sci 2011; 29(8):783–789.
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
ARTICLE IN PRESS Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Journal of Science and Medicine in Sport journal homepage: www.elsevier.com/locate/jsams
Original research
Validation and reliability of two activity monitor for energy expenditure assessment Anne-Sophie Brazeau a , Nadia Beaudoin a,b , Virginie Bélisle a , Virginie Messier a , Antony D. Karelis c , Rémi Rabasa-Lhoret a,b,d,e,∗ a
Montreal Clinical Research Institute (IRCM), Montreal, Quebec, Canada Department of Medicine, Université de Montréal, Montreal, Quebec, Canada c Department of Kinanthropology, University of Quebec in Montreal, Montreal, Quebec, Canada d Department of Nutrition, Université de Montréal, Montreal, Quebec, Canada e Montreal Diabetes Research Center (MDRC) of the Université de Montréal Hospital Research Center (CR-CHUM), Montreal, Quebec, Canada b
a r t i c l e
i n f o
Article history: Received 11 July 2014 Received in revised form 10 October 2014 Accepted 1 November 2014 Available online xxx Keywords: Test–retest Activity monitoring Accelerometer Motion sensors Doubly labeled water
a b s t r a c t Objectives: This study explores the reliability and validity of the SenseWear Armband (SWA) and Actical (ACT) for free-living total energy expenditure, and energy expenditure during rest and light-to-moderate exercises (walking, ergocycling). Design: Participants wore the 2 devices during 7 days (free-living) and then participated to 3 days of testing in our laboratory. Methods: SWA and ACT estimates of total energy expenditure was compared to the doubly labeled water technique (7 days), and energy expenditure during rest (60 min), treadmill (45 min; intensities ∼22% to ∼41% VO2peak ) and ergocycling (45 min; ∼50% VO2peak ) were compared to indirect calorimetry over the following 3 days. Paired T-tests and intra-class correlation coefficient (ICC) with 95% confidence interval (CI95 ) were computed. Results: Twenty adults were recruited (BMI 23.1 ± 2.3 kg/m2 ). Compared to doubly labelled water, SWA overestimated energy expenditure by 94 kcal/d (±319; P = 0.2) and ACT underestimated by −244 kcal/d (±258; P = 0.001). Energy expenditure during rest (SWA 210 ± 116, ACT 124 ± 133 kcal/d; p < 0.05) and treadmill (according on intensity: SWA 54 ± 46 to 67 ± 38, ACT 68 ± 25 to 84 ± 40 kcal; p < 0.05) were overestimated and underestimated during ergocycling (SWA −93 ± 65, ACT −269 ± 111 kcal; p < 0.05) compared to indirect calorimetry. High ICC were observed at rest (SWA 0.994 CI95 0.987–0.997; ACT 0.998 CI95 0.996–0.999) and during ergocycling (SWA 0.941 CI95 0.873–0.975; ACT 0.854 CI95 0.687–0.939). Conclusion: Acceptable estimation of total energy expenditure was observed with the SWA. Both devices were reliable but not accurate for energy expenditure’s estimations during rest and for specific exercises. © 2014 Sports Medicine Australia. Published by Elsevier Ltd. All rights reserved.
1. Introduction It is important in clinical, nutritional and exercise sciences as well as in epidemiological research to develop simple and accurate devices for the evaluation of energy expenditure (EE). There are many methods available to estimate EE depending on financial and human resources as well as on research objectives. On one side, there are self-report measures such as activity diaries and questionnaires which are relatively inexpensive but also less
∗ Corresponding author. E-mail address: [email protected] (R. Rabasa-Lhoret).
accurate mainly due to recall bias.1 On the opposite, objective methods, such as doubly labeled water and calorimetry chambers, are more accurate but also more expensive. In between, there are portable activity monitors, such as accelerometers and motion sensors, which objectively estimate total energy expenditure (TEE)2 and provide detailed physical activity (PA) information (duration and intensity).3,4 They offer a unique opportunity to capture freeliving EE. Many devices are commercially available, including the Actigraph, Actical (ACT) and SenseWear Armband (SWA). The ACT, an omnidirectional accelerometer, is used, for example, for population surveillance.5 This device produces «count» value as its output. To allow comparison, counts are processed using predictions
http://dx.doi.org/10.1016/j.jsams.2014.11.001 1440-2440/© 2014 Sports Medicine Australia. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
G Model JSAMS-1104; No. of Pages 5
ARTICLE IN PRESS A.-S. Brazeau et al. / Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
2
equations into METS (i.e. metabolic equivalent) or calories. Validation studies among healthy adults, using published equations, have provided inconsistent results.6,7 The ACT provided valid resting energy expenditure (REE) estimation compared to indirect calorimetry for normal to overweight adults,8 but discrepancies were observed for EE during short PA periods.7,9 In a mechanical set-up, the ACT presented a good intra-instrument reliability (coefficient of variation (CV) = 0.5%).10 Despite its wide use, the ability of ACT to evaluate with accuracy the daily TEE has not been established. To our knowledge, this device and its software has not been validated against doubly labeled water (DLW), the gold standard to evaluate TEE.11 The SWA is another device commercially available. This device includes a 2-axis accelerometer but also considers physiological measures (e.g. skin temperature) to estimate EE. It showed good reliabilities over a large range of activities, from rest2,4 to cycling12,13 but underestimated EE during low speed walking on treadmill14 and during high intensity exercises.15,16 Its validity to estimate TEE has been assessed using the DLW technique; improved precision was found with newer versions of the sensor (Pro3) and its software (Innerview 6.1)2,3,17 compared to previous (Pro 2; Innerview 4.1),4 but SWA still underestimates TEE compared to DLW (−1122 and −1174 kcal/day). Comparisons between the two software versions have been previously published.3,17 Technologies improve rapidly and additional independent validation studies are needed to confirm the accuracy of these newer generation devices and software. Moreover, to compare EE estimated by these devices (i.e., comparing one only measuring accelerations to one measuring accelerations but also capturing physiological measures), head to head validation is required. Freeliving studies allow evaluating accuracy in real-life condition where intensities and type of activities varies, which is more representative of an actual lifestyle. Laboratory studies allow for controlling additional parameters, thus extending our understanding of the various limits of the devices (e.g., such as bicycling). We aimed at investigating the accuracy of the ACT and the SWA (i) for free-living TEE (main objective), (ii) for EE during different intensities of PA, (iii) for REE. A sub-objective was to examine the reproducibility of these two devices for EE during ergocycling (iv) and rest (v). We hypothesized that both devices would be accurate for TEE and REE.
2. Methods The project was designed to evaluate the SWA and the ACT in out-clinic (i.e., real life) and in-clinic (i.e., in the laboratory) settings (see supplementary figure). Participants wore the 2 devices for a 7-day period of free-living (out-clinic) and then participated to 3 days of testing in our laboratory (in-clinic). During the outclinic period, we evaluated TEE against DLW. During the last 3 visits, we evaluated the accuracy of these devices to estimate REE and EE at different intensities on the treadmill and during ergocycling against indirect calorimetry. We also examined the SWA and ACT’s reliability for EE during rest and ergocycling. The Institut de recherches cliniques de Montreal ethic review board approved the study. Written informed consent was obtained prior to data collection. Twenty healthy adults (85% Caucasian) participated in the study. They were recruited through word of mouth. Inclusion criteria were: aged between 18 and 45 years old, normal or overweight, non-smoker, without major disease, illness or medications use that would impact metabolism. During the first visit, the participants underwent a medical exam. Body weight and height were measured (light clothing, shoes removed). Then participants performed a graded exercise test on an ergocycle Ergoline 900 (Bitz, Germany) until voluntary exhaustion to measure their VO2 peak . During the
test, power output was increased by 25 W every 3 min. The highest oxygen consumption obtained was considered the VO2 peak (mL/kg/min). Breath-by-breath gas samples were analyzed using an Ergocard (software version 6, MediSoft, Dinant, Belgium) cardiopulmonary exercise test station. Participants were then trained on how to wear the SWA and the ACT according to the manufacturers’ instructions. Participants were asked to wear them for the complete study period (i.e., 10 days), to only remove them when showering and to restrain from water activities. The SWA Pro 3 (Bodymedia, Pittsburgh, PA) was worn on the upper right arm (on the triceps at the mid-humerus point). The SWA captures data through a 2-axis accelerometer and heat flux, galvanic skin response, skin temperature, and near-body ambient temperature sensors. The data as well as gender, age, body weight, height, handedness and smoking status is used to calculate EE using the Innerview Research Software version 6.1 developed by the manufacturer. The Actical device (Mini Mitter, Bend, OR) is a small electronic omni-directional monitor capturing movements in multiple planes, although it is most sensitive to vertical accelerations when worn on the hip.18 This sensor uses a piezoelectric technology to register accelerations called activity counts. Participants wore this device on the hip, positioned laterally above the iliac crest and secured with an elastic belt. The device was initialized using 60 s epochs. Data was analyzed using the manufacturer software (version 2.1) and counts were transformed to EE measures using a 2-regression equation.19 During this first visit, urine samples were collected and isotope solutions for DLW measurement were administered to assess TEE over a 7-d period. Each participant ingested 0.3 g/kg body weight of 2 H18 O. Complete procedures are described elsewhere.4 Briefly, urine sample obtained on visit 1 were used to assess the amount of tracers ingested. The four other samples (2 on visit 2 and 2 on visit 3) were used to evaluate the rate of disappearance of 2 H and 18 O. All samples were measured in triplicate. An Isoprime Stable Isotope Ratio Mass Spectrometer connected to a Multiflow-Bio module for Isoprime and a Bilson 222XL Autosampler (GV Instruments, Manchester, United Kingdom) were used for the measurement of TEE. Data were processed with the MassLynx 3.6 software (Waters Corp, Milford, MA). Stability tests were performed each day, which yield a standard deviation of 0.026% for 2 H and 0.004% for 18 O. The next day (visit 2), participants came to provide urine samples. A week after the first visit, following the 7 days out-clinic period, participants participated to the 3-day (10-h/day) in-clinic reliability trial. Each day, they followed, under supervision and with a strict time schedule, the same program of activities. They arrived, fasting, at the laboratory at 07h30 and participated to 60 min of lying awake and two periods of 45-min standardized physical activities (treadmill: intensities from ∼22% to ∼41% VO2peak ; ergocycle: ∼50% VO2peak ). The 3 days were identical except for the walking speed on the treadmill which increased between each visit to induce higher EE allowing validation over a wider range of speeds. Over the residual 7.5 h, which included the standardized breakfast, lunch and two snacks, participants remained seated. On the first day, they completed a chart of their sedentary activities (e.g. reading, watching TV) in order to reproduce them on the subsequent days. REE was measured during a fasting state by indirect calorimetry during one of the 3 mornings of the in-clinic trial. Carbon dioxide and oxygen were measured with a SensorMedics Delta Track II with the ventilated-hood technique. Measurement of gas concentrations were used to determine the 24-h REE with the Weir equation.20 Participants were lying in a supine position, without speaking or sleeping for 60 min. The first 10-min was an acclimatization period and was discarded. The last 50 min were used in the analysis. The calorimeter gas analyzers were calibrated prior each measurement.
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
G Model
ARTICLE IN PRESS
JSAMS-1104; No. of Pages 5
A.-S. Brazeau et al. / Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
During the standardized physical activities, EE was evaluated, at 1-min intervals, by indirect calorimetry with the same ergocard used to assess VO2peak . Data was analyzed with the Statistical Package for Social Sciences 20.0 (SPSS, Inc., Chicago, IL) and the MedCalc Software version 9 (Mariakerke, Belgium). All study variables were examined for normality of distribution prior to analysis (Kolmogorov–Smirnov). Data are presented as means ± standard deviation. The extent of agreement between measures of EE was evaluated with Pearson correlations, ICC and Bland and Altman plots. Mean differences between methods were also compared with paired t-tests. To evaluate the reliability ICC and CV were computed with 95% confidence interval (CI95 ). 3. Results Twenty healthy adults (85% Caucasian) participated in the study (age: 26.2 ± 3.6 years; body mass index was 23.1 ± 2.3 kg/m2 (range 20.0–30.3 kg/m2 ; 15% >25 kg/m2 ) and VO2peak 41.0 ± 11.6 mL/kg/min (range 22.6–65.7). The SWA was worn 97.7% of the time during the 7-day out-clinic period and 100% of the time during the three 10-h periods (in-clinic). The ACT was also worn 100% of the time during the in-clinic period (direct observation). However, given that the ACT does not capture information on wear time this information is not available for the 7-day out-clinic period. As per instructions, it should be similar to the SWA. No discomfort or negative effects were reported for the two devices. Table 1 showed comparisons of EE estimations. Mean TEE over 7 days estimated by both the SWA and ACT were significantly correlated with TEE measured with DLW. Bland–Altman analysis (Fig. 1) shows that there was no tendency of over or underestimation with SWA across the wide range of TEE. Meanwhile ACT showed a
3
tendency to increased underestimation with higher TEE. Accordingly, ACT significantly underestimated TEE by a mean of 244 kcal compared to DLW. Although a significant relationship between REE estimated by both devices with REE measured by indirect calorimetry was shown, SWA and ACT overestimated REE by 210 kcal (12.4%) and 124 kcal (7.4%), respectively. Most of the EE estimated by the 2 devices during the standardized exercises were significantly correlated with the indirect calorimetry measures (r = 0.80–0.90; p < 0.05) except for SWA during the lowest speed walk (mean 3.1 km/h) and ACT during ergocycling. For every intensity that was tested both SWA and ACT overestimated EE during the treadmill exercise (see supplementary online figures). Fig. 2 shows the repartition of EE during the 10-h periods spent in the research facility and illustrates that REE and EE during ergocycling and sedentary activities were reproducible over the three visits. Both devices showed high ICC at rest (SWA 0.994, CI95 0.987, 0.997; ACT 0.998, CI95 0.996, 0.999) and during ergocycling (SWA 0.941, CI95 0.873, 0.975; ACT 0.854, CI95 0.687, 0.939). Low CV were observed for both the SWA and the ACT at rest (1.8% CI95 1.3, 2.3 and 1.1% CI95 0.8, 1.4 respectively). CV was smaller for the SWA (12.1% CI95 8.8, 15.4) compared with ACT (20.5% CI95 14.0, 26.9) during the ergocycling exercise. Increases in EE over the 10-h period from visit 3 to visit 4 to visit 5 followed the increased intensities on the treadmill, as designed. 4. Discussion This study offers further validation on the accuracy of two monitors for estimating TEE and EE during rest and specific activities in healthy adults under free-living and control conditions. The design included out-clinic and in-clinic assessments and concomitant estimations allowing direct comparisons.
Table 1 Comparisons of energy expenditure estimated by activity monitors (SWA and ACT) to a reference method (DLW or indirect calorimetry). Energy expenditure
Mean difference
Pearson’s correlation
Intra-class correlation
r
p
R
CI95%
2.8 ± 12.4 −9.6 ± 9.9
0.804 0.877
0.00003 0.00003
0.892 0.906
0.705–0.956 0.757–0.964
12.4 ± 6.8 7.5 ± 7.7
0.891 0.856
0.717 0.674
0.267–0.891 0.154–0.874
−37.0 ± 25.8 −374.2 ± 206.6
0.796 0.356
0.00005 0.135
0.886 0.388
0.705–0.956 −0.588 to 0.764
kcal
kcal
TEE (kcal/day) Reference method: DLWa SWA ACT
2711 ± 507 2806 ± 514 2468 ± 361*
94 ± 319 −244 ± 258
REE (kcal/day) Reference method: Indirect calorimetry SWA ACT
1487 ± 246 1697 ± 250* 1611 ± 249*
210 ± 116 124 ± 133
Ergocycle at 50% of VO2peak (kcal)a,b Reference method: Indirect calorimetry SWA ACT
358 ± 116 266 ± 84* 89 ± 53*
Treadmill at 3.1 km/ha (kcal) Reference method: Indirect calorimetry SWA ACT
136 ± 40 191 ± 52* 220 ± 64*
55 ± 50 84 ± 40
23.3 ± 36.3 36.5 ± 12.4
0.443 0.796
0.07 0.00008
0.598 0.835
−0.074 to 0.850 0.559–0.938
Treadmill at 4.3 km/ha (kcal) Reference method: Indirect calorimetry SWA ACT
173 ± 57 240 ± 61* 241 ± 48*
67 ± 38 68 ± 25
28.0 ± 12.8 29.3 ± 12.9
0.794 0.897
0.00008 <0.0001
0.885 0.940
0.692–0.957 0.838–0.977
Treadmill at 6.4 km/ha (kcal) Reference method: Indirect calorimetry SWA ACT
255 ± 99 309 ± 73* 327 ± 100*
54 ± 46 71 ± 47
19.2 ± 13.8 21.8 ± 16.9
0.895 0.89
<0.0001 <0.0001
0.923 0.941
0.800–0.970 0.848–0.977
−93 ± 65 −269 ± 111
%
<0.0001 <0.0001
Data are means ± standard deviation. * Significantly different from DLW or indirect calorimetry (p < 0.05). a Missing data for 1 participant. b Data obtained during visit 3. ACT = Actical, DLW = doubly labeled water, REE = resting energy expenditure, SWA = SenseWear Armband, TEE = total energy expenditure, CI95% = 95% confidence interval.
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
G Model
ARTICLE IN PRESS
JSAMS-1104; No. of Pages 5
A.-S. Brazeau et al. / Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
4
A)
TEE by SWA - TEE by DLW
1000 +1.96 SD 720,3 500 Mean 0
94,2
-500
-1.96 SD -531,9
-1000 2000
2500
3000
3500
4000
4500
Mean of TEE by SWA and DLW
B) 600
TEE by ACT - TEE by DLW
400 +1.96 SD 261,3
200 0
Mean
-200
-243,5 -400 -600 -1.96 SD -748,4
-800 -1000 1500
2000
2500
3000
3500
4000
Mean of TEE by ACT and DLW Fig. 1. Bland–Altman plots. (A) TEE estimated by the SenseWear Armband and measured by doubly labeled water. (B) TEE estimated by the ACT and measured by doubly labeled water. ACT = Actical, DLW = doubly labeled water, REE = resting energy expenditure, SWA = SenseWear Armband, TEE = total energy expenditure.
Energy expenditure (kcal)
Similar to previous studies, the SWA presented a good estimate of TEE when compared to DLW with less than 100 kcal mean daily difference. A 7-day period offered TEE’s estimate as accurate as 10–14 days2,4 and, as previously observed,2 the SWA’s software version 6.1 showed an improved accuracy compared to the version 4.2 in the study by St-Onge et al.4 Even if REE was adequately REE Sedentary activities Ergocycle Treadmill
1400 1200 1000 800 600 400 200 0
V3
V4
SWA
V5
V3
V4
V5
ACT
Fig. 2. Repartition of energy expenditure during the 10-h period (V refers to visit number). ACT = Actical, DLW = doubly labeled water, REE = resting energy expenditure, SWA = SenseWear Armband.
estimated by the SWA in studies using a previous software’s version,21 similar to our findings, more recent studies support an overestimation of REE by approximately 12% in healthy older people.22,23 This study is the first, to our knowledge, to propose a validation of TEE estimate by the ACT using DLW. In this group of young adults, ACT significantly underestimated by 244 kcal the daily TEE. Oppositely, REE was significantly overestimated by the ACT compared to indirect calorimetry. This overestimation was greater than previously observed in adults (<32 kcal).8 The large standard deviations observed for all the differences between activity monitors and objective EE assessment indicates large inter-individual variations. When comparing to other validation studies performed among Caucasian individuals, reasons for difference in EE estimations may result in the variety of ethnicity of the participants.13 Since REE estimations relied on prediction equations,24 which have shown discrepancies between different ethnicities,25,26 it may explain part of the differences between our results and others. For specific exercises, both devices presented similar accuracy on the treadmill, except the SWA at the lowest speed. Both devices were more accurate at a higher speed. This was previously reported by Spierer et al.7 who observed that the ACT was more accurate at jogging speeds compared to walking. The increase accuracy in EE estimations associated with increased speeds make it difficult to detect accurate changes in EE associated with increased speed walk. For example, in this study, the treadmill speed increased from a mean of 3.1 (visit 3) to 4.3 (visit 4) to 6.4 km/h (visit 5). Accordingly, ACT was more accurate between visit 4 and 5 as compared to indirect calorimetry (ACT: 80 ± 59 vs. indirect calorimetry: 81 ± 59 kcal) than from visit 3 and 4 (ACT: 18 ± 39 vs. indirect calorimetry: 35 ± 27 kcal). SWA was more accurate in EE estimation as speed increased. One limitation of accelerometers is their inability to measure EE when there is no acceleration as confirmed by the incapacity of the ACT to show increased EE during ergocycling, a static activity. However, the SWA, without great accuracy, indicated increased EE during static activities as shown previously in other studies.15,27 Both monitors were reliable to estimate REE, EE during ergocycling and for the sedentary tasks performed during the remaining 7.5-h period. By design, we observed significant differences during the treadmill exercises. The SWA has been previously shown to offer reproducible values over a wide range of activities in humans.3,12 The ACT has been shown to provide reproducible measures in a mechanical setting.10 This study offers an additional validation of the ACT by showing adequate reliability in day-to-day activities. The acceptability to wear any of these devices was adequate as no participants reported any burden during the 10-day trial. Both devices had acceptable practicability28 as they require minimal time for staff training and it is easy to upload patient information and to download data. However, the SWA presents the advantage that data can be downloaded directly to the computer using a USB cable compared to the ACT which needs an intermediate port. Moreover, SWA’s software proposes a more user-friendly report of the results. The present study has several limitations. Our study design would have been improved by additional measures of REE. Each subject was only measured once, thereby not allowing estimating the intra-individual variation in REE. This validation study only tested two standardized activities at four different intensities. Additional studies using other activities may be needed to extend the validation. During the free-living period, we did not ask the participants to report their water activities, although we asked them to restrain from these activities. We did not impute for the missing
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001
G Model JSAMS-1104; No. of Pages 5
ARTICLE IN PRESS A.-S. Brazeau et al. / Journal of Science and Medicine in Sport xxx (2015) xxx–xxx
2.3% of the time over the 7 days that the two PA monitors were not worn. Also, although each participant used the same device for the entire protocol, multiple ACT and SWA were used throughout the study. However, it should be noted that multiple devices are often used in research studies. Larger sample size would allow sub-group analysis (e.g. impact of sex, age, body mass index or ethnicity). Accuracy and reliability of activity monitors in obese people should also be assessed as they would greatly benefit from the changes in EE (e.g. overweight or obese patients with cardio metabolic complications). As reported previously, REE was quite variable among body mass index categories.24 This direct comparison of two activity monitors showed differences in accuracy between monitors. The SWA seems to better predict TEE than the ACT. Consequently, the ACT should be limited to expressing PA duration and intensity using counts cut-offs.29,30 The selection of one monitor over the other depends on the research objectives. For example, in the context of population surveillance, accuracy and reliability of an activity monitor has to be addressed over days and not only for specific physical activities. Additionally, in a research context, both activity monitors may benefit from a direct measure of REE by indirect calorimetry to eliminate some bias in the EE estimations introduced by predictions equations. 5. Conclusion The SWA provided acceptable estimations of TEE under freeliving conditions but not the ACT. Both devices were reliable but not accurate for REE and for specific physical activities. Since accuracy varies according to the types and/or intensities of exercises, it may be difficult to evaluate with exactness the size of the changes in EE associated with the changes in the type of exercise or intensity. Practical implications • This study provides an additional validation of the ACT and SWA. • The differences between EE estimation by the two devices indicate that their data cannot be pooled. • ACT should expressed physical activity duration and intensity using counts cut-offs. Acknowledgments We thank Diane Mignault for DLW measurements. This work was supported by the J-A De Sève research chair awarded to RRL. At the time of the study, ASB held a scholarship from the Canadian Institute for Health Research (CIHR) and RRL was awarded by the Fonds de Recherches du Québec en Santé (FRQ-S). The authors declare that they have no conflict of interest. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jsams.2014.11.001. References 1. Craig CL, Marshall AL, Sjostrom M et al. International physical activity questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003; 35(8):1381–1395.
5
2. Johannsen DL, Calabro MA, Stewart J et al. Accuracy of armband monitors for measuring daily energy expenditure in healthy adults. Med Sci Sports Exerc 2010; 42(11):2134–2140. 3. Mackey DC, Manini TM, Schoeller DA et al. Validation of an armband to measure daily energy expenditure in older adults. J Gerontol A Biol Sci Med Sci 2011; 66(10):1108–1113. 4. St-Onge M, Mignault D, Allison DB et al. Evaluation of a portable device to measure daily energy expenditure in free-living adults. Am J Clin Nutr 2007; 85(3):742–749. 5. Tremblay M, Langlois R, Bryan S et al. Canadian Health Measures Survey pre-test: design, methods, results. Health Rep 2007; 18(Suppl.):21–30. 6. Lyden K, Kozey SL, Staudenmeyer JW et al. A comprehensive evaluation of commonly used accelerometer energy expenditure and MET prediction equations. Eur J Appl Physiol 2011; 111(2):187–201. 7. Spierer DK, Hagins M, Rundle A et al. A comparison of energy expenditure estimates from the Actiheart and Actical physical activity monitors during low intensity activities, walking, and jogging. Eur J Appl Physiol 2011; 111(4):659–667. 8. Dellava JE, Hoffman DJ. Validity of resting energy expenditure estimated by an activity monitor compared to indirect calorimetry. Br J Nutr 2009; 102(1):155–159. 9. Crouter SE, Churilla JR, Bassett Jr DR. Estimating energy expenditure using accelerometers. Eur J Appl Physiol 2006; 98(6):601–612. 10. Esliger DW, Tremblay MS. Technical reliability assessment of three accelerometer models in a mechanical setup. Med Sci Sports Exerc 2006; 38(12):2173–2181. 11. Melanson Jr EL, Freedson PS. Physical activity assessment: a review of methods. Crit Rev Food Sci Nutr 1996; 36(5):385–396. 12. Brazeau AS, Karelis AD, Mignault D et al. Test-retest reliability of a portable monitor to assess energy expenditure. Appl Physiol Nutr Metab 2011; 36(3):339–343. 13. Fruin ML, Rankin JW. Validity of a multi-sensor armband in estimating rest and exercise energy expenditure. Med Sci Sports Exerc 2004; 36(6): 1063–1069. 14. Harrison SL, Horton EJ, Smith R et al. Physical activity monitoring: addressing the difficulties of accurately detecting slow walking speeds. Heart Lung 2013; 42(5):361–364, e361. 15. Benito PJ, Neiva C, Gonzalez-Quijano PS et al. Validation of the SenseWear armband in circuit resistance training with different loads. Eur J Appl Physiol 2012; 112(8):3155–3159. 16. Drenowatz C, Eisenmann JC. Validation of the SenseWear Armband at high intensity exercise. Eur J Appl Physiol 2011; 111(5):883–887. 17. Farooqi N, Slinde F, Haglin L et al. Validation of SenseWear Armband and ActiHeart monitors for assessments of daily energy expenditure in freeliving women with chronic obstructive pulmonary disease. Physiol Rep 2013; 1(6):e00150. 18. John D, Freedson P. ActiGraph and Actical physical activity monitors: a peek under the hood. Med Sci Sports Exerc 2012; 44(1 Suppl. 1):S86–S89. 19. Crouter SE, Bassett Jr DR. A new 2-regression model for the Actical accelerometer. Br J Sports Med 2008; 42(3):217–224. 20. Weir JB. New methods for calculating metabolic rate with special reference to protein metabolism. J Physiol 1949; 109(1–2):1–9. 21. Malavolti M, Pietrobelli A, Dugoni M et al. A new device for measuring resting energy expenditure (REE) in healthy subjects. Nutr Metab Cardiovasc Dis 2007; 17(5):338–343. 22. Heiermann S, Khalaj Hedayati K, Muller S et al. Accuracy of a portable multisensor body monitor for predicting resting energy expenditure in older people: a comparison with indirect calorimetry. Gerontology 2011; 57(5):473–479. 23. Casiraghi F, Lertwattanarak R, Luzi L et al. Energy expenditure evaluation in humans and non-human primates by sensewear armband. Validation of energy expenditure evaluation by sensewear armband by direct comparison with indirect calorimetry. PLOS ONE 2013; 8(9):e73651. 24. Hasson RE, Howe CA, Jones BL et al. Accuracy of four resting metabolic rate prediction equations: effects of sex, body mass index, age, and race/ethnicity. J Sci Med Sport 2011; 14(4):344–351. 25. Frankenfield D, Roth-Yousey L, Compher C. Comparison of predictive equations for resting metabolic rate in healthy nonobese and obese adults: a systematic review. J Am Diet Assoc 2005; 105(5):775–789. 26. Brazeau AS, Suppere C, Strychar I et al. Accuracy of energy expenditure estimation by activity monitors differs with ethnicity. Int J Sports Med 2014. 27. Brazeau AS, Karelis AD, Mignault D et al. Accuracy of the SenseWear Armband during ergocycling. Int J Sports Med 2011; 32(10):761–764. 28. Fitzpatrick R, Davey C, Buxton MJ et al. Evaluating patient-based outcome measures for use in clinical trials. Health Technol Assess 1998; 2(14):i–iv, 1–74. 29. Wong SL, Colley R, Connor Gorber S et al. Actical accelerometer sedentary activity thresholds for adults. J Phys Act Health 2011; 8(4):587–591. 30. Colley RC, Tremblay MS. Moderate and vigorous physical activity intensity cutpoints for the Actical accelerometer. J Sports Sci 2011; 29(8):783–789.
Please cite this article in press as: Brazeau A-S, et al. Validation and reliability of two activity monitor for energy expenditure assessment. J Sci Med Sport (2015), http://dx.doi.org/10.1016/j.jsams.2014.11.001