Electronic Supplementary Material Table S1 Summary of Key Methodological Issues Related
Electronic Supplementary Material Table S1 Summary of key methodological issues related to GT3X/+ data collection protocols and data processing in validation/calibration studies in preschoolers (the rest of studies are available on request)
References / Study design(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Preschoolers (n=24, Validation, calibration and comparison studies reviewed for objectives 1 and 2: n=11)
Butte et al. [1] / Part 1:
Lab/controlled
Part 2:
Free-living observational
(7 days, waking hours protocol)
[Criterion validity: room calorimetry and doubly labelled water] / SED/PA intensity classification
EE algorithm
To develop and validate cross-sectional time series and multivariate adaptive regression splines models based on accelerometry and heart rate for the prediction of EE using room calorimetry and doubly labelled water and established accelerometry cut points for PA levels / Part 1:
50,
4.5±0.8 y
Part 2:
105,
4.6±0.9 y / GT3X+ (right hip) / Normal / NS
60 s / Part 1:
Not used
Part 2:
≥20-0-0
≥1000 min/day
≥4 days/week / Cut-points:
Butte SED, LPA/MPA, MPA/VPA VA and VM cut-points / Cross-sectional time series and multivariate adaptive regression splines models are acceptable for the prediction of EE in preschool-age children. Cut points were satisfactory for the classification of SED, LPA, and MVPA in preschoolers
Costa et al. [2] / Part 1:
Lab/controlled
Part 2:
Lab/controlled
[Criterion validity: direct observation] / SED/PA intensity classification
To calibrate and validate against direct observation (video recorded) the GT3X+ to measure PA and SED in preschool children / Part 1:
18,
2.8±0.5y
Part 2:
38,
2-3 y / GT3X+ (right hip) / LFE / 80 Hz.
5 s and 15 s / Not used / Cut-points:
Costa SED and MVPA VA and VM cut-points / The 5 s VA cut-points showed smaller biases in estimated sedentary behaviour and PA time in 2-3 years. MVPA showed a high overestimation time
References / Study design
(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Flynn et al. [3] / Part 1
Lab/controlled
Part 2
Lab/controlled
Part 3
Free-living observational
(1 school-day -8:30 a.m. to 4:00 p.m.-)
[Calibration study: reliability] / Others: light sensor
1. To assess reliability of the GT3X+ ambient light sensor
2. To identify a lux threshold to accurately discriminate between indoor and outdoor activities in children
3. To test the accuracy of the lux threshold in a free-living environment / Part 1
20 GT3X+
Part 2
18,
7.0±2.3 y
Part 3
18,
4.4±0.4. y / GT3X+
(right hip) / NS / 60 Hz.
60 s (only for part 3) / Part 1:
Not used
Part 2:
Not used
Part 3:
Not used
≥4 h/day
NS / Not used / In part 1, there was high interinstrument reliability for the light sensor.
In part 2, the optimal lux threshold was determined to be 240 lux.
In part 3, results of the school-day validation demonstrated the monitor was 97% accurate for overall detection of indoor and outdoor conditions.
Martin et al. [4] / Free-living observational
(7 days, waking hours protocol)
[Criterion validity: ActivPAL] / SED/PA intensity classification
To compare activPAL and GT3X monitors in free-living children. / 23,
4.5±0.7 y / GT3X (right hip)
ActivPAL / NS / 30 Hz.
60 s / NS
≥6 h/day
≥3 days/week / Cut-points:
Reilly SED VA cut-point / Differences exist between activPAL and GT3X in SED. GT3X presented lower values compared to activPAL
References / Study design
(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Janssen et al. [5] / Lab/controlled
[Criterion validity: room calorimetry and direct observation] / SED/PA intensity classification
EE algorithm
1. To examine the predictive validity of ActiGraph EE equations against room calorimetry
2. To compare the classification accuracy of ActiGraph cut-points for classifying SED and PA intensity in 4-6 years old against direct observation / 40,
4-6 y / GT3X
(right hip) / NS / 30 Hz.
15 s or 60 s depending on the cut-points used / Not used
Not used / Cut-points:
Evenson, Pate, Puyau, Reilly, Sirard and
Van Cauwenberghe SED, LPA and MVPA VA cut-points
EE algorithm
Pate and Puyau equations / EE MVPA-Pate equation.
For the rest, authors did not recommend to apply Pate or Puyau due to underestimation.
Evenson cut-points showed significantly higher classification accuracy for SED. Pate cut-point showed significantly higher accuracy for MVPA
Jimmy et al. [6] / Lab/controlled
[Criterion validity: indirect calorimetry] / SED/PA intensity classification
To develop and validate cut-points for the VA and the VM in 5 to 9 year old against indirect calorimetry / 32
5-9 y / GT3X
(right hip) / Normal / 30 Hz.
5 s / Not used / Cut-points:
Jimmy MVPA and VPA VA and VM cut-points / Current cut-points adequately reflect MVPA and VPA in young children. Cut-off points based on VM counts did not appear to reflect the intensity categories better than cut-off points based on VA counts alone
Johansson et al. [7] / Part 1: Lab/controlled
Part 2: Free living observational
(7 days, waking hours protocol)
[Criterion validity: direct observation] / Placement
SED/PA intensity classification
To calibrate and validate the GT3X+ for wrist-worn placement in young preschoolers against direct observation (video recorded) / 38,
15-36 month / GT3X+
(left wrist) / Normal / 30 Hz.
5 s / Part 1:
Not used
Part 2:
NS
≥11h/day of SED removed / Cut-points:
Johansson SED, LPA, VPA VA and VM cut-points / The developed intensity thresholds appear valid in order to categorize sedentary and physical activity intensity categories in preschool children
References / Study design
(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Kahan et al. [8] / Part 1: Lab/controlled
(3 daily unstructured morning play periods)
Part 2: Lab/controlled
(3 daily unstructured morning play periods)
[Convergent validity] / SED/PA intensity classification
To evaluate convergent validity of accelerometry with direct observation using methods that are either consistent with current practise/recommendation / Part 1: 57,
4.7±0.3 y
Part 2: 12, (NS) / GT3X (right hip) / NS / 30 Hz.
Part 1:
15 s
Part 2:
5 s / ≥30-0-0
NS / Cut-points:
Evenson, Pate, Sirard and Van Cauwenberghe SED and MVPA VA cut-points / Cut-points of Sirard are the best converged with the direct observation measure for SED and MVPA.
Also, Sirard and Pate cut-points are more sensitive in detecting SED and MVPA, respectively
Meredith-Jones et al. [9] / Free-living observational
(7 days, 24h protocol)
[Comparison study without a criterion] / SED/PA intensity classification
Sleep algorithm
To compare the effect of 6 different sleep-scoring rules on PA and SED / 291,
4-8 y / GT3X
(right hip) / Normal / 30 Hz.
15 s / ≥60-0-0
≥8 h/days
≥3 days/week / Cut-points:
Puyau SED, LPA and MVPA VA cut-points
Sleep Algorithm:
Sadeh sleep algorithm for method 5 / Different methods of removing sleep from 24h data markedly affect estimates of SED, yielding values ranging from 556 to 1145 min/day. Estimates of NWT (33–193 min), wear time (736–1337 min) and CPM (384–658) also showed considerable variation. By contrast, estimates of MVPA were similar, varying by less than 1 min/day
Pulakka et al. [10] / Free-living observational
(7 days, 24h protocol)
Two free play sessions recorded at home during these days.
[Criterion validity: direct observation] / SED/PA intensity classification
Test the feasibility and validity of the GT3X in measuring PA of rural Malawian toddlers against direct observation (video recorded) / 56,
16-18.5 months / GT3X
(right hip) / Normal / 30 Hz.
15 s / NS / Cut-points:
Pulakka SED/LPA and MVPA VA and VM cut-points / The acc. proved a feasible and valid method of assessing PA among Malawian toddlers
References / Study design
(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Zakeri et al. [11] / Lab/controlled
[Criterion validity: room calorimetry] / EE algorithm
To develop cross-sectional time series and multivariate adaptive regression splines models for prediction of EE based on accelerometry, heart rate and calorimetry method and validate them against room calorimetry / 69,
3-5 y / GT3X+ (right hip) / Normal / 30 Hz.
60 s / Not used
Not used / EE algorithm:
cross-sectional time series and multivariate adaptive regression splines statistical models / Cross-sectional time series and multivariate adaptive regression splines models should prove useful in capturing the complex dynamics of EE and movement that are characteristics of preschoolers
Acc: accelerometer, BMI: body mass index, CPM: counts per minute, EE: energy expenditure, h: hours, Lab: laboratory condition, LFE: low-frequency extension, LPA: light physical activity, MPA: moderate physical activity, MVPA: moderate-to-vigorous physical activity, NS: not specified, NWT: non-wear time, PA: physical activity, s: seconds, SED: sedentary time, VA: vertical axis, VM: vector magnitude, VPA: vigorous physical activity
a NWT definition expressed as: minimum minutes of 0 CPM – minimum minutes for before and after allowance windows – maximum minutes of allowance
1
Electronic Supplementary Material Table S2 Summary of key methodological issues related to GT3X/+ data collection protocols and data processing in validation/calibration studies in children and adolescents (the rest of studies are available on request)
References / Study design(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Children and adolescents (n=81, Validation, calibration and comparison studies reviewed for objectives 1 and 2: n=26)
Aibar et al. [12] / Free-living observational
(7 days, waking hours protocol)
[Comparison study without a criterion] / Epoch length
To examine the effect of different epoch lengths (3-60 s) on MVPA, 10 minute bouts of MVPA and compliance with World Health Organization guidelines. / 401,
14.5±0.7 y / GT3X
(right hip) / NS / 30 Hz.
3 s
5 s
10 s
15 s
30 s
45 s
60 s / ≥10-0-0
NS / Cut-points:
Evenson MVPA VA cut-points / Significant epoch effect for time spent in MVPA, 10 minute bouts of MVPA and the extent of compliance with guidelines percentage of compliance of guidelines. Shorter epochs such as 5 s, 10 s and 15 s are proposed for comparative studies carried out with adolescents
Aittasalo et al. [13] / Lab/controlled
[Criterion validity: direct observation
Convergent validity] / SED/PA intensity classification
To compare mean amplitude deviation of raw acceleration signal and to develop cut-points based on the mean amplitude deviation in two different acc. brands. / 20,
14.2±0.7 y / GT3X
(left hip)
AM13
(right hip) / Not used / 30 Hz.
5 s / Not used / Cut-points:
Aittasalo SED, LPA, MPA and VPA VA cut-points for raw data VM / The cut-points were almost identical in the two brands and indicate that it is possible to find a method, which classifies similarly the intensity of adolescents’ PA from raw acceleration data irrespective of acc. brand.
References / Study design
(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Barreira et al. [14] / Free-living observational
(7 days, 24h protocol)
[Criterion validity: daily logs] / Sleep algorithm
1. To add layers and features to a previously published fully automated algorithm designed to identify children’s nocturnal sleep and to exclude episodes of night time non-wear/wakefulness and potentially misclassified daytime sleep episodes
2. To validate this refined sleep algorithm against daily logs / 45,
10.0±0.4 y / GT3X+
(hip) / LFE / 80 Hz.
60 s / Not used / Sleep algorithm:
Tudor-Locke refined sleep algorithm / Refined sleep algorithm total sleep episode time (36 min) was significantly different from Log sleep period time (24 min), but not different from Log + Acc. sleep period time (24 min). Significant and moderately high correlations were apparent between refined sleep algorithm determined variables and those using the other methods. There were no differences between refined sleep algorithm and Log + Acc. estimates of sleep onset or sleep offset and log wake time
Chandler et al. [15] / Lab/controlled
[Criterion validity: direct observation] / SED/PA intensity classification
Placement
1. To determine the cut-points for the GT3X+, non-dominant wrist-mounted acc. in children aged 8-12 y
2. To compare classification accuracies among the acc.’s three axes and VM values against direct observation (video recorded) / 45,
8-12 y / GT3X+
(non-dominant wrist) / Normal / NS
5 s / Not used / Cut-points:
Chandler SED, LPA, MPA, VPA VA, axis 2, axis 3 and VM cut-points / Results found comparable activity intensity classification accuracies from the GT3X+ wrist-worn acc. to previously published studies. Based on ROC and regression analyses, activity intensities can be distilled from this acc. using the 3 axes or VM values with similar classification accuracy
References / Study design
(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Crouter et al. [16] / Part 1:
Lab/controlled
Part 2:
Free living observational (unstructured PA session)
[Criterion validity: indirect calorimetry] / SED/PA intensity classification
Placement
To develop cut-points for classifying PA intensity through dominant wrist accelerations and validate the against indirect calorimetry in an unstructured PA session / 181,
12.0±1.5 y / GT3X/+
(dominant wrist) / LFE / 30 Hz.
5 s / Not used / Cut-points:
Crouter SED, LPA, MPA and VPA VA and VM cut-points determined by ROC curves and regression analyses / Compared to measured values, the VA and VM regression models developed on wrist acc. data had insignificant mean bias for child-METs and time spent in sedentary behavior, LPA, MPA and VPA; however, they had large individual errors
Crouter et al. [17] / Lab/controlled
[Criterion validity: indirect calorimetry] / EE algorithm
To examine the validity of seven child-specific equations compared with indirect calorimetry. / 72,
12±0.8 y / GT3X+
(right hip) / LFE / 30 Hz.
10 s / Not used / EE algorithm:
Crouter VM
Crouter VA
Evenson
Freedson
Trost
Treuth
Puyau / Crouter 2-regression model and Puyau were the most accurate equations
Crouter et al. [18] / Lab/controlled
[Criterion validity: indirect calorimetry] / EE algorithm
To develop two new two-regression models for use in children, that estimate energy expenditure (EE) using the ActiGraph GT3X mean VM counts or VA counts against indirect calorimetry / 59,
11.0±1.7 y / GT3X
(right hip) / LFE / 30 Hz.
10 s / Not used / EE algorithm:
VA and VM two regression-models developed / The new 2 regression models in children with the GT3X provide a closer estimate of mean measured MET than other currently available prediction equations. In addition, they improve the individual prediction errors across a wide range of activity intensities
References / Study design
(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Dowd et al. [19] / Lab/controlled
[Criterion validity: ActivPAL] / SED/PA intensity classification
1. To investigate the criterion validity, the concurrent validity, perform and validate a value calibration of the activPAL
2. To compare the estimating posture by activPAL with sedentary tresholds used with the GT3X / 30,
17.2±0.9 y / GT3X
(right hip)
ActivPAL / NS / 30 Hz.
15 s / Not used / Cut-points:
Ridgers SED VA cut-point / These findings suggest that the activPAL is a valid, objective measurement tool that can be used for both the measurement of PA and SED in an adolescent female population
Fairclough et al. [20] / Free-living observational
(7 days, waking hours protocol)
[Convergent validity: GT3X+ and GENEA] / Placement
SED/PA intensity classification
1. To explore children’s compliance to wearing wrist and hip-mounted acc.
2. To compare children’s PA derived from wrist and hip raw accelerations
3. To examine differences in raw and counts PA measured by hip-worn accelerometry / 109,
9-10 y. / GT3X+
(right hip)
GENEA
(Non-dominant wrist) / NS / 100 Hz.
1 s / ≥20-0-0
≥10 h/days
≥3 days/week / Not used / Wrist placement promotes superior compliance than hip. Raw accelerations were significantly higher for wrist compared to hip, possibly due to placement location and technical differences between devices. Hip PA calculated from raw accelerations and counts differed substantially.
References / Study design
(days of assessment, protocol)
[validation method] / Topics discussed
Aims / Subjects
N, mean±SD or age range / Acc. model
(placement) / Filter / Sampling frequency
Epoch / NWT definitiona
Valid day
Valid week / Cut-Points/
EE algorithm/
MET equation/
Sleep Algorithm / Main findings/Conclusions
Flynn et al. [3] / Part 1
Lab/controlled
Part 2
Lab/controlled
Part 3
Free-living observational
(1 school-day -8:30 a.m. to 4:00 p.m.)
[Calibration study: reliability] / Others: light sensor
1. To assess reliability of the GT3X+ ambient light sensor
2. To identify a lux threshold to accurately discriminate between indoor and outdoor activities in children
3. To test the accuracy of the lux threshold in a free-living environment / Part 1
20 GT3X+
Part 2
18,
7.0±2.3 y
Part 3
18,
4.4±0.4. y / GT3X+
(right hip) / NS / 60 Hz.
60 s (only for part 3) / Part 1:
Not used
Part 2:
Not used
Part 3:
Not used
≥4 h/day
NS / Not used / In part 1, there was high interinstrument reliability for the light sensor.
In part 2, the optimal lux threshold was determined to be 240 lux.
In part 3, results of the school-day validation demonstrated the monitor was 97% accurate for overall detection of indoor and outdoor conditions.
Hänggi et al. [21] / Lab/controlled
[Criterion validity: indirect calorimetry] / SED/PA intensity classification