Gesundheit & Bewegung
ORIGINALIA
ACCELEROMETRY AND ENERGY EXPENDITURE IN OBESITY

Accelerometry and Energy Expenditure in
Obese Adults and Normal Weight Controls

Accelerometrie und Energieverbrauch bei adipösen Erwachsenen
und normalgewichtigen Kontrollpersonen

ZUSAMMENFASSUNG

Fragestellung: Die Studie evaluiert ein neues triaxiales Accelerometer zur Vorhersage des Energieverbrauchs, gemessen als VO2/kg, bei adipösen und normalgewichtigen Erwachsenen während Alltagsaktivitäten. Methoden: 37 adipöse Erwachsene (Body Mass Index (BMI) 37±5,4) und 17 normalgewichtige Kontrollpersonen (BMI 23 ±1,8)) führten acht verschiedene Aktivitäten während 5 bis 8 Minuten durch. Dabei trugen sie am rechten Oberschenkel (unterhalb Hüftgelenk) ein triaxiales Accelerometer zur Messung der bewegungsinduzierten Accelerometereinheiten (AE) (DiaTrace System). Zeitgleich erfolgte die Messung der Sauerstoffaufnahme (VO2) und Kohlendioxidabgabe (VCO2) mittels eines tragbaren Spiroergometrie-Systems. Die Beziehung zwischen AE und VO2/kg wurde mittels Spline-Regression und linearen gemischten Modellen analysiert. Ergebnisse: Für alle Aktivitäten wiesen die Adipösen eine niedrigere VO2/kg auf als die Normalgewichtigen. Für beide Gruppen zeigte sich eine lineare Beziehung zwischen AE und VO2/kg für AE-Werte zwischen 0 und 300 Einheiten/min mit einer Zunahme des VO2/kg von 3,7ml/min (95%-Vertrauensintervall (VI) 3,4-4,1) bei adipösen und 3,9ml/min (95%-VI 3,4-4,3) pro 100 AE bei normalgewichtigen Personen. Die lineare Modellierung des gesamten AE-Wertebereiches zeigt grosse Vorhersage-Intervalle von ±6,3ml/min bei Adipösen und ±7,3ml/min VO2/kg bei Normalgewichtigen. Fazit: Es ist bei Adipösen wie Normalgewichtigen nur begrenzt möglich, mittels AE den Energieverbrauch, definiert als VO2/kg, für Alltagsaktivitäten von variierenden Intensitäten und Formen der Muskelarbeit vorauszusagen.

Schlüsselwörter: Accelerometrie, Alltagsaktivitäten, Energieverbrauch, Adipositas.

SUMMARY

Objective: To evaluate a new triaxial accelerometer device for prediction of energy expenditure, measured as VO2/kg, in obese adults and normal-weight controls during activities of daily life. Subjects and methods: Thirty-seven obese adults (Body Mass Index (BMI) 37±5.4) and seventeen controls (BMI 23±1.8) performed eight activities for 5 to 8 minutes while wearing a triaxial accelerometer on the right thigh. Simultaneously, VO2 and VCO2 were measured using a portable metabolic system. The relationship between accelerometer counts (AC) and VO2/kg was analysed using spline regression and linear mixed-effects models. Results: For all activities, VO2/kg was significantly lower in obese participants than in normalweight controls. A linear relationship between AC and VO2/kg existed only within accelerometer values from 0 to 300 counts/min, with an increase of 3.7 (95%-confidence interval (CI) 3.4-4.1) and 3.9 ml/min (95%-CI 3.4-4.3) per increase of 100 counts/min in obese and normal-weight adults, respectively. Linear modelling of the whole range yields wide prediction intervals for VO2/kg of ±6.3 and ±7.3 ml/min in both groups. Conclusion: In obese and normal-weight adults, the use of AC for predicting energy expenditure, defined as VO2/kg, from a broad range of physical activities, characterized by varying intensities and types of muscle work, is limited.

Key Words: Accelerometer, physical activity, energy expenditure, obesity.

INTRODUCTION

Accelerometers provide a measure of physical activity and are increasingly being used in research and in the population (20). Previous studies calibrated records of accelerometer counts (AC) in a given time interval against measured energy expenditure (VO2; METs [metabolic equivalent]) during a variety of activities (15, 21). Regression equations to estimate MET values from AC were created and categories defining sedentary, easy, moderate and vigorous activity were generated (8, 19). However, despite the widespread use of accelerometers, the relation between AC and exercise energy expenditure in obese adults has rarely been assessed in obese adults (6, 14). Recently a triaxial accelerometer has been developed for integration into a mobile phone. The phone is equipped with specialized software (DiaTrace System (2)) that allows the recording of acceleration data and offers mobile assistance in supporting exercise behaviour in daily life.
This study sought to evaluate a new triaxial accelerometer device for prediction of energy expenditure, measured as VO2/kg, by accelerometer counts (AC) in obese adults and normal weight controls during eight activities of daily life.

SUBJECTS AND METHODS

The Fraunhofer-Institut für Graphische Datenverarbeitung, Rostock, Germany provided the recording software (DiaTrace System) for the assessment of triaxial accelerometer data (2).

Study population
Thirty-seven obese adults were recruited from a 3-year outpatient weight loss program of the University of Bern. Seventeen normal weight individuals served as control group (Table 1). Inclusion criteria were: Age 18 to 63 years; Body mass index (BMI) 30 to 50 for the obesity group and BMI 20 to 25 for the control group. Exclusion criteria were: Diagnosed coronary artery disease with exercise induced myocardial ischemia; diagnosed left ventricular dysfunction, heart failure and symptomatic peripheral occlusive vascular disease; reduced haemoglobin level; untreated arterial hypertension; obstructive lung problems (FEV1%VC <70); musculoskeletal disorders that limit physical activities to be assessed; significant atrial and/or ventricular arrhythmia.

Study procedures
The study protocol was approved by the local Ethics Committee of the University of Bern, Switzerland. Prior to participation, obese and normal weight individuals signed an informed consent.

Physical activities
Physical activities to be assessed (Table 2) were performed in a random order. The activities were chosen to provide a variety of intensities from sedentary to moderate-vigorous, and to reflect typical activities of daily life, including “lifestyle” activities, general housework
activities and leisure activities for relaxation. Between activities, a
resting phase of 10 minutes was inserted.

Equipment and parameters assessed
The hardware of the accelerometer features an STMicroelectronics triaxial sensor (ST LIS302DL), a sampling rate of 20Hz, 26bit/g quantization and a measuring range of ±2g (2). The data assessment software (DiaTrace System) provides sensor raw data and in addition a metric for physical activity without the offset of gravity or influence of orientation. The accelerometer measures the physical force at the fixation point for each axis (ax, ay, az), subtracts the gravity influence and computes the physical force in AC by the following equation:

The gravity offset can be achieved by the moving average operation for each axis:

Ttable 1:Characteristics of obese and normal weight participants.

Table 2:Description of physical activities with total exercise time in brackets.

For the sampling rate of 20Hz we defined m=0.01 which allows a dynamic offset compensation. For further processing, all recorded acceleration raw data tuples were stored with a time stamp. Because the acceleration sensor provides only discrete data, the AC were normalized to a second and summarized in an epoch of 3.2 seconds. The AC is SI-conform and measured in m/s3:

The most frequent location of carrying a mobile phone is the front trouser pocket (13). Therefore, participants wore the accelerometer attached to a belt at the right thigh 15cm below the Spina iliaca anterior superior during all activities.
Oxygen uptake and CO2production were measured breath by breath by means of a portable metabolic unit (Metamax; Cortex Company, Leipzig, Germany). The heart rate was measured by a Polar heart rate transmitter unit. Additionally, rating of perceived exertion (RPE; Borg scale 6- 20) was assessed at the end of each activity, except for resting and sitting/watching TV.

Study preparation
Prior to study entry, all participants underwent a ramp treadmill test (Woodway, Weil/Germany) to determine peak VO2. Three hours prior to assessment, participants were not allowed to smoke or eat, except drinking fruit tea and water. One hour before starting the measurements, participants were introduced to the study equipment and procedure. Before starting the assessment, the accelerometer and the metabolic unit were synchronized. Accelerometer data, data on gas exchange and data on heart rate were documented every three seconds. During all study activities, participants were accompanied by two trained staff members, who were responsible for speed and/or intensity control of the exercise.

Data preparation
All data were checked for correct synchronization between AC and VO2/kg measures. To assure a steady-state phase of all measurements during each activity, values collected during the first two minutes and the last 30 seconds of each 5- and 8-minute activity were discarded. To address measurement errors of the heart rate transmitter, heart rate values above 180beats/min (0.2% of all observations) and values below 50beats/min for resting activities (lying; watching TV) or 60beats/min for remaining activities (0.7% of all observations) were discarded.
For biological reasons, there is a time lag between the activity as immediately measured by the accelerometer and the resulting energy expenditure. Therefore, averages of each measurement were calculated for the constant phase of each activity within each patient and then used for the statistical analysis.

Statistical analysis
Baseline characteristics, AC and VO2/kg of obese and normal weight participants were compared using unpaired t-tests. To characterize the relationship between AC and VO2/kg over a range of activities of different intensities, an explorative cubic spline regression with knots at 0, 200, 400, 600, and 1000counts/min was calculated and plotted separately for both groups. Linear mixed-effects regression analyses were performed for both groups accounting for the dependence among repeated measurements within the same subject and controlling for age and BMI, once for an accelerometer range of 0 to 300counts/min and once for the whole range. A scatter plot was drawn with separate linear regression lines and corre-sponding 95%-prediction intervals for each group. The effect of the different types of activities on the relationship between AC and VO2/kg is depicted by a scatter plot with different markers for each type of activity. Moreover, a linear mixed-effects regression was calculated that also accounted for the type of activity. All P-values and 95%-confidence intervals (CI) are two-sided. P-values <0.05 are considered significant. All analyses were done using Stata 11 (Stata Corporation, College Station, Texas).

RESULTS

Energy expenditure in obese compared to normal weight adults
With the exception of two exercises (cycling and push-ups), the respiratory exchange rate (RER) was always within the aerobic range (between 0.80 and 0.87) in both obese and normal weight adults. When cycling, the mean RER was 0.95 in both groups. For push-ups, the mean RER was 0.93 and 0.95 in normal weight and obese adults, respectively. For all exercises, there were no significant differences in RER between groups.
For all physical activities, including watching TV and lying down, the obese individuals demonstrated significantly higher absolute values of VO2l/min. The mean difference was lowest for lying, at 8%, and highest for walking at slope, at 35%. On the contrary, except for walking at slope, obese patients exhibited significantly lower VO2/kg values than normal weight individuals. This is also true for cycling despite the fact that both groups performed the exercise at the same relative work load (64% of peak VO2, Fig. 1a) and the same RPE of “somewhat hard”.
Figure 1:Comparison of measures (mean) of energy expenditure (a, b) and accelerometer activity counts (c) between obese adults (black bars) and normal weight adults (grey bars) for different activities. * denotes a significant difference between the two groups as shown by an unpaired t-test.





For all activities, energy expenditure expressed as kJ/10min was significantly higher in the obese compared to the normal weight group (Fig. 1b) (range of increase between 13% for sitting and lying, and 66% for walking at slope).
AC per minute varied considerably by type of exercise, with no statistically significant difference between the obese and normal weight groups (Fig. 1c).

Relationship between physical activity and energy expenditure
A linear relationship between AC and energy expenditure (VO2/kg) only exists at lower AC up to a threshold of about 300counts/min (Fig. 2). An adjusted mixed-effects linear regression within this range demonstrated a significant relationship between AC and VO2/kg in both the obese and normal weight adults (Table3a). The VO2/kg increased by 3.7ml/min (95%-CI 3.4- 4.1) in obese and 3.9ml/min (95%-CI 3.4- 4.3) in normal weight participants if AC increased by 100 counts/min. Age and BMI had a significant effect on the AC only in obese patients. When modelling the whole AC range linearly (Table3b and Fig. 2), the relationship with VO2/kg remained highly significant in both groups, with narrow confidence intervals around the regression lines (±0.5 and ±0.9ml/min in obese and normalweight participants, respectively), but rather wide prediction intervals (±6.3 and ±7.3ml/min, respectively).
Table 3: Relationship between accelerometer counts (AC) and energy expenditure (VO 2 /kg) for obese and normal weight participants assessed by mixed-effects linear regression. The effects show the increase in VO 2 /kg (ml/min) per increase in the dependent variables (AC, age and BMI).

Figure 2:Scatter plot showing relationship between accelerometer counts and VO 2 /kg for obese (solid circles) and normal weight adults (hollow triangles). Cubic spline regression lines are depicted as grey lines, linear regression lines with 95%-prediction intervals as black lines (obese: solid lines; normal weight adults: dash-dotted lines).
When additionally correcting for the type of activity in the regression analysis, the association between AC and VO2/kg diminished, and the extent of energy expenditure was mostly explained by age, BMI and type of activity for obese subjects and by type of activity for normal weight adults (data not shown). Compared to lying down, all activities except watching TV resulted in an increase in energy expenditure in both groups. Fig. 3 shows the distribution of the different types of ac-tivities in relation to AC and VO2/kg. In both groups, lying and watching TV resulted in low AC and low VO2/kg. Vacuum cleaning resulted in rather low AC but fairly high VO2/kg. Comparing the distributions of slow walking with walking at slope reveals that both activities resulted in similar AC (approximately 450counts/min), but led to very different VO2/kg values (Fig. 3).

DISCUSSION

This study provides new insight into the relationship between accelerometer output and energy expenditure, expressed as VO2/kg, in obese adults and normal weight controls. The assessed physical activities (Table 2) represented typical activities from the daily life of both obese and normal weight adults, and are characterized by dynamic concentric, dynamic-eccentric, and isometric muscle work.

Energy expenditure in obese compared to normal weight adults
For all activities including lying and sitting, obese adults exhibited higher absolute VO2but lower VO2/kg values than adults with normal weight (Fig. 1a). The higher absolute VO2 cannot only be explained by the absolute greater body mass but also by the greater effort of moving heavier legs contributing significantly to the oxy-gen cost as shown for cycling (17). The absolute VO2/kg was 13% to 66% higher in obese than in normal weight adults, corresponding to a difference of 541kJ per hour of activity (Fig. 1b).

Relationship between physical activity and energy expenditure
There exists a linear relationship between AC and VO2/kg only at low AC (Fig. 2, Table 3a). Nevertheless, a significant linear relationship can be modelled over the whole AC range, but the fairly wide prediction intervals indicate that a prediction of energy expenditure from AC using this model equation is not appropriate. In the study of Treuth et al. (19), accelerometer and VO2/kg measures showed a similar distribution as in our study, however, the authors only showed a 95%-CI.
When accounting for the type of activity, the association between AC and VO2/kg diminished, and the extent of energy expenditure could mostly be explained by the type of activity. Therefore, recording the type of activity provided more information on the estimation of VO2/kg than measuring the AC. The following example depicts the shortcomings of the AC: Vacuum cleaning and carrying grocery bags caused similar energy expenditure but obviously the pattern of movement was very different, which resulted in different AC.
The limited association between AC and VO2/kg seems to be explained by two issues: the different types of muscle work generated by different modes of physical activity (i.e. dynamic-concentric and/or isotonic; dynamic-eccentric; isometric muscle work), and the location of the accelerometer. In normal weight adults, a closer correlation of AC and VO2/kg was reported with rhythmic-dynamic activities such as walking and running (5, 8, 10, 16) than with complex activities (12, 21), which are characterized by multiple and non-axial movements of the body or limbs, and by complex types of muscle work (11, 12). The more isometric muscle work is involved in physical activities, the less AC is measured for a similar VO2/kg, and energy expenditure will thus be underestimated by AC. This seems to be independent of whether an accelerometer measures uniaxial or triaxial (14, 21). Two triaxial accelerometers that were worn by obese adults over two weeks in everyday life and were simultaneously evaluated by a doubly labelled water method did not provide accurate estimates of activity-related energy expenditure at individual levels (14). In normal weight adults, the more varying physical activities and body movements are considered, the more inaccurate is the regression equation describing the relation between AC and VO2/kg (15). In this context the location of wearing the accelerometer is relevant. The closer the accelerometer is located to the centre of gravity of the body, the less accurate physical activity can be measured by accelerometers. If the device is fixed at the hip, even small variations of location (e.g. the medial axillar line versus the Spina iliaca anterior) resulted in different AC for the same physical activity (18). The most frequent location of carrying a mobile phone is the front trouser pocket (13). In our study, participants wore the accelerometer attached to a belt, which is close to the trouser pocket. However, the two locations might still lead to differences in the movement pattern.
The aim of obtaining dependencies between metrical physical parameters and VO2/kg as a measure of energy expenditure needs precise transformation of the given parameters for each individual and the current class of activity. Because of the sensor noise, the accuracy of the acceleration data is not sufficient to use kinematic relations, e.g. velocity or distance (1). Bouten et al. (4) are identifying a set of parameters; hereby the sum of the integral of the acceleration of each axis is a promising parameter for the correlation between physical activity and energy expenditure.
Recently, more sophisticated methods to predict energy expenditure from accelerometer data were published that also account for different types of activity (3, 7, 9). These studies suggest that in the future an artificial neural network algorithm as recently proposed by Freedson et al. (9) probably could help to improve estimation of accelerometer-based physical activity metrics across a range of activity types and intensities.
In conclusion, we found a linear relation between AC and VO2/kg only for low AC in both obese and normal weight adults. We suggest that further developments in multidimensional accelerometer technologies should focus on recognizing patterns of physical activity (i.e. body movement including gravity aspects) and the characteristics of muscle work itself in order to differentiate between activities. A combination of hip placed sensors and new generation, smart wrist watches with integrated sensors might lead to a significant improvement in the estimation of VO2/kg as measure of energy expenditure during everyday life activities.

Acknowledgments
We thank Sven Trelle, MD, for statistical and methodological support.

Conflict of interest
None of the authors have conflicts of interest.

LITERATURE

  1. Bao L Physical activity recognition from acceleration data under seminaturalistic conditions (Master thesis). Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, USA, 2003.
  2. Bieber B, Peter C Using Physical Activity for User Behavior Analysis. 1st International Conference on Pervasive Technologies Related to Assistive Environments, Athens. ACM, 2008.
  3. Bonomi AG, Goris AH, Yin B, Westerterp KR Detection of type, duration, and inten-sity of physical activity using an accelerometer. Med Sci Sports Exerc 41 (2009) 1770- 1777.
    doi:10.1249/MSS.0b013e3181a24536
  4. Bouten CV, Koekkoek KT, Verduin M, Kodde R, Janssen JD A triaxial accelerometer and portable data processing unit for the assessment of daily physical activity. IEEE Trans Biomed Eng 44 (1997) 136- 147.
    doi:10.1109/10.554760
  5. Brage S, Wedderkopp N, Franks PW, Andersen LB, Froberg K Reexamination of valid-ity and reliability of the CSA monitor in walking and running. Med Sci Sports Exerc 35 (2003) 1447- 1454.
    doi:10.1249/01.MSS.0000079078.62035.EC
  6. Cooper AR, Page A, Fox KR, Misson J Physical activity patterns in normal, overweight and obese individuals using minute-by-minute accelerometry. Eur J Clin Nutr 54 (2000) 887- 894.
    doi:10.1038/sj.ejcn.1601116
  7. Crouter SE, Clowers KG, Bassett DR JR A novel method for using accelerometer data to predict energy expenditure. J Appl Physiol 100 (2006) 1324- 1331.
    doi:10.1152/japplphysiol.00818.2005
  8. Freedson PS, Melanson E, Sirard J Calibration of the Computer Science and Applica-tions, Inc. accelerometer. Med Sci Sports Exerc 30 (1998) 777- 781.
    doi:10.1097/00005768-199805000-00021
  9. Freedson PS, Lyden K, Kozey-Keadle S, Staudenmayer J Evaluation of artificial neural network algorithms for predicting METs and activity type from accelerometer data: validation on an independent sample. J Appl Physiol 111 (2011) 1804- 1812.
    doi:10.1152/japplphysiol.00309.2011
  10. Goulding A, Jones IE, Taylor RW, Piggot JM, Taylor D Dynamic and static tests of balance and postural sway in boys: effects of previous wrist bone fractures and high adiposity. Gait Posture 17 (2003) 136- 141.
    doi:10.1016/S0966-6362(02)00161-3
  11. Hendelman D, Miller K, Baggett C, Debold E, Freedson P Validity of accelerometry for the assessment of moderate intensity physical activity in the field. Med Sci Sports Exerc 32 (2000) S442- S449.
    doi:10.1097/00005768-200009001-00002
  12. Howe CA, Staudenmayer JW, Freedson PS Accelerometer prediction of energy ex-penditure: vector magnitude versus vertical axis. Med Sci Sports Exerc 41 (2009) 2199- 2206.
    doi:10.1249/MSS.0b013e3181aa3a0e
  13. Ichikawa F, Chipchase J, Grignani R Where's the phone? A study of Mobile Phone Location in Public Spaces. 2nd International Conference on Mobile Technology, Ap-plications and Systems, Guangzhou. IEEE, 2005, 1- 8.
  14. Jacobi D, Perrin AE, Grosman N, Dore MF, Normand S, Oppert JM, Simon C Physical activity-related energy expenditure with the RT3 and TriTrac accelerometers in overweight adults. Obesity (Silver Spring) 15 (2007) 950- 956.
    doi:10.1038/oby.2007.605
  15. Matthews CE Calibration of accelerometer output for adults. Med Sci Sports Exerc 37 (2005) S512- S522
    doi:10.1249/01.mss.0000185659.11982.3d
  16. Nichols JF, Morgan CG, Chabot LE, Sallis JF, Calfas KJ Assessment of physical activity with the Computer Science and Applications, Inc., accelerometer: laboratory versus field validation. Res Q Exerc Sport 71 (2000) 36- 43.
  17. Prentice AM, Goldberg GR, Murgatroyd PR, Cole TJ Physical activity and obesity: problems in correcting expenditure for body size. Int J Obes Relat Metab Disord 20 (1996) 688- 691.
  18. Swartz AM, Strath SJ, Bassett DR Jr, O'Brien WL, King GA, Ainsworth BE Estimation of energy expenditure using CSA accelerometers at hip and wrist sites. Med Sci Sports Exerc 32 (2000) S450- S456.
    doi:10.1097/00005768-200009001-00003
  19. Treuth MS, Schmitz K, Catellier DJ, McMurray RG, Murray DM, Almeida MJ, Going S, Norman JE, Pate R Defining accelerometer thresholds for activ-ity intensities in adolescent girls. Med Sci Sports Exerc 36 (2004) 1259- 1266.
  20. Troiano RP A timely meeting: objective measurement of physical activity. Med Sci Sports Exerc 37 (2005) S487- S489.
    doi:10.1249/01.mss.0000185473.32846.c3
  21. Welk GJ, Blair SN, Wood K, Jones S, Thompson RW A comparative evaluation of three accelerometry-based physical activity monitors. Med Sci Sports Exerc 32 (2000) S489- S497.
    doi:10.1097/00005768-200009001-00008
Corresponding Author:
Prof. Dr. Katharina Meyer
Universitätsklinik Bern
Univ. Poliklinik für Endokrinologie & Diabetologie
Murtenstr. 21
3010 Bern
Schweiz
E-Mail: katharina.meyer@insel.ch