Comparison of total energy expenditure assessed by

European Journal of Sport Science, 2014 http://dx.doi.org/10.1080/17461391.2014.949309

ORIGINAL ARTICLE

Comparison of total energy expenditure assessed by two devices in controlled and free-living conditions

SYLVIE ROUSSET1, ANTHONY FARDET1, PHILIPPE LACOMME2, SYLVIE NORMAND3, CHRISTOPHE MONTAURIER1, YVES BOIRIE1, & BÉATRICE MORIO1 INRA, Human Nutrition Unit UMR1019, CRNH d’Auvergne, Clermont-Ferrand, France, 2LIMOS, UMR CNRS 6158, Laboratory of Computer Science, Clermont-Ferrand University, Aubière Cedex, France, 3Lyon 1 University, CRNH RhôneAlpes, and CENS, F-69310 Lyon, France

Downloaded by [Sylvie Rousset] at 01:19 29 August 2014

1

Abstract The objective of this study was to evaluate the validity of total energy expenditure (TEE) provided by Actiheart® and Armband®. Normal-weight adult volunteers wore both devices either for 17 hours in a calorimetric chamber (CC, n = 49) or for 10 days in free-living conditions (FLC) outside the laboratory (n = 41). The two devices and indirect calorimetry or doubly labelled water, respectively, were used to estimate TEE in the CC group and FLC group. In the CC, the relative value of TEE error was not significant (p > 0.05) for Actiheart® but significantly different from zero for Armband®, showing TEE underestimation (−4.9%, p < 0.0001). However, the mean absolute values of errors were significantly different between Actiheart® and Armband®: 8.6% and 6.7%, respectively (p = 0.05). Armband® was more accurate for estimating TEE during sleeping, rest, recovery periods and sitting–standing. Actiheart® provided better estimation during step and walking. In FLC, no significant error in relative value was detected. Nevertheless, Armband® produced smaller errors in absolute value than Actiheart® (8.6% vs. 12.8%). The distributions of differences were more scattered around the means, suggesting a higher inter-individual variability in TEE estimated by Actiheart® than by Armband®. Our results show that both monitors are appropriate for estimating TEE. Armband® is more effective than Actiheart® at the individual level for daily light-intensity activities. Keywords: Calorimetric chamber, doubly labelled water, total energy expenditure estimation, Actiheart®, Armband®, free-living

Introduction Today, epidemics of overweight and obesity are dramatically increasing worldwide. Such chronic diseases generally result from an energy imbalance, i.e., an excess of energy intake and physical inactivity. In preventive nutrition, the evaluation of energy expenditure and consumption is therefore of utmost importance, especially in epidemiological studies. Evaluating variations in free-living energy expenditure during the day and on a day-to-day basis is also of major interest in clinical trials as well as for individual use. Total energy expenditure (TEE) may be calculated from physical activity questionnaires that generally lack precision because of misestimation by volunteers, or via specific monitors. Several devices are available today for research purposes. Their principle is

based on either accelerometry (Actigraph®, RT3®, ActiReg®; Arvidsson, Slinde, & Hulthen, 2009; Lyden, Kozey, Staudenmeyer, & Freedson, 2011; Rothney, Brychta, Meade, Chen, & Buchowski, 2010), heart rate (HR) and accelerometry (Actiheart®; Brage et al., 2004), or accelerometry, temperature, heat flux and impedance (Armband®; St-Onge, Mignault, Allison, & Rabasa-Lhoret, 2007). It is important to validate existing devices against criterion methods and against each other so that researchers can make informed decisions about their choice of monitor. The present study therefore aimed to investigate the validity of two portable monitoring devices – Actiheart® and Armband® – in normal-weight subjects and in both controlled (calorimetric chamber, CC) and free-living conditions (FLC). To the best of our knowledge, such a comparison between

Correspondence: Sylvie Rousset, INRA, Human Nutrition Unit UMR1019, CRNH d’Auvergne, F-63000 Clermont-Ferrand, France. E-mail: [email protected] © 2014 European College of Sport Science

2

S. Rousset et al.

TEE estimated by the devices and the results given by two reference methods, indirect calorimetry (IC) and doubly labelled water (DLW) has never been carried out. In the end, the results of this study will allow us to determine the strengths and weaknesses of Actiheart® and Armband® for estimating the intensity levels of scheduled activities and TEE in both CC and FLC.

metabolic task (MET). The MET values were calculated from paEE and rEE. Thus, TEE was calculated as: TEE ¼ rEE þ paEE þ thermogenesis

Physical activity intensity was calculated as (Schutz, Weinsier, & Hunter, 2001): MET ¼

Methods


Participants Two independent groups of participants were recruited in Clermont-Ferrand (France) through advertising in the local newspapers. The first one was composed of 49 normal-weight volunteers who were confined in a CC. For the second group, an independent sample of 41 normal-weight volunteers participated in a 10-day test in FLC. Before participating in the study, all volunteers signed an informed consent form. The protocol was submitted to and approved by the French Ethics Committee for the Protection of Human Subjects (Sud-Est 6). This protocol was registered under the reference NCT012095572 in the clinical trials system. Participants were free of disease and medication known to alter energy metabolism and were nonsmokers.

Anthropometric measurements Body weight was measured using a scale (Seca, Hamburg, Germany), with the participants wearing light clothing and no shoes. Height was measured using a height stadiometer. Fat-free mass (FFM) and fat mass (FM) were determined by DEXA (dual energy X-ray absorptiometry, QDR-4500A, Hologic Inc., Waltham, MA, USA). ®

®

Description of Actiheart and Armband devices and of their use Both devices were worn by the volunteers in the CC and FLC. The Actiheart® unit (CamNtech, Cambridge, UK) is placed on the breast at horizontal level. We used long-term recording with a 1-minute epoch. Actiheart® provides two outputs: activity expressed as counts per minute, and HR as beats per minute. On the basis of these two outputs and sleeping HR, the physical activity energy expenditure (paEE) calculation was done using the group calibration branched model (Brage et al., 2007). Rest energy expenditure (rEE) was estimated from the Schofield equation, taking gender and weight into account (Schofield, 1985). Thermogenesis was evaluated as 10% of TEE. The physical activity intensities were expressed in

ð1Þ

rEE þ paEE 0.9 TEE ¼ rEE rEE

ð2Þ

The SenseWear Pro-3 Armband® device (version 6.0, BodyMedia, Pittsburgh, PA, USA) is worn on the right arm triceps. Armband® measures biaxial accelerometry, body temperature, heat flow and impedance. The maker of the SenseWear Armband® device did not provide information about its prediction algorithm. Armband® gives TEE values minuteby-minute with no detail about paEE or rEE. In the CC, a rest period after sleeping (7:00–8:00 am, Table I) made it possible to evaluate rEE. In FLC, the rest period was detected by two outputs produced by Armband®: lying position and awakened state. For each volunteer in FLC, we looked for a 30-minute rest period after sleeping, responding to these two criteria. The MET values were calculated in the same way as for Actiheart® (equation 2). Physical activities were then categorised into light([0.9–3]), moderate- ([3–6]) or vigorous- ([6–9]) MET intensity levels, minute-by-minute, from TEE and rEE evaluated by Actiheart, and Armband in both CC and FLC. MET values were also calculated from IC measures in CC. Finally, the percentages of time at each intensity level were calculated over the whole measurement period.

Indirect calorimetry protocol in controlled conditions The volunteers of the first group were confined in a CC. They arrived at the laboratory at 5:00 pm and measurements began at 00:00 am until 5:00 pm the next day. Thus, the first measurements were made during sleep. Volunteers were awake as of 7:00 am and measurements concerned the waking and active period. The laboratory was equipped with two CC (10 m2), each with a toilet and a sink, a desk, a chair, a bed, a treadmill, a phone, a television, a DVD player and a stereo system. The air extracted from each chamber was blown through a flow meter that supplies an electrical signal proportional to the air mass flowing through it. The differences in oxygen and carbon dioxide contents between air entering and leaving the chamber were determined with differential gas analysers: Oxymat 6 (Siemens), scale: 21–20%, accuracy: ± 0.5% of the scale (i.e., 0.005%) for oxygen; and Ultimat 6 (Siemens), scale:

Comparison of total energy expenditure assessed

3

Table I. Activities schedule in the calorimetric chamber


Activity

Chronological order

Starting time

Duration (min)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

00:00 07:00 09:15 09:30 09:35 09:50 10:05 10:30 10:50 11:00 11:20 12:10 14:00 14:20 14:50 15:10 15:40 16:00 16:30

360 60 15 5 15 15 25 20 10 20 25 20 20 30 20 30 20 35 20

Sleep Rest Writing Sitting–standing After sitting–standing: Recovery 1 Slow walking After slow walking: Recovery 2 Walking at 3 km.h−1 After walking at 3 km.h−1: Recovery 3 Walking at 4 km.h−1 After walking at 4 km.h−1: Recovery 4 Lunch Walking at 5 km.h−1 After walking at 5 km.h−1: Recovery 5 Aerobic step After aerobic step: Recovery 6 Walking at 6 km.h−1 After walking at 6 km.h−1: Recovery 7 House work

0–1%, accuracy: ± 0.5% of the scale (i.e., 0.005%) for carbon dioxide. One file containing minuteby-minute measurements of chamber temperature, relative humidity, O2 and CO2 concentrations, atmospheric pressure and air flow were loaded for each volunteer. Volunteers successively carried out 12 activities (sleep, rest, writing, slow walking, walking at 3, 4, 5 and 6 km.h−1, aerobic step, standing–sitting, lunch and housework) and 7 recovery periods (after standing–sitting: r1; the five walking speeds: r2, r3, r4, r5 and r7; and aerobic step: r6), in chronological order (Table I). Volunteers used a treadmill to walk at different speeds, whereas slow walking was performed by walking around the chamber. Participants were informed by phone of the beginning and the end of each scheduled activity. An activity schedule file with the accurate starting time and duration of each activity was completed and loaded by us for each volunteer.

Doubly labelled water protocol in free-living conditions The TEE of 41 normal-weight volunteers of the second group was calculated using the DLW technique: the two tracers (18O and 2H) are distributed in body water and deuterium is eliminated only as water, whereas 18O is eliminated as water and carbon dioxide so that its rate of excretion is higher than that of deuterium. The difference between the two elimination rates is therefore a measurement of carbon dioxide production over the period of observation.

Volunteers arrived at the laboratory in a postabsorptive state at 7:30 am. They were given a single oral dose of 10% 18O H218O mixed with 99.9% 2 H 2H2O (Eurisotop, St. Aubin, France), and each subject received 0.07 g/kg body weight of 2H2O and 1.5 g/kg BW of H218O, followed by a water rinse. Urine samples were collected at baseline pre-dose, 3 h and 4 h post-dose, and on days 5 and 10 postdose. All samples were stored at −20°C until isotopic analysis. Urine samples were purified and analysed as previously described (Sauvinet et al., 2011). Aliquots of 0.1 µL of the treated samples were injected into a high-temperature conversion elemental analyser coupled with a Delta Advantage isotope ratio mass spectrometer (IRMS) via a Conflow IV Interface (all from Thermo Scientific, Bremen, Germany). Water from the samples was reduced to H2 and CO gas in a glassy carbon reactor heated to 1420°C, then separated on a gas chromatography (GC) column at 90°C, before sequential analysis of deuterium and 18-oxygen isotopic abundances in the IRMS. Three water standards from Iso-Analytical Ltd. (Crewe, UK) were used to establish the calibration curve for normalisation of the values. The 2H- and 18Oenrichments were expressed in δ‰ vs. VSMOW (Vienna Standard Mean Ocean Water). The production of CO2 was calculated according to Schoeller’s equation (Schoeller et al., 1986), and the TEE was derived using Weir’s equation (Weir, 1949) and assuming a respiratory quotient of 0.85. As for the volunteers in the CC, an activity schedule file was loaded by us for each volunteer

4

S. Rousset et al.

studied in FLC. These files contained only one activity starting at the time when they drank DLW and finishing when they collected the last urine sample on day 10 post-dose.


Data collection and preparation For the whole experiment, we loaded a volunteer file with the characteristics of volunteers: volunteer code, birth date, gender, experimental condition (CC or FLC), weight, height, FFM and FM. This file, all the activity schedule files, plus the data files from Actiheart®, Armband® and the CC are in Excel 2003 format. Finder2E software, developed in-house and downloaded at https://activcollector.clermont.inra.fr//Finder2E/Finder2E_v4.zip, was used for preparing the data (Rousset et al., 2011). First, all the files mentioned above were uploaded by Finder2E. Files from Actiheart®, Armband® and the CC were then standardised so that they all started and ended at the same time. Finder2E then replaced the missing data corresponding to the time when the devices were not worn by an estimated TEE values. Non-wearing time represented 2% of the time in FLC and 0% in the CC. The estimated values were automatically calculated as the mean of 10 values: 5 before and 5 after the non-wearing period. After the data preparation, Finder2E computed the TEE in kcal, minute-by-minute, calculated from O2 and CO2 concentrations, air flow, temperature and atmospheric pressure measured in a CC, according to Weir’s equation (Weir, 1949), and taking the possible drifts of gas analysers and flow meters into account. Technical details of the CC were previously described (Bitar, Vermorel, Fellmann, & Coudert, 1995). Finder2E offered the possibility of computing means of TEE (kcal.min−1) during each activity period using the volunteer’s activity schedule and his/her Armband®, Actiheart® and CC files. Finally, for each activity, Finder2E calculated the relative values of errors between the TEE estimated by Actiheart® or Armband® and the TEE of the reference (IC) according to the equation: ðTEEDevice TEERef Þ 100 ð3Þ ErrorDevice ð%Þ ¼ TEERef where TEEDevice is the TEE estimated by the device (Actiheart® or Armband®), and TEERef is the TEE measured by IC (reference). The absolute value of error is positive or null. These operations require a few seconds and give the user the possibility to estimate the accuracy of each TEE predicted by a monitor in comparison to Weir’s equation.

For the volunteers in FLC, only the sum of the predictive values of TEE for both monitors over the total duration (10 days) was compared with the value of TEE calculated by the DLW technique. Statistical analysis All t-tests were performed using SAS 9 software. For each gender, the characteristics of the two groups (age, body mass index [BMI], FFM, FM) were compared by t-tests. The percentage of time estimated in an activity category (light-, moderate- or vigorous-intensity) by each device was compared to that determined by the reference method in the CC by performing paired t-tests. For each activity, the error of TEE estimation (%) was expressed either in relative (equation 3) or in absolute value (positive value of error). First, four t-tests (two devices × two conditions) were carried out to determine if the mean of relative error values calculated over the whole measurement period was significantly different from zero. In order to determine if the two devices provided significantly different errors over the activity periods, we carried out paired t-tests per activity in the CC and over the whole period in FLC. Statistical significance was set at p < 0.05. Agreement between portable monitoring devices, TEE estimations and reference measurements was evaluated by Bland–Altman plots (Bland & Altman, 1986). For each experimental condition, two plots were drawn showing the mean difference between TEE values estimated by one device and calculated by the reference method against the mean of the two methods. The bias is estimated by the mean difference (M) and the standard deviation (s). Statistically, 95% of the differences will lie between M ± 2s (agreement limits). The validity of Actiheart® and Armband® was evaluated in each experimental condition by comparing agreement level between each device and the reference method. Results All subjects were of normal weight (18.5 < BMI < 25.0 kg/m2) and middle-aged (Table II). For men, there was no difference in age, BMI and percentage of FM between the two groups (CC and FLC). For women, FFM was significantly higher in FLC than in CC; other parameters were similar between the two groups. Intensities of activity carried out in the CC and FLC Light-intensity activities ([0.9–3] MET) took up most of the time: 93% and 90% according to

5

Comparison of total energy expenditure assessed Table II. Characteristics of the participants (mean ± s) Men Experimental conditions N Age (years) BMI (kg.m−2) Fat-free mass (kg) Fat mass (%)

CC 23 44.2 ± 23.6 ± 60.0 ± 17.2 ±

Women FLC

4.6 1.4 6.1 4.3

21 41.7 ± 23.8 ± 63.3 ± 16.2 ±

6.9 1.5 5.2 3.7

CC 26 46.2 ± 21.6 ± 43.2 ± 24.6 ±

FLC

5.6 1.8 4.3 4.4

42.4 22.4 45.9 25.4

20 ± 8.0 ± 2.2 ± 3.5* ± 5.3


*p < 0.05 for the comparison between the CC and FLC. CC, calorimetric chamber; FLC, free-living conditions.

Armband®, and 96% and 93% according to Actiheart®, in the CC and FLC, respectively. Moderateintensity activities ([3–6] MET) accounted for 7% and 9% of the time according to Armband®, and 4% and 6% according to Actiheart®. Otherwise, there were few vigorous-intensity activities ([6–9] MET): less than 1% of time. Moreover, the percentage of time determinated by Actiheart and Armband dedicated to the moderate-intensity activity was slightly higher in free-living than in controlled conditions. It was possible only in the CC to compare the duration of light-, moderate- or vigorous-intensity activities determined by the reference method and estimated by the devices. The paired t-tests showed a significant mean difference between Actiheart® and IC for light-intensity activities of 3.6% (overestimation), p < 0.0001; for moderate-intensity activities of −3.6% (underestimation), p < 0.001; and for vigorous-intensity activities of −0.04%, p = 0.70. No difference between Armband® and IC was noted for estimating the time of light- and moderate-intensity activities (0.6%, p = 0.13; −0.4%, p = 0.34). Conversely, there was a significant difference for the duration of vigorous-intensity activities estimated by Armband®: −0.2%, p = 0.02 (underestimation).

Errors produced by both devices in the CC In the CC, the mean relative value of error for estimating TEE was not significantly different from zero in the case of Actiheart®, −1.8 ± 10.6% (p = 0.23), but significantly different from zero for Armband® − 4.9 ± 6.7% (p < 0.001). However, the mean difference between the device errors in absolute value was significantly different from zero (p = 0.05). Thus, the error was higher for Actiheart® (8.6 ± 6.3%) compared to Armband® (6.7 ± 5.1%). The Actiheart® and Armband® absolute errors were similar in women for both monitors, 7.7 vs. 6.8% (p = 0.58), while Actiheart® errors were higher in men compared to Armband®, 9.7 vs. 6.5% (p = 0.03). Analysis by activity showed that Actiheart® better predicted TEE than Armband during walking at 4 km.h−1, exercising with a stepper and standing recovery after sitting–standing. In contrast, the prediction of TEE by the Armband® during postabsorptive rest, sleeping and the rests after slow walking and stepper activity was better than that of Actiheart® (Figure 1). No significant difference was observed for the other activities (walking at 3, 5 and

Figure 1. Signiﬁcant differences in errors between Actiheart® (AH) and Armband® (AB) for six sedentary behaviours and two activities (mean ± s) performed in calorimetric chambers.


6

S. Rousset et al.

Figure 2. Bland and Altman plots (mean difference ± 2 s) of the agreement level between total energy expenditure (TEE) measured in calorimetric chambers (CC) and estimated using (a) Actiheart® (AH) and (b) Armband® (AB), and between TEE measured over 10 days in free-living conditions (FLC) by the doubly labelled water (DLW) and estimated using (c) Actiheart® (AH) and (d) Armband® (AB). : Mean difference (M) : Mean difference plus or less two standard deviations (M ± 2s)

6 km.h−1, other recovery periods, writing and standing–sitting; Figure 1). In the CC, the bias M calculated according to Bland and Altman’s method was small for both devices: −0.08 kcal.min−1 for Armband® and −0.03 kcal.min−1 for Actiheart® (Figure 2a and b). However, the upper and lower limits of agreement were closer for Armband® (−0.32 to 0.15 kcal.min−1; Figure 2b) compared to Actiheart® (−0.39 to 0.32 kcal.min−1; Figure 2a). Thus, lower and upper limits were −19% and 9% for Armband, and −24% and 22% for Actiheart. Except for two values for Actiheart®, all differences between portable devices and the reference were within the limits (M ± 2s).

Errors produced by both devices in FLC In FLC, the relative mean errors produced by both monitors were not significantly different from zero: −3.4 ± 15.5%, and −0.7 ± 10.3%, for Actiheart®

and Armband®, respectively. Conversely, the mean difference between the device errors in absolute value was significantly higher for Actiheart® (12.8 ± 9.1%) compared to the SenseWear Armband® (8.6 ± 5.5%; t = 2.78, p = 0.008). The TEE errors averaged 15.5% and 8.8% in men, and 10.0% and 8.5% in women, for Actiheart® and Armband®, respectively. The mean difference between device errors was significant for men (p = 0.005) but not for women. The mean bias M between DLW and the two monitors was low: −0.07 and −0.03 kcal.min−1 for Actiheart® (Figure 2c) and Armband® (Figure 2d), respectively. However, the agreement limits were larger for Actiheart® (−0.69 to 0.54 kcal.min−1) than for Armband (−0.43 to 0.36 kcal.min−1). For two subjects, the TEE calculated by DLW was 1.92 and 2.66 kcal.min−1 and estimated to be as high as 2.84 and 3.30 kcal.min−1 by Actiheart®. In the two cases, Actiheart® dramatically overestimated the TEE

Comparison of total energy expenditure assessed (Figure 2c). Thus, lower and upper limits were −23% and 19% for Armband, and −36% and 29% for Actiheart.


Discussion In the present study, the performances of Actiheart® and Armband® for TEE estimation were compared with the results given by two reference methods used in controlled and FLC, the IC and the DLW method, respectively. The CC tool makes it possible to evaluate TEE during short activity periods in controlled conditions and the DLW method to measure the TEE in FLC for 10 days. Numerous studies focus on controlled conditions and, as stressed by Koehler et al. (2011) and Rothney et al. (2010), only very few publications encompass the two conditions for measuring TEE in the CC and FLC. In these last two studies, the same volunteers were studied in both the CC and the FLC. However, this scientific approach is less powerful than the study of the two conditions with participants belonging to independent groups. In addition, these two studies validated Actigraph® and Armband®, respectively, with smaller groups than ours (22 vs. 49 and 14 vs. 41 volunteers). Thus, the majority of the studies developed TEE prediction models under controlled conditions (using a CC or a face mask; Arvidsson, Fitch, Hudes, & Fleming, 2011; Arvidsson et al., 2009; Brage et al., 2004; Colbert, Matthews, Havighurst, Kim, & Schoeller, 2011; Dorwny, Cho, Akohoue, Chen, & Buchowski, 2008; Howe, Staudenmayer, & Freedson, 2009; King, Torres, Potter, Brooks, & Coleman, 2004; Lyden et al., 2011; Rothney, Neumann, Beziat, & Chen, 2007; I. Zakeri, Adolph, Puyau, Vohra, & Butte, 2008; I. F. Zakeri, Adolph, Puyau, Vohra, & Butte, 2010). Before this study, to the best of our knowledge, Actiheart® was compared only twice to the results given by the DLW technique in FLC (Assah et al., 2011; Villars et al., 2012). In both studies, the mean bias of physical activity TEE was −9.1 kJ.kg−1. day−1 (Assah et al., 2011) and −5.0 kJ.kg−1.day−1 (Villars et al., 2012). These results are close to those found in our study, i.e., −6.8 kJ.kg−1.day−1. Armband® was more often tested than Actiheart® in FLC, with good estimation of TEE. In the majority of cases, less than 10% of error was observed. However, activity-related EE was sometimes poorly estimated. Several studies showed that Armband® underestimated EE in the high value range of EE related to high-intensity physical activity (Arvidsson et al., 2009, 2011; Fruin & Rankin, 2004; Johannsen et al., 2010; Koehler et al., 2011; Mackey et al., 2011; St-Onge et al., 2007). Overall, it is difficult to compare the performance of monitors or prediction models because of

7

the different methodologies and criteria used by authors, i.e., controlled vs. FLC, various populations, activities of various intensity levels, errors expressed in percentage, absolute values of errors in percentage, and either in kilojoules or in kilocalories per minute or day, per kilograms or not (Ryan & Gormley, 2013). This fact justified our experiment and the present study made it possible to compare Actiheart® and Armband® performances in the same conditions. In the present study, the TEE estimated by the two devices was compared for rest, physical activities and recovery periods in the CC. The main results of this study showed that Actiheart® was more effective than Armband® in predicting TEE during walking at 4 km.h−1, exercising with a stepper and standing recovery after sitting–standing. The reason is probably due to the fact that there are obvious HR differences between such activities and rest. Moreover, mean TEE estimated by Actiheart® for the group in the CC was close to the mean TEE of the reference. Since Armband® is more efficient for estimating energy expenditure of light-intensity activities, it was not surprising that the prediction of TEE during post-absorptive rest, sleeping and the rests after slow walking and stepper activity was better estimated by this portable device. That may also explain the overall TEE underestimation of Armband® in controlled conditions during which several moderate-intensity activities were performed. The agreement limits showed that Armband could underestimate (−19%) but only slightly overestimate TEE (+9%), whereas Actiheart could both underestimate (−24%) and overestimate TEE (19%). Thus, limits were more scattered for Actiheart than Armband. In other words, a given individual Actiheart® estimation could be farther from the reference than an Armband® estimation. The intensity of the activities carried out in the CC was more precisely predicted by Armband® than Actiheart® because most of the activities were either light or moderate. Conversely, Actiheart® was better for estimating activities of vigorous-intensity such as walking at 6 km.h−1 or aerobic step. The studies of IC in controlled conditions, such as in the CCs, have the great merit to measure TEE for relatively homogeneous activity periods but do not reflect actual life conditions. The study in FLC was complementary to the previous ones because life is composed of spontaneous, short or long, discontinuous activities. The DLW technique gives a result of TEE averaged over 10 days. Bland and Altman’s method applied in FLC showed that the deviations around the mean of differences were larger in FLC than in the CC. It may be more difficult to estimate TEE of spontaneous, diverse and brief activities (less than 1 minute) in FLC than that of stereotyped and


8

S. Rousset et al.

long duration activities in the CC. Moreover a smaller range between the lower and upper limits was found for Armband® (−24% to 19%) than for Actiheart® (−36% to 29%). Note that Armband® contains more sensors to measure variables linked to energy expenditure (biaxial accelerometers, skin and proximal temperatures, impedance and heat flux) than Actiheart® (uniaxial accelerometer and HR). In conclusion, most of the time in FLC is spent on light-intensity activities, a small part is spent on moderate-intensity activities, and a very short time is spent on vigorous-intensity activities. In these conditions, bias was larger for Actiheart® than for Armband®. Armband® is thus a more valid portable monitoring device within the context of daily physical activities. More generally, Armband® is best suited for light-intensity activities of daily life and Actiheart® for sportsmen carrying out an individual calibration. Nevertheless, there are still great opportunities available to develop other simpler and cheaper monitors adapted to daily activities and dedicated to the general public, using novel technologies. Acknowledgement We would like to thank V. Sauvinet, M. Duclos and A. Lebert for their helpful suggestions, A. Gerard for her technical assistance, and all of the volunteers who participated in this study. Funding No funding outside INRA was received for this work. References Arvidsson, D., Fitch, M., Hudes, M. L., & Fleming, S. E. (2011). Accuracy of multisensor activity monitors in normal versus high BMI African American children. Journal of Physical Activity & Health, 8, 1124–1134. Arvidsson, D., Slinde, F., & Hulthen, L. (2009). Free-living energy expenditure in children using multi-sensor activity monitors. Clinical Nutrition, 28, 305–312. doi:10.1016/j.clnu.2009.03.006 Assah, F. K., Ekelund, U., Brage, S., Wright, A., Mbanya, J. C., & Wareham, N. J. (2011). Accuracy and validity of a combined heart rate and motion sensor for the measurement of free-living physical activity energy expenditure in adults in Cameroon. International Journal of Epidemiology, 40, 112–120. doi:10.1093/ ije/dyq098 Bitar, A., Vermorel, M., Fellmann, N., & Coudert, J. (1995). Twenty-four-hour energy expenditure and its components in prepubertal children as determined by whole-body indirect calorimetry and compared with young adults. American Journal of Clinical Nutrition, 62, 308–315. Bland, J. M., & Altman, D. G. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. The Lancet, 327, 307–310. doi:10.1016/S0140-6736(86)90837-8 Brage, S., Brage, N., Franks, P. W., Ekelund, U., Wong, M.-Y., Andersen, L. B., … Wareham, N. J. (2004). Branched equation modeling of simultaneous accelerometry and heart rate monitoring imp-roves estimate of directly measured physical activity

energy expenditure. Journal of Applied Physiology, 96, 343–351. doi:10.1152/japplphysiol.00703.2003 Brage, S., Ekelund, U., Brage, N., Hennings, M. A., Froberg, K., Franks, P. W., & Wareham, N. J. (2007). Hierarchy of individual calibration levels for heart rate and accelerometry to measure physical activity. Journal of Applied Physiology, 103, 682–692. doi:10.1152/japplphysiol.00092.2006 Colbert, L. H., Matthews, C. E., Havighurst, T. C., Kim, K., & Schoeller, D. A. (2011). Comparative validity of physical activity measures in older adults. Medicine and Science in Sports and Exercise, 43, 867–876. doi:10.1249/MSS.0b013e31 81fc7162 Dorwny, C. A., Cho, L., Akohoue, S. A., Chen, K. Y., & Buchowski, M. S. (2008). Validity of a multisensor armband in estimating 24-h energy expenditure in children. Medicine and Science in Sports and Exercise, 40, 699–706. doi:10.1249/ MSS.0b013e318161ea8f Fruin, M. L., & Rankin, J. W. (2004). Validity of a multi-sensor armband in estimating rest and exercise energy expenditure. Medicine and Science in Sports and Exercise, 36, 1063–1069. doi:10.1249/01.MSS.0000128144.91337.38 Howe, C. A., Staudenmayer, J. W., & Freedson, P. S. (2009). Accelerometer prediction of energy expenditure: Vector magnitude versus vertical axis. Medicine and Science in Sports and Exercise, 41, 2199–2206. doi:10.1249/MSS.0b013e3181aa3a0e Johannsen, D. L., Calabro, M. A., Stewart, J., Franke, W., Rood, J. C., & Welk, G. J. (2010). Accuracy of armband monitors for measuring daily energy expenditure in healthy adults. Medicine and Science in Sports and Exercise, 42, 2134–2140. doi:10.1249/ MSS.0b013e3181e0b3ff King, G. A., Torres, N., Potter, C., Brooks, T. J., & Coleman, K. J. (2004). Comparison of activity monitors to estimate energy cost of treadmill exercise. Medicine and Science in Sports and Exercise, 36, 1244–1251. doi:10.1249/01.MSS.0000132379.09364.F8 Koehler, K., Braun, H., De Marees, M., Fusch, G., Fusch, C., & Schaenzer, W. (2011). Assessing energy expenditure in male endurance athletes: Validity of the sensewear armband. Medicine and Science in Sports and Exercise, 43, 1328–1333. doi:10.1249/MSS.0b013e31820750f5 Lyden, K., Kozey, S. L., Staudenmeyer, J. W., & Freedson, P. S. (2011). A comprehensive evaluation of commonly used accelerometer energy expenditure and met prediction equations. European Journal of Applied Physiology, 111, 187–201. doi:10.1007/ s00421-010-1639-8 Mackey, D. C., Manini, T. M., Schoeller, D. A., Koster, A., Glynn, N. W., Goodpaster, B. H., … Cummings, S. R. (2011). Validation of an armband to measure daily energy expenditure in older adults. Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 66, 1108–1113. doi:10.1093/gerona/glr101 Rothney, M. P., Brychta, R. J., Meade, N. N., Chen, K. Y., & Buchowski, M. S. (2010). Validation of the ActiGraph tworegression model for predicting energy expenditure. Medicine and Science in Sports and Exercise, 42, 1785–1792. doi:10.1249/ MSS.0b013e3181d5a984 Rothney, M. P., Neumann, M., Beziat, A., & Chen, K. Y. (2007). An artificial neural network model of energy expenditure using nonintegrated acceleration signals. Journal of Applied Physiology, 103, 1419–1427. doi:10.1152/japplphysiol.00429.2007 Rousset, S., Lasnes, M., Spriet, C., Walter, A., Morio, B., & Lacomme, P. (2011). Finder2e: A software to characterise activity and energy expenditure over short time-intervals in free living volunteers. Proceedings of the Second International Conference on Ambulatory Monitoring of Physical Activity and Movement, Glasgow. Ryan, J., & Gormley, J. (2013). Measurement of energy expenditure by activity monitors: A review of the literature. Physical Therapy Reviews, 18, 239–262. doi:10.1179/1743288X13Y.0000000063

Comparison of total energy expenditure assessed


Sauvinet, V., Gabert, L., Alligier, M., Normand, S., Roth, H., Laville, M., & Desage, M. (2011). Comparison of hightemperature conversion and equilibration methods for the determination of d(31)-palmitic acid oxidation in man using continuous-flow isotope ratio mass spectrometry. Rapid Communications in Mass Spectrometry, 25, 2749–2759. doi:10.1002/ rcm.5173 Schoeller, D. A., Ravussin, E., Schutz, Y., Acheson, K. J., Baertschi, P., & Jequier, E. (1986). Energy-expenditure by doubly labeled water – Validation in humans and proposed calculation. American Journal of Physiology, 250, R823–R830. Schofield, W. N. (1985). Predicting basal metabolic rate, new standards and review of previous work. Human Nutrition Clinical Nutrition, 39 (Suppl. 1), 5–41. Schutz, Y., Weinsier, R. L., & Hunter, G. R. (2001). Assessment of free-living physical activity in humans: An overview of currently available and proposed new measures. Obesity Research, 9, 368–379. doi:10.1038/oby.2001.48 St-Onge, M., Mignault, D., Allison, D. B., & Rabasa-Lhoret, R. (2007). Evaluation of a portable device to measure daily energy

9

expenditure in free-living adults. American Journal of Clinical Nutrition, 85, 742–749. Villars, C., Bergouignan, A., Dugas, J., Antoun, E., Schoeller, D. A., Roth, H., … Simon, C. (2012). Validity of combining heart rate and uniaxial acceleration to measure free-living physical activity energy expenditure in young men. Journal of Applied Physiology, 113, 1763–1771. doi:10.1152/japplphysiol.01413.2011 Weir, J. B. D. B. (1949). New methods for calculating metabolic rate with special reference to protein metabolism. The Journal of Physiology, 109, 1–9. Zakeri, I., Adolph, A. L., Puyau, M. R., Vohra, F. A., & Butte, N. F. (2008). Application of cross-sectional time series modeling for the prediction of energy expenditure from heart rate and accelerometry. Journal of Applied Physiology, 104, 1665–1673. doi:10.1152/ japplphysiol.01163.2007 Zakeri, I. F., Adolph, A. L., Puyau, M. R., Vohra, F. A., & Butte, N. F. (2010). Multivariate adaptive regression splines models for the prediction of energy expenditure in children and adolescents. Journal of Applied Physiology, 108(1), 128–136. doi:10.1152/ japplphysiol.00729.2009