versão On-line ISSN 1980-0037
Rev. bras. cineantropom. desempenho hum. vol.14 no.1 Florianópolis 2012
Marcelo RomanziniI; Edio Luiz PetroskiII; Felipe Fossati ReichertIII
IUniversidade Estadual de Londrina. Centro de Educação Física e Esporte. Departamento de Educação Física. Londrina, PR. Brasil
IIUniversidade Federal de Santa Catarina. Centro de Desportos. Departamento de Educação Física. Florianópolis, SC. Brasil
IIIUniversidade Federal de Pelotas. Escola Superior de Educação Física. Pelotas, RS. Brasil
The aim of this study was to verify the criterion and cross-validity of accelerometer thresholds for distinguishing different physical activity intensities and identifying sedentary behavior in children and adolescents. A systematic literature review was conducted using the PubMed, Scopus, Sports Discus and Web of Science databases. Inclusion criteria were: a) derivation and/or validation of accelerometer thresholds related to intensity of physical activity in youth (age 2 to 18 years); b) use of indirect calorimetry or direct observation as the reference method; c) original research articles published in English, Portuguese or Spanish. Nineteen studies were selected. The accelerometers most often investigated were ActiGraph, RT3 and Actical. Thresholds showed good to moderate validity in the calibration phase (sensitivity = 68 to 100%; specificity = 61 to 100%). Generalizability of the thresholds was higher when they were tested on independent samples (Kappa = 0.72 to 0.91; sensitivity = 79 to 94%; specificity = 72 to 98%) than during independent activities (Kappa = 0.46 to 0.71; sensitivity = 27 to 97%; specificity = 52 to 95%). One calibration study tested the validity of thresholds in independent samples and activities, and only one threshold validation study was found. In conclusion, limited information is available on the generality of accelerometer thresholds for physical activity monitoring in children and adolescents. Validation studies are needed to identify appropriate thresholds for each type of accelerometer.
Key words: Calibration; Motion; Motor activity; Validity of tests.
The use of accelerometry in studies of physical activity in children and adolescents is becoming increasingly common.1 Accelerometers are electronic devices that measure the acceleration of body movement2 and enable objective quantification of the frequency, duration, and intensity of physical activity. Although accelerometry does not provide contextual information on physical activity and is unable to measure certain forms of activity correctly,3,4 its use in child and adolescent research prevents information bias, provides an improved understanding of the relationship between physical activity and health, and enables identification of findings that would otherwise be undetectable by subjective measurements.5
From an operational standpoint, accelerometer counts must be translated into a biologically or behaviorally significant variable.6 This process, known as calibration, involves the identification of thresholds associated with the intensity of physical activity or the conversion of accelerometer counts into units of energy expenditure, using calorimetry or direct observation as a reference method. As the relationship between counts and biological or behavioral measures is influenced by physical and physiological parameters6 and that different accelerometer models collect and store data differently,2 it is recommended that population- and accelerometer-specific thresholds or predictive models of energy expenditure be developed.
Over the past few years, a variety of accelerometer types have been calibrated in samples of children and adolescents, and several thresholds and prediction models have been made available. However, authors have noted that the time spent by children and adolescents on moderate-to-vigorous physical activities varies significantly according to the adopted threshold.7-9 This may hinder comparison between studies on the prevalence of physical activity, and may affect the precision of effect measures in studies of the association between physical activity and health outcomes.
de Graauw et al.10 recently reviewed the validity of predictive models derived from accelerometer counts and found them able to provide precise measurements of physical activity-related energy expenditure in children and adolescents only at the group level. On the other hand, no published studies have systematically reviewed the thresholds available for each accelerometer model and their indicators of validity in this population group. Therefore, the aim of the present study is to provide a systematic review of the criterion and cross-validity of accelerometer count thresholds for the classification of intensity of physical activity in children and adolescents.
The PubMed, Scopus, Sports Discus, and Web of Science databases were searched for studies that derived and/or validated accelerometer counts for determination of physical activity intensity in children and adolescents.
The search was limited to articles published until January 2011. Table 1 describes the search strategy used in each database.
The criteria for inclusion were as follows: a) establishment and/or validation of accelerometer count thresholds for determination of the intensity of physical activity; b) sample composed of children and/or adolescents (218 years); c) use of indirect calorimetry or direct observation as a reference method; d) original research articles published in English, Spanish, or Portuguese. Articles that mentioned accelerometer calibration only as a secondary portion of the Methods section were excluded, as were abstracts, review articles, dissertations, theses, monographs, book chapters, and duplicates. The references of all selected articles were also reviewed in an attempt to identify relevant studies not revealed by the literature search. No additional studies of relevance were thus identified.
In order to enable comparison among thresholds and indicators of validity derived for each accelerometer model, data on a variety of parameters (sample profile, physical activity profile, reference measure, and method used for derivation of thresholds) were extracted from each study. Furthermore, two investigators (MR and FFR) carried out independent assessments of the methodological quality of each study, using a modified version of a checklist that has been previously employed elsewhere in the literature10,11 (Table 2). Any divergences in data extraction were reviewed by a third investigator (ELP).
The criterion validity and cross-validity (generality) of thresholds were analyzed on the basis of the statistical measures adopted and their magnitude. Criterion validity was assessed by comparison of the intensity of physical activity as determined by each threshold versus a criterion measure. Cross-validity was assessed by comparison of the intensity of physical activity as determined by each threshold versus that determined by reference methods in independent samples and/or the same sample engaging in independent activities. Adequate measures of validity included sensitivity, specificity, and 95% limits of agreement (BlandAltman plots). Other measures of validity included percent agreement, Cohens kappa (k), and intraclass, Pearsons product-moment (r), and Spearmans rank correlation coefficients.12 Arbitrarily, a sensitivity and specificity of ≥80% was defined as indicative of good validity (+), ≥60% as indicative of moderate validity (±), and <60% as indicative of poor validity (-). Alternatively, good validity (+) was defined as k>0.60 and ≥90% agreement or r>0.75, moderate validity (±) as k>0.40 and ≥70% agreement or r>0.50, and poor validity (-) as k≤0.40 and <70% agreement or r≤0.50.13
Our search strategy yielded 1558 studies, 19 of which were selected (Figure 1). These studies provided thresholds for seven different models of accelerometers. The most commonly investigated models were ActiGraph, RT3, and Actical. Overall, 16 thresholds for identification of sedentary behavior (SED), 23 for identification of moderately intense activity (MOD), and 20 for detection of vigorous physical activity (VIG) were identified. Five studies used direct observation as the reference method for comparison,14-18 whereas all others used indirect calorimetry for calibration. One study assessed the cross-validity of previously published thresholds for the ActiGraph model.19 Eight provided cross-validation analysis, but only one used an independent sample and independent activities.17
Checklist-derived scores suggest the included studies were of fair methodological quality (mean score, 5.6±1.4 points; range, 3.58.0). Five studies were of high quality (>6 points),14,16,17,20,21 and all others were of moderate quality (3.56.0 points). A single study provided feasibility data, noting an acceptable amount of data loss due to malfunctioning of the ActiGraph accelerometer (<5%).20
THRESHOLDS AND VALIDITY
Overall, thresholds had good to moderate validity in the calibration phase (sensitivity 68100%, specificity 61100%) (Table 3). The ActiGraph model exhibited good validity for SED-related thresholds (sensitivity 86100%, specificity 91100%) and moderate to good validity for MOD-related thresholds (sensitivity 7796%, specificity 61100%) and VIG-related thresholds (sensitivity 68100%, specificity 8095%). In children and adolescents (age 618 years) specifically, ActiGraph thresholds ranged from 100 to 800 counts·min-1 for SED, 19003600 counts·min-1 for MOD, and 39008200 counts·min-1 for VIG. In preschoolers (age 25 years), thresholds ranged from 11001600 counts·min-1 for SED, 16803560 counts·min-1 for MOD, and 33705020 counts·min-1 for VIG.
RT3 thresholds were calculated only in child and adolescent studies. Thresholds ranged from 40420 counts·min-1 for SED, 9501860 counts·min-1 for MOD, and 23304110 counts·min-1 for VIG. Only one study reported validation parameters,21 which were indicative of good threshold validity across all levels of physical activity intensity. The Actical accelerometer exhibited good to moderate validity for SED-related thresholds (sensitivity 8697%, specificity 7298%), MOD-related thresholds (sensitivity 7897%, specificity 7392%), and VIG-related thresholds (sensitivity 7798%, specificity 6179%). In children and adolescents, thresholds were set at 44100 counts·min-1, 15002030 counts·min-1, and 28806500 counts·min-1 for SED, MOD, and VIG respectively. In preschoolers, only one threshold was determined for MOD (715 counts·15s-1) and one for VIG (1411 counts·15s-1).
CROSS-VALIDATION OF THRESHOLDS
Four ActiGraph calibration studies tested the cross-validity of thresholds determined on independent samples and/or activities.17,18,22,26 Cross-validation with independent samples showed good results for the thresholds reported by Vanhelst et al.22 (k=0.720.85) and Reilly et al.18 (sensitivity 83%, specificity 82%). With independent activities, good to moderate cross-validity was found for the MOD and VIG thresholds determined by Pate et al.26 (sensitivity 97 and 66%, specificity 86 and 95% respectively). With independent samples and activities, there was poor to moderate cross-validity for the thresholds reported by Sirard et al.17 (r=0.460.71). An independent study tested the cross-validity of ActiGraph thresholds19 and found that those reported by Evenson et al.23 performed better across different physical activity intensities (k=0.68) as compared with other published thresholds (k=0.620.36).20,24,25 Overall, the Evenson et al. thresholds23 had good to moderate cross-validity for identification of SED, MOD, and VIG activities (Table 2).
Two RT3 calibration studies tested the cross-validity of thresholds determined with independent samples.21,27 Vanhelst et al.27 reported good cross-validity of thresholds across all intensities of physical activity (k=0.750,91), whereas Chu et al.21 found good cross-validity for SED-related thresholds (sensitivity 94%, specificity 98%) and moderate cross-validity of MOD- and VIG-related thresholds (sensitivity 84 and 79%, specificity 72 and 84% respectively). The only cross-validation study of the Actical accelerometer was by Pfeiffer et al.,30 who tested the cross-validity of their thresholds with independent activities and reported moderate cross-validity for the MOD-related threshold (k=0.46) and good cross-validity for the VIG-related threshold (k=0.71). Overall, cross-validity was greater when thresholds were tested on independent samples (k=0.720,91; sensitivity 7994%; specificity 7298%)18,21,22,27 rather than independent activities (k=0.460.71; sensitivity 2797%; specificity 5295%)14,16,26,30 (Table 2).
The present study reviewed the criterion validity and cross-validity of accelerometer count thresholds for classification of physical activity intensity in children and adolescents. The ActiGraph, Actical, and RT3 accelerometers were those most commonly calibrated, and their thresholds were strikingly different for each intensity of physical activity. Overall, there was good to moderate validity for discrimination of SED, MOD, and VIG activities. However, limited information was provided on the cross-validity of these thresholds when tested with independent samples and independent activities.
Differences in the criteria used to define the intensity of physical activity, sample size and profile, study protocol, and statistical procedures may have contributed to the discrepancies in thresholds determined for a single accelerometer model. It bears stressing that there is no clear understanding of which procedures are most adequate for derivation of accelerometer thresholds, which explains the lack of methodological standardization of calibration studies. As an illustrative example, a wide range of criteria have been used to categorize physical activity in the literature, and there is ongoing debate as to whether 3 or 4 METs should be used to define moderate activity in children and adolescents.33,34
Another important methodological aspect concerns the method used for threshold derivation. Traditionally, regression models or ROC curves have been used. The pros and cons of these methods have been discussed at length elsewhere.35,36 In short, although regression models enable derivation of thresholds adjusted to subject characteristics, their high standard error is a major limitation.10 ROC curves, in turn, enable empirical testing of all possible thresholds on the ROC curve plot, which gives investigators the possibility of choosing the appropriate threshold based on the optimal balance between sensitivity and specificity.
Regardless of the accelerometer model, thresholds exhibited good to moderate validity for determination of physical activity intensity as compared with the reference criterion measures adopted in calibration studies. However, in general, thresholds for moderate and vigorous physical activity derived from protocols based on ambulatory activities (walking and running) had better validity15,17,21,30 than thresholds derived from protocols which included a combination of ambulatory and non-ambulatory activities.23,31,32 Indeed, certain non-ambulatory activities (dribbling a basketball, climbing stairs, jumping jacks, step aerobics, martial arts, ball tossing) tend to give lower accelerometer counts than ambulatory activities with a lower energy expenditure.
Accordingly, most of the accelerometers identified in this review are more sensitive to activities with a major vertical acceleration component, such as walking and running. Furthermore, accelerometers tend to be less precise when recording movement of body segments other than that on which they were placed.37 This set of factors may explain the superior validity of thresholds derived from ambulatory activities. However, as the daily activities of children and adolescents are in no way restricted to ambulation, it is advisable for calibration studies to include activities that are truly representative of daily living in this population.36
There was limited information on the cross-validity of the thresholds identified in our review of the literature. Overall, threshold cross-validity indicators were superior when tested on independent samples than when tested with independent activities. Likewise, Corder et al.38 found that the accuracy of predictive models of energy expenditure derived from accelerometer counts is more dependent on the tested activities than on participant characteristics.
Ideally, threshold cross-validity should be tested on independent samples and activities. Sirard et al.17 monitored 269 preschoolers within the school environment on different days and found poor to moderate correlation (r=0.460.70) between the sum of 15-second periods of physical activities of differing intensity when categorized by direct observation and when categorized by thresholds determined during the calibration stage. Trost et al.19 tested the cross-validity of various sets of ActiGraph thresholds in a sample of 206 participants between the ages of 5 and 15, using a protocol consisting of 12 activities of varying intensity (sedentary to vigorous), and found that the thresholds reported by Evenson et al.23 outperformed those reported by Treuth et al.,24 Mattocks et al.,20 and Puyau et al.25 across all levels of physical activity.
The ActiGraph accelerometer is the most widely used model in child and adolescent research,11 and that for which the most thresholds have been published. However, existing ActiGraph thresholds were developed with the uniaxial 7164 and GT1M models. Although the anteroposterior axis of the GT1M accelerometer was made available with second-generation models, current GT1M thresholds use information obtained from the vertical axis alone. The ActiGraph motion sensor currently available on the market is the GT3X model. This device consists of a triaxial accelerometer that collects information from three axes (vertical, medio-lateral, and anterior-posterior) and combines this information into a magnitude vector. Therefore, although acceleration data recorded by the vertical axis of the GT1M model is comparable to that provided by the GT3X model39 for exploration of triaxial GT3X data, thresholds for the magnitude vector of this model have yet to be determined.
In short, the present review found that accelerometer count thresholds have good to moderate validity for estimation of physical activity intensity in children and adolescents. However, there is limited information on the cross-validity of these thresholds on independent samples and activities. Presently, there is evidence to indicate use of the Sirard et al. thresholds17 in preschoolers (poor to moderate cross-validity) and of the Evenson et al. thresholds23 in children and adolescents (good to moderate cross-validity). Future validation studies are required to determine the most appropriate thresholds for each accelerometer model. Count thresholds of the magnitude vector of the GT3X accelerometer are required for exploration of the triaxial capability of this new version of the ActiGraph device.
1. Rowlands AV. Accelerometer assessment of physical activity in children: an update. Pediatr Exerc Sci 2007;19(3):252-66. [ Links ]
2. Chen KY, Bassett DR Jr. The technology of accelerometry-based activity monitors: current and future. Med Sci Sports Exerc 2005;37(Suppl 11):S490-500. [ Links ]
3. Matthew CE. Calibration of accelerometer output for adults. Med Sci Sports Exerc 2005;37(Suppl 11):S512-22. [ Links ]
4. Oliver M, Schofield GM, Kolt GS. Physical activity in preschoolers: understanding prevalence and measurement issues. Sports Med 2007;37(12):1045-70. [ Links ]
5. Reilly JJ, Penpraze V, Hislop J, Davies G, Grant S, Paton JY. Objective measurement of physical activity and sedentary behaviour: Review with new data. Arch Dis Child 2008;93(7):614-9. [ Links ]
6. Freedson P, Pober D, Janz KF. Calibration of accelerometer output for children. Med Sci Sports Exerc 2005;37(Suppl 11):S523-30. [ Links ]
7. Cliff DP, Okely AD. Comparison of two sets of accelerometer cut-off points for calculating moderate-to-vigorous physical activity in young children. J Phys Act Health 2007;4(4):509-13. [ Links ]
8. Guinhouya CB, Hubert H, Soubrier S, Vilhelm C, Lemdani M, Durocher A. Moderate-to-vigorous physical activity among children: discrepancies in accelerometry-based cut-off points. Obesity (Silver Spring) 2006;14(5):774-7. [ Links ]
9. Mota J, Valente M, Aires L, Silva P, Santos MP, Ribeiro JC. Accelerometer cut-points and youth physical activity prevalence. Eur Phy Educ Rev 2007;13(3):287-99. [ Links ]
10. de Graauw SM, de Groot JF, van Brussel M, Streur MF, Takken T. Review of prediction models to estimate activity-related energy expenditure in children and adolescents. Int J Pediatr 2010;489304. doi:10.1155/2010/489304 [ Links ]
11. De Vries SI, Van Hirtum HW, Bakker I, Hopman-Rock M, Hirasing RA, Van Mechelen W. Validity and reproducibility of motion sensors in youth: a systematic update. Med Sci Sports Exerc 2009;41(4):818-27. [ Links ]
12. Szklo M, Javier Nieto F. Epidemiology: beyond the basics. Sudbury: Jones and Bartlett Publishers; 2007. [ Links ]
13. Innes E, Straker L. Validity of work-related assessments. Work. 1999;13(2):125-52. [ Links ]
14. De Bock F, Menze J, Becker S, Litaker D, Fischer J, Seidel I. Combining accelerometry and HR for assessing preschoolers' physical activity. Med Sci Sports Exerc 2010;42(12):2237-43. [ Links ]
15. van Cauwenberghe E, Labarque V, Trost SG, de Bourdeaudhuij I, Cardon G. Calibration and comparison of accelerometer cut points in preschool children. Int J Pediatr Obes 2011;6(2-2):e582-9. [ Links ]
16. Welk GJ, Eisenmann JC, Schaben J, Trost SG, Dale D. Calibration of the Biotrainer Pro Activity Monitor in Children. Pediatr Exerc Sci 2007;19(2):145-58. [ Links ]
17. Sirard JR, Trost SG, Pfeiffer KA, Dowda M, Pate RR. Calibration and Evaluation of an Objective Measure of Physical Activity in Preschool Children. J Phys Act Health 2005;2:345-57. [ Links ]
18. Reilly JJ, Coyle J, Kelly L, Burke G, Grant S, Paton JY. An objective method for measurement of sedentary behavior in 3- to 4-year olds. Obes Res. 2003;11(10):1155-8. [ Links ]
19. Trost SG, Loprinzi PD, Moore R, Pfeiffer KA. Comparison of accelerometer cut points for predicting activity intensity in youth. Med Sci Sports Exerc. 2011;43(7):1360-8. [ Links ]
20. Mattocks C, Leary S, Ness A, Deere K, Saunders J, Tilling K, et al. Calibration of an accelerometer during free-living activities in children. Int J Pediatr Obes. 2007;2(4):218-26. [ Links ]
21. Chu EYW, McManus AM, Yu CCW. Calibration of the RT3 Accelerometer for Ambulation and Nonambulation in Children. Med Sci Sports Exerc 2007;39(11):2085-91. [ Links ]
22. Vanhelst J, Beghin L, Turck D, Gottrand F. New validated thresholds for various intensities of physical activity in adolescents using the Actigraph accelerometer. Int J Rehabil Res 2011;34(2):175-7. [ Links ]
23. Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG. Calibration of two objective measures of physical activity for children. J Sports Sci 2008;26(14):1557-65. [ Links ]
24. Treuth MS, Schmitz K, Catellier DJ, McMurray RG, Murray DM, Almeida MJ, et al. Defining accelerometer thresholds for activity intensities in adolescent girls. Med Sci Sports Exerc 2004;36(7):1259-66. [ Links ]
25. Puyau MR, Adolph AL, Vohra FA, Butte NF. Validation and calibration of physical activity monitors in children. Obes Res 2002;10(3):150-7. [ Links ]
26. Pate RR, Almeida MJ, McIver KL, Pfeiffer KA, Dowda M. Validation and calibration of an accelerometer in preschool children. Obesity (Silver Spring) 2006;14(11):2000-6. [ Links ]
27. Vanhelst J, Béghin L, Rasoamanana P, Theunynck D, Meskini T, Iliescu C, et al. Calibration of the RT3 accelerometer for various patterns of physical activity in children and adolescents. J Sports Sci 2010;28(4):381-7. [ Links ]
28. Kavouras SA, Sarras SE, Tsekouras YE, Sidossis LS. Assessment of energy expenditure in children using the RT3 accelerometer. J Sports Sci 2008;26(9):959-66. [ Links ]
29. Rowlands AV, Thomas PWM, Eston RG, Topping R. Validation of the RT3 Triaxial Accelerometer for the Assessment of Physical Activity. Med Sci Sports Exerc 2004;36(3):518-24. [ Links ]
30. Pfeiffer KA, McIver KL, Dowda M, Almeida MJCA, Pate RR. Validation and Calibration of the Actical Accelerometer in Preschool Children. Med Sci Sports Exerc 2006;38(1):152-7. [ Links ]
31. Puyau MR, Adolph AL, Vohra FA, Zakeri I, Butte NF. Prediction of Activity Energy Expenditure Using Accelerometers in Children. Med Sci Sports Exerc 2004;36(9):1625-31. [ Links ]
32. Tanaka C, Tanaka S, Kawahara J, Midorikawa T. Triaxial Accelerometry for Assessment of Physical Activity in Young Children. Obesity (Silver Spring). 2007;15(5):1233-41. [ Links ]
33. Harrell JS, McMurray RG, Baggett CD, Pennell ML, Pearce PF, Bangdiwala SI. Energy costs of physical activities in children and adolescents. Med Sci Sports Exerc 2005;37(2):329-36. [ Links ]
34. Ridley K, Olds TS. Assigning energy costs to activities in children: a review and synthesis. Med Sci Sports Exerc 2008;40(8):1439-46. [ Links ]
35. Jago R, Zakeri I, Baranowski T, Watson K. Decision boundaries and receiver operating characteristic curves: new methods for determining accelerometer cutpoints. J Sports Sci 2007;25(8):937-44. [ Links ]
36. Welk GJ. Principles of design and analyses for the calibration of accelerometry-based activity monitors. Med Sci Sports Exerc 2005;37(Suppl 11):S501-11. [ Links ]
37. Cliff DP, Reilly JJ, Okely AD. Methodological considerations in using accelerometers to assess habitual physical activity in children aged 0-5 years. J Sci Med Sport 2009;12(5):557-67. [ Links ]
38. Corder K, Brage S, Mattocks C, Ness A, Riddoch C, Wareham NJ, et al. Comparison of two methods to assess PAEE during six activities in children. Med Sci Sports Exerc 2007;39(12):2180-8. [ Links ]
39. Sasaki JE, John D, Freedson PS. Validation and comparison of ActiGraph activity monitors. J Sci Med Sport 2011;14(5):411-6. [ Links ]
Address for correspondence Received: 27 May 2011
Departamento de Educação Física
Universidade Estadual de Londrina
Rod. Celso Garcia Cid, km 380, Campus Universitário
CEP 86051-990, Londrina Paraná, Brasil
Accepted: 21 September 2011
Address for correspondence
Received: 27 May 2011