VIEWS: 7 PAGES: 17 POSTED ON: 2/12/2010
‘Objective Measurement of Physical Activity’ Satellite Meeting of the ICDAM6 Conference Programme & Abstracts University of Southern Denmark Odense, Denmark 26th April 2006 http://www.eyhs-satellite.sdu.dk/ http://www.icdam6.dk/ 1 Organisers The Satellite Seminar is organised by the Institute of Sports Science and Clinical Biomechanics, University of Southern Denmark in co‐operation with the Institute of Public Health, University of Cambridge. Organising Committee Karsten Froberg, University of Southern Denmark (Chair) Søren Brage, University of Cambridge, England Niels Wedderkopp, University of Southern Denmark Jes Bak Sørensen, University of Southern Denmark Niels Chr. Møller, University of Southern Denmark Peter Lund Kristensen, University of Southern Denmark Lone Holm Petersen, University of Southern Denmark Anders Grøntved, University of Southern Denmark Mathias Ried‐Larsen, University of Southern Denmark Scientific Committee Chris Riddoch, Middlesex University, London, England Angie Page, Bristol University, England Sigmund Anderssen, Norwegian School of Sport Sciences, Oslo, Norway Luís Sardinha, Technical University of Lisbon, Portugal Lars Bo Andersen, Norwegian School of Sport Sciences, Oslo, Norway Søren Brage, University of Cambridge, England Karsten Froberg, University of Southern Denmark Contact us Karsten Froberg, Institute of Sports Science and Clinical Biomechanics University of Southern Denmark, Campusvej 55 DK‐5230 Odense M. Denmark E‐mail: email@example.com Phone: +45 6550 3457 Alt. phone: +45 6550 3450 2 Programme 3 12.00‐13.00 Registration and lunch 13.00‐13.10 Karsten Froberg, University of Southern Denmark. Opening message 13.10‐13.50 Nick Wareham, University of Cambridge. Why measure Physical Activity better? 13.55‐14.35 Greg Welk, Iowa State University. Overview of Methods 14.40‐15.20 Søren Brage, University of Cambridge. The need for individual calibration 15.20‐15.40 Coffee break 15.40‐16.20 Ulf Ekelund, University of Cambridge. Summation of free living physical activity intensity data 16.25‐17.05 Lars Bo Andersen, Norwegian School of Sport Sciences. Statistical considerations N. C. Møller, University of Southern Denmark. Unit specific calibration of CSA accelerometers model 7164 in a mechanical setting ‐ is it worth the effort? The effect on inter‐ instrumental reliability in the laboratory and in the field 17.10‐17.50 Ashley Cooper, University of Bristol. Using accelerometry and GPS to investigate the influence of the environment on physical activity 17.55‐18.45 Oral presentation 1. Rusk Z, Metcalf BS, Jeffery AN, Voss LD, Wilkin TC. Demonstration and Validation of a Novel Dashboard to Simplify the Analysis of MTI/CSA Accelerometer Data (The EarlyBird Study) Poster session 1. Besson H, Brage S, Jakes R, Ekelund U, Wareham N. Validation of the Recent Physical Activity Questionnaire (RPAQ) 2. Rumo M. Estimating sleeping and waking periods using globally fixed compared to individually assessed time points 3. Brandes M, Schomaker R, Möllenhoff G, Rosenbaum D. Physical activity vs. quality of gait and quality of life 4. Grant PM, Ryan CG,Tigbe WW, Granat MH. Objective measurement of free‐living physical activity using the activpaltm monitor 5. Mäder U, Ruch N, Rumo M, Martin BW. Simultaneous use of heart rate and accelerometry to estimate energy expenditur. 6. de Vries SI, Bakker I, Hopman‐Rock M, Remy A. Hirasing RA, van Mechelen W. Selecting an appropriate motion sensor in children and adolescents 7. Corder K, Brage S, Wareham NJ, Ekelund U. Comparison of the Actigraph Models 7164 and GT1M during 7 days of free‐living activity in Indian children 8. Slinger J, Kruisifikx J, van Breda E, Kuipers H. Validation of the PAM accelerometer in 11‐15 year old children 18.45‐19.00 Final remarks 19.00‐ Standing buffet 4 Abstracts 5 Satellite Meeting, April 26th 13.10 – 17.50 Objective Measurement of Physical Activity Professor Nick Wareham, University of Cambridge Why measure Physical Activity better? Previous epidemiological studies have clearly demonstrated the importance of physical activity in the aetiology of a range of chronic disorders. Although the instruments used in such studies to measure activity have some limitations, the fact that the associations observed are strong and consistent does rather beg the question ʺWhy bother to measure activity better ?ʺ. This talk will consider this question in relation to descriptive epidemiological studies of variation in activity by time, place and person, analytical studies focusing on specificity and dose‐response, trials and studies of gene‐physical activity interaction. Associate Professor Greg Welk, Iowa State University Overview of Methods This session will provide a broad overview of objective assessments of physical activity. Emphasis will be placed on the use of accelerometry‐based physical activity monitors but other promising technologies will also be described. The presentation will chronicle the evolution of technology for activity assessment and describe the strengths and limitations of the various tools. Issues involved in the calibration of accelerometers will be described along with recommendations for data collection, and the processing and interpretation of accelerometer data. The presentation will conclude with an overview of promising new technologies and approaches that will further advance research on the assessment of physical activity. Career Development Fellow Søren Brage, University of Cambridge The need for individual calibration Combining accelerometry with heart rate monitoring may improve precision of physical activity measurement. Considerable variation exists in relationships between activity intensity and heart rate and accelerometry, which may be reduced by individual calibration. However, the need for individual calibration limits the feasibility of these techniques in population studies and thus it is important to develop simpler methods for calibration whilst maintaining a high degree of validity. In this talk, different alternative procedures for individual calibration are presented in comparison with a reference calibration procedure. 6 Associate Professor Ulf Ekelund, University of Cambridge Summary of free living physical activity intensity data The development of small, light‐weight activity monitors based on accelerometry or the combination of movement sensing and physiological measures has made it possible to assess free‐ living physical activity in large scale epidemiological studies. However, the cleaning, analysis, and interpretation of free‐living physical activity data based on movement sensing need consideration. The current presentation will summarise and discuss different possibilities to summarise free living physical activity intensity data. Professor Lars Bo Andersen, The Norwegian School of Sport Science Statistical considerations EYHS has pooled data from many culturally diverse countries. These diversities have created problems in the assessment of key variables of physical activity, fitness and CVD risk. Patterns of activity differ, e.g. cycling, which make comparisons of MTI measurements difficult. Cycling habits may affect assessment of fitness assessed by cycle ergometry, and some blood measurements are not quite comparable between countries. These problems will be discussed. ‐ Secondly, different approaches of analysing longitudinal data will be given. Doctoral Student Niels Christian Møller, University of Southern Denmark Unit specific calibration of CSA accelerometers model 7164 in a mechanical setting ‐ is it worth the effort? The effect on inter‐instrumental reliability in the laboratory and in the field Introduction Quantifying instrument validity and reliability has become an increasingly important issue as the use of accelerometers has become more and more common. When applying accelerometry in epidemiological studies where multiple units are used, effective unit specific calibration seems crucial in order to minimize the inter‐instrumental out‐put variation which has been observed under standardized conditions in a mechanical setup. In the present study, we hypothesised that if calibration factors were applied to accelerometer out‐ put, a reduced random population variation would be the result. Therefore, the purpose was to calculate unit specific calibration factors in a multiple number of accelerometers in order to test the impact on inter‐instrumental reliability. Methods Considerable variations were observed across the population of instruments in a mechanical setting in the laboratory as 53 accelerometers were undergoing a rather comprehensive quality check before, and during, the data collecting period in the Danish part of the European Youth Heart Study II. Therefore, individual calibration factors were calculated for all units since our intention was to examine whether calibration could increase inter‐instrumental reliability. The effect on inter‐instrumental reliability was analysed by multiplying unit specific calibration factors 7 to a) data obtained in a mechanical setup in the laboratory during the field data collecting period, and to b) data collected during free living conditions in the field. Results When applying unit specific calibration factors to data derived in the mechanical setup in the laboratory, accelerometer out‐put went from 5025 count*min‐1 to 4927 counts*min‐1. Calibration caused impact on increased inter‐instrumental reliability was quite marked as the standard deviation was reduced dramatically from 361 (95% CI=338;388) counts*min‐1 to 137 (95% CI=128;147) counts*min‐1 (62%). When applying unit specific calibration factors to field data collected during free living conditions, raw average activity intensity went from 446 counts*min‐1 to 440 counts*min‐1. Calibration had literally no effect on inter‐instrumental reliability, as the group variation (standard deviation) decreased by a minimum from 162 (95% CI=150;176) to 157 (95% CI=146;171) counts*min‐1 (3%). Conclusion Unit specific calibration factors shown to be appropriate to increase inter‐instrumental reliability in the experimental setting in the laboratory should be addressed as highly questionable when applied to field data reflecting more complex and heterogeneous movement of the human body. Associate Professor Ashley Cooper, Bristol University Using accelerometry and GPS to investigate the influence of the environment on physical activity The environment, both real (eg access to parks or other green spaces) and perceived (eg beliefs about how safe the environment is for play), is suggested to influence childrens levels of physical activity. There is some evidence that perceptions of the environment are associated with physical activity levels but little data about childrens use of the built environment to be active. Minute by minute accelerometry is now a common method of measuring the level and patterns of childrens activity, but does not tell you where activity takes place. The global positioning system (GPS) allows the location of an individual to be accurately mapped. Integration of these two methods will enable us to identify where around the school or home environment physical activity takes place. This presentation will describe these techniques and present preliminary data describing use of the environment for physical activity by primary school aged children. 8 Oral and Poster presentations 17.55 – 18.45 Oral presentations Abstract 1 Demonstration and Validation of a Novel Dashboard to Simplify the Analysis of MTI/CSA Accelerometer Data (The EarlyBird Study) Zoe Rusk, Brad S Metcalf, Alison N Jeffery, Linda D Voss, Terence J Wilkin Dept of Endocrinology & Metabolism, Peninsula Medical School, Plymouth, UK Introduction MTI/CSA accelerometers are an objective and accurate means of recording the physical activity (PA) of children. However, manual processing of the data is time consuming, vulnerable to error and liable to inconsistency. There is pressing need for a user‐friendly ‘dashboard’, a macro based on menu choices which set the criteria for the cleaning and capture of data in advance. Methods The ‘EarlyBird PA dashboard’ is an EXCEL™ macro designed to automate the steps routinely taken when manually processing physical activity files. The macro was designed around six principles, the dashboard’s menu allowing the user to: • Select thresholds to define up to 5 different intensities of physical activity • Select up to 7 time periods for analysis. • Set threshold for and patch in strings of zeros assumed to be when the monitor is removed during waking time. • Select earliest time in the morning and latest time in the evening/night where only PA data between these times would be considered. • Patch in periods of unrealistic intensity i.e. ≥14,000cpm for young children. • Display patched‐in data and % contribution it makes to the total data analysed. ‘Patching in’ means replacing cells with the mean activity recorded during the same time period on other equivalent days. Results We have put together an animated PowerPoint presentation demonstrating the EarlyBird dashboard in practice, showing the macro processing a week’s activity file in approximately three minutes (manually 10‐15mins). We compared the macro data with manual data obtained from processing files of 100 EarlyBird children (7‐10y). The macro‐v‐manual mean differences in: total PA counts and minutes spent at low, medium and high intensities were all <3.0% (p<0.05, power>99%). E.g. for total PA counts the mean difference between the two methods was 0.5% where 95% of individual differences were within 4.5%. The two methods were highly correlated (r=0.99) and the slope coefficient from a linear regression analysis was 0.994. Conclusions The dashboard is reliable in practice, and takes less than a quarter of the time to process the data. It is envisaged that an Internet version could be made available for wider use. 9 Posters presentations Abstract 1 Validation of the Recent Physical Activity Questionnaire (RPAQ) H Besson, S Brage, R Jakes, U Ekelund, N Wareham MRC Epidemiology Unit, Cambridge CB1 9NL, United Kingdom Introduction We previously reported the validity of the EPIC‐Norfolk Physical Activity Questionnaire (EPAQ2) assessing usual activity in work, travel, recreation and domestic life using the past year as the frame of reference. As the validity was limited, we shortened the frame of reference to develop a new Recent Physical Activity Questionnaire (RPAQ) aimed at assessing usual physical activity in the past month. We now report the repeatability of RPAQ and its validity in estimating energy expenditure compared to objective measures of energy expenditure from the Doubly Labelled Water (DLW) method. Methods Twenty five women and 25 men, aged 21 to 55 years old were recruited. Total energy expenditure (TEE) was measured for 14 days with the DLW method. Last month physical activity was reported at the end of the DLW measurements. Repeatability was measured for a sub‐sample of 38 subjects who completed the questionnaire for a second time 4 months later. Energy costs of activities were scored according to the Physical Activity Compendium. Results Mean body mass index of participants was 24.5 kg/m2 (SD 3.5) and their physical activity level ranged from 1.19 to 2.56. The partial correlation coefficient for total score of activity (MET‐h/day) adjusted on time between administrations was 0.36 (p = 0.03). The lowest level of repeatability in a sub‐domain of activity was for activity at work (r=0.24; p = 0.14), whereas the highest was for vigorous activities (r = 0.70; p < 0.0001). Age and gender did not affect these associations. Calculated TEE from RPAQ was associated with TEE from DLW (r = 0.72; p < 0.0001). Similarly, a correlation was observed between estimated physical activity energy expenditure (PAEE = TEE – RMR [resting metabolic rate]) and measured PAEE (r = 0.46; p = 0.0008). None of these associations was affected by age and gender. Conclusions This study suggests that the RPAQ is a valid instrument for estimating TEE and PAEE in healthy adults. Although the results from the repeatability study are modest, they reflect more the stability of activity rather than test‐retest reliability since the two administrations of the questionnaire did not overlap the same time‐period. 10 Abstract 2 Estimating sleeping and waking periods using globally fixed compared to individually assessed time points. Martin Rumo Swiss Federal Office of Sport Introduction Accelerometers are widely used in objective measurement of physical activity. In order to adequately interpret behavioural patterns concerning physical activity, it is crucial to define waking and sleeping periods. If no self‐reports are available from the subjects, two methods are possible: Either one globally fixes points of time for rising and going to bed or one tries to estimate these points of time from the accelerometry data. The application of both methods is compared, by using an algorithm to assess sleeping and waking periods from the accelerometry data. Methods Accelerometry data was recorded from 50 individuals (26 males, 24 females, Age: 39 ± 11 years) wearing an Actigraph (Manufacturing Technology Inc. (MTI), Fort Walton Beach, FL, USA) for 4 to 7 consecutive days. Counts were recorded minute‐by‐minute. For each day the longest time period, during which the mean counts over 10 minutes do not exceed 180 were considered to be the sleeping period. For the accordingly assessed waking periods, the lengths of periods of low, moderate and vigorous activity as well as the mean counts per minute (mcpm) were compared with each other in a paired t‐test with the respective values when the waking period was globally set to 7.00 to 23.00. Results The differences between the two methods did reach statistical significance in all four variables (low: p<.001, moderate: p<.001, vigorous: p=.016, mcpm: p<.001). Their respective means were 731.1 ± 78.9 min/day , 157.8 ± 55.7, 11.9 ± 10.8, 437.4 ± 143.95 for the individually assessed method and 802.4 ± 52.4, 147.5 ± 51.2, 11.0 ± 10.17 380.4 ± 129.29 for the globally fixed method. Conclusions Globally fixing the waking period can lead to over‐ or underestimation of durations of all activity intensities as well as mean counts per minute. Since the automated method identifies the waking period with the interval that contains the moderate and vigorous activities, that can be lost by using globally fixed time points, it is recommended to use the individually assessed time points when no self‐reports are available. 11 Abstract 3 Physical activity vs. quality of gait and quality of life Mirko Brandes, Ralph Schomaker, Gunnar Möllenhoff, Dieter Rosenbaum Motion Analysis Lab, Orthopaedic Department, University Hospital of Muenster, Germany Introduction This study investigated the relationship between 1. physical activity (quantity of gait) measured with objective devices, 2. quality of gait assessed by clinical gait analysis and 3. quality of life. To ensure a larger spectrum in the quantity of gait, subjects designated to joint replacement due to unilateral osteoarthritis in the knee or hip joint were chosen. Methods Physical activity (PA) of 26 patients (58.6 years) was assessed with a) The accelerometer‐based DynaPort ADL‐monitor measured posture and locomotion for one day and b) The SAM Step‐ Activity‐Monitor, a small microprocessor‐operated acceleration sensor, recorded the number of gait cycles in 1‐minute‐intervals for one week. PA was compared to a group of 26 healthy subjects. The quality of gait was assessed with a six camera system in combination with two force plates. Quality of life was evaluated with the SF‐36 questionnaire. Spearman‐rank‐correlations were calculated between PA level, quality of gait and quality of life. Results The patients had worn the ADL‐monitor for 13.8 ± 1.5 hours and the SAM for 14.0 ± 1.4 hours per day. According to the ADL‐monitor, the subjects used 10.5% of the recorded time for locomotion, 32.6% for standing, 43.9% for sitting and 12.5% for lying. The patients performed 4782 ± 2116 gait cycles per day. Nearly 80% of these gait cycles were performed with intensities <21 gait cycles per minute. The number as well as the intensity of patients’ gait cycles differed significantly from healthy adults. The sub‐categories of the SF‐36 for physical functioning and general health showed 38 and 56 out of 100 points, respectively. Moderate correlations were found between the PA level and the quality of life but not between the PA level and the quality of gait. Conclusions The PA of the control group was nearly identical with previously reported healthy adults (1), whereas patients’ activity level was lower than values reported for hip and knee patients after surgery (2). Regarding the poor to moderate correlations between PA, quality of gait and quality of life, it appears mandatory for an objective judgement of a subject’s activity level to measure PA directly using appropriate methods. References (1) Busse, M. E., et al.: Quantified measurement of activity provides insight into motor function and recovery in neurological disease. J Neurol Neurosurg Psychiatry. 75:884‐888, 2004. (2) Silva, M., et al.: Average patient walking activity approaches 2 million cycles per year: pedometers underrecord walking activity. J Arthroplasty. 17:693‐697, 2002. 12 Abstract 4 Objective measurement of free‐living physical activity using the activpaltm monitor P. M. Grant, C. G. Ryan, W. W. Tigbe, M. H. Granat School of Health & Social Care, Glasgow Caledonian University, Glasgow, Scotland Introduction Objective information about activity patterns and the amount of physical activity undertaken in the population is needed to provide appropriate recommendations for health programmes. However, even with recent advances in technology allowing for the use pedometery and accelerometery, accurate measurement of physical activity remains problematic. The activPALTM professional physical activity monitor (PAL Technologies Ltd, Glasgow, UK), is a single unit device, requiring no calibration, that identifies in real time, episodes of walking, sitting and standing allowing the measurement of both activity and inactivity. In addition, the monitor records step number and instantaneous cadence. The studies aimed to evaluate the validity and reliability of the activPAL monitor as a measure of free‐living activity. Study 1 Method Ten adults wore three activPAL monitors to carry out everyday activities. Activities were videoed and classified by three observers into sitting, standing and walking. Data from the activPAL similarly classified and compared to those of observation (criterion method). Results Inter‐observer reliability [ICC (1)] was excellent (>0.97) and inter‐device reliability [ICC (1)] good‐ excellent (0.78‐0.99). Agreement between activPAL and observation by the Bland‐Altman method (2) was good‐excellent, mean difference of ‐2.0% to 0.2% (limits of ‐16.1% to 12.1%). Second‐by‐ second analysis yielded a percentage agreement of 95.9% with sensitivity and predictive values 88.1% to 99.6% for different activities. Study 2 Method Step‐count and cadence in 20 adults (12F, 8M) were evaluated during treadmill walking for five different speeds and outdoor walking at three self selected speeds. Trials were videoed and observation used as the criterion measure. Each participant wore four activPAL monitors. Results Inter‐device reliability [ICC (1)] was ≥0.99 for step‐count and cadence. Agreement by Bland‐ Altman method (2), between the activPAL and observation, was excellent for step‐count and cadence at all speeds, mean difference of 0% ‐ 1.0% (limits of ‐2.6% to +3.2%). Conclusion The activPAL activity monitor is a valid and reliable objective measure of free‐living physical activity. References Portney LG and Watkins MP. Foundations of Clinical Research. Applications to Practice. Norwalk, Conn.: Appleton & Lange, 1993 Bland JM, Altmann DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999; 8: 135‐60. 13 Abstract 5 Simultaneous use of heart rate and accelerometry to estimate energy expenditure. Urs Mäder, Nicole Ruch, Martin Rumo, Brian W. Martin. Federal Institute of Sports Magglingen, Switzerland Introduction The importance of habitual physical activity for disease prevention is widely recognized. However, physical activity remains difficult to be measured accurately and the available measurement techniques have inherent limitations. The simultaneous use of heart‐rate and accelerometry may be useful to overcome some limitations of those methods, commonly used separately. The purpose of this study was to validate an approach that used heart rate and accelerometry (HRA) data to estimate energy expenditure (EE). Method Heart rate and accelerometry (biaxial) data were collected simultaneously by a chest mounted device every 10 sec during several activities. 6 of the volunteers were women (30.5 ± 11.6 y, BMI = 22.8 ± 5.0), 10 were men (37.2 y, BMI = 26.6 ± 3.3). EE was measured by a portable metabolic system. Data of walking at 3 velocities on the flat and sitting were used to calibrate all measured values before scanning for activity specific data patterns and defining 8 activity classes. Prediction equations were developed to calculate EE for each activity class. Estimated EE based on HRA data was validated on 12 volunteers (6 females, 41.5 ± 14 y, BMI = 21.4 ± 2.6; 6 males, 39.3 ± 16.3 y, BMI = 22.3 ± 1.0) against whole‐body indirect calorimetry (WIC) during 10 h. The protocol included walking, biking, playing, and stepping, each for 30 min. Before entering the WIC, the volunteers performed the calibration procedure described above. Results: Total EE, averaged over 10 h, calculated by HRA and measured by WIC was equal 2.5 ± 0.3 kcal/min. Average difference between the two method was 4.1 ± 3.1 kcal/min (Min = 0.7 %, Max = 10.1 %). There was no significant differences between EE determined by HRA and WIC during specified activities (walking: 1.1, p = 0.3; playing: 1.6, p = 0.1; stepping: ‐1.8 kcal/min, p = 0.1) except for biking (‐3.9 kcal/min, p= 0.002). Discussion The simultaneous use of heart rate and accelerometry seems to be a valid approach to estimate EE. However, additional comprehensive studies are needed to evaluate the validity of EE measurement by (HRA) during longer periods under free living conditions. 14 Abstract 6 Selecting an appropriate motion sensor in children and adolescents Sanne de Vries, Ingrid Bakker, Marijke Hopman‐Rock, Remy A. Hirasing, Willem van Mechelen. TNO Quality of Life, Department of Physical Activity and Health, Leiden, The Netherlands Introduction In the past decades, motion sensors have evolved from simple mechanical devices to three‐ dimensional accelerometers that can be used to assess physical activity or to estimate energy expenditure. Because many children and adolescents have difficulties in accurately recalling their physical activities, motion sensors are being used with increasing regularity. The purpose of the present study was to systematically review published evidence on the reproducibility, validity, and feasibility of motion sensors used to assess physical activity in healthy children and adolescents (2‐18 yr). Methods A systematic literature search was performed in PubMed, Embase, and SpycINFO. The clinimetric quality of two pedometers (Digi‐Walker, Pedoboy), four one‐dimensional accelerometers (LSI, Caltrac, Actiwatch, CSA/ActiGraph), and three three‐dimensional accelerometers (Tritrac‐R3D, RT3, Tracmor2) was evaluated and compared using a 20‐item checklist. Results Overall, the quality of the studies (n = 35) and therefore the level of evidence for the reproducibility, validity, and feasibility of the motion sensors was modest (mean = 6.4 ± 1.6 out of 14 points). There was strong evidence for a good reproducibility of the Caltrac in adolescents (12‐ 18 yr), a poor reproducibility of the Digi‐Walker in children (8‐12 yr), a good validity of the CSA/ActiGraph in children and adolescents (8‐18 yr), and a good validity of the Tritrac‐R3D in children (8‐12 yr). Conclusions From this study it can be concluded that: 1. The CSA/ActiGraph is the most studied motion sensor in children and adolescents. There is extensive evidence for a good reproducibility, validity, and feasibility of the CSA/ActiGraph in healthy children and adolescents (reproducibility: 4‐18 yr; validity: 3‐18 yr); 2. There is no information on the reproducibility of motion sensors in preschool children (2‐4 yr); 3. There is no information on the reproducibility of three‐dimensional accelerometers. Researchers and practitioners are strongly encouraged to regularly assess and report the clinimetric properties of the motion sensors they use, although not without improving the quality of the reported information. 15 Abstract 7 Comparison of the Actigraph Models 7164 and GT1M during 7 days of free‐living activity in Indian children Kirsten Corder, Søren Brage, Nicholas J. Wareham and Ulf Ekelund MRC Epidemiology Unit, Elsie Widdowson Laboratory, 120 Fulbourn Road, Cambridge, CB1 9NL, U.K Introduction The Actigraph (formerly CSA/MTI) is the most commonly used accelerometer in physical activity research. It has been proved valid to measure physical activity in children, both in a controlled laboratory environment and during free‐living (1,2). The new version (GT1M) replaces the discontinued Model 7164 and the two monitors differ substantially. To our knowledge, data from these two monitors has not been directly compared. This study aimed to determine whether data from the new Actigraph (GT1M) is comparable to that from the commonly used model (7164). Methods 30 adolescents (15.8±0.6y) from Chennai, India wore the 7164 and GT1M Actigraph accelerometers simultaneously for 7 days. Two time‐synchronised Actigraphs, one 7164 and one GT1M, were attached to the same tight elastic waist belt, one monitor placed centrally on each hip. The GT1M and 7164 were set to record at 5 and 60 second epochs respectively, the shortest epoch allowing for 7 days of continuous step and count measurement. All analyses were carried out using mean counts per minute (cpm), a measure which is unaffected by the time resolution of the activity data. The agreement between monitors was assessed using the Bland‐Altman method (3), Pearson correlation coefficients, paired t‐test and regression analyses. Results The correlation between monitors was r=0.89. The GT1M measured on average 7% higher than model 7164 (372.9±19.5 and 347.9±19.0 cpm, P=0.0121). A correction factor of 0.928 is suggested for comparison between 7164 and GT1M data. Bland‐Altman plots revealed no evidence of heteroscedasticity or trend for the difference between the two monitors to be associated with activity level. Conclusions The correlation between the monitors was high, indicating that data from the GT1M and 7164 should be comparable using a correction factor. The increased memory capacity and reduced inter‐ monitor variability of the GT1M (unpublished results from mechanical calibration) suggest that the GT1M should produce more accurate and standardised physical activity data than the Model 7164, which could still be compared to historical data on mean counts per minute. References 1. Trost, S. et al. MSSE 30:629‐633,1998. 2. Eston, R., et al. JAP 84:362‐371,1998. 3. Bland, J. and Altman, D. Lancet 8:307‐310,1986. 16 Abstract 8 Validation of the PAM accelerometer in 11‐15 year old children Jantine Slinger, Liesbeth Kruisifikx, Eric van Breda, Harm Kuipers Department of Health Sciences, Maastricht University, 0031‐433881383 Introduction It is important to test the validity of accelerometers in children, to be able to use the instruments in studying the relationship between physical activity and health. The Physical Activity Monitor (PAM) is a new Dutch accelerometer and is valid in use with adults. The aim of this study is to examine the validity of the PAM in children 11‐15 years old. Methods The PAM was validated against VO2 intake in 36 children. The children performed three different types of activities: treadmill walking in different velocities, four different exercises on the treadmill and cycling on an ergo meter. Results No significant correlations were found between PAM and VO2 for the tested activities. The linear mixed models method produces the following formula to predict the VO2 value from the PAM value in the walking part: VO2ij = 452 + 12,5 pamij + αi + βipamij + eij. It was not possible to analyse the variable exercises part with this method. The linear mixed models of the cycling part showed that an increasing VO2 value does not affect the PAM value. Conclusions Contrary to the expectations we did not find significant correlations between PAM and VO2 values. This is probably due to inter individual differences. The linear mixed models method shows a formula to predict the VO2 out of the PAM values, but because of a large error term, this formula can not be used in studying the physical activity level with the PAM in children. As expected in the cycling part there were no significant correlations, because the PAM is placed on the hip and measures movement of the torso and during cycling the torso does not move enough. We conclude that the PAM is not valid to evaluate the physical activity level in 11 to 15 year old children. The seminar is supported by: The Danish Ministry of the Interior and Health The Sports Science Research Council of the Ministry of Culture The Faculty of Health Sciences, University of Southern Denmark 17
"Children_ Physical Activity _ Health"