Skip to main content
NCBI home page
Preventive Medicine Reports logoLink to Preventive Medicine Reports
. 2020 Jun 8;19:101142. doi: 10.1016/j.pmedr.2020.101142

Calibration and validation of accelerometry using cut-points to assess physical activity in paediatric clinical groups: A systematic review

Mayara S Bianchim a,1, Melitta A McNarry a,⁎,1, Lillebeth Larun b, Alan R Barker c,1, Craig A Williams c,1, Kelly A Mackintosh a,1
PMCID: PMC7327836  PMID: 32637301

Highlights

  • •

    Five studies generated physical activity cut-points across seven conditions in youth. Recommendations for future studies calibrating accelerometry in paediatric clinical are:

  • •

    To account for the pathophysiology of the disease in the calibration protocol.

  • •

    To integrate a measurement of energy expenditure to the calibration protocol.

  • •

    To move towards using machine learning techniques.

  • •

    To include a control group.

  • •

    To cross-validate the cut-points.

Abstract

Regular physical activity is associated with physiological and psychosocial benefits in both healthy and clinical populations. However, little is known about tailoring the analysis of physical activity using accelerometers to the specific characteristics of chronic conditions. Whilst accelerometry is broadly used to assess physical activity, recommendations on calibration in paediatric clinical groups are warranted. The aim of this systematic review was to provide a critical overview of protocols used to calibrate accelerometry in children and adolescents with clinical conditions, as well as to develop recommendations for calibration and validation of accelerometry in such populations. The search was performed between March to July 2017 using text words and subject headings in six databases. Studies had to develop moderate-to-vigorous intensity physical activity (MVPA) cut-points for paediatric clinical populations to be included. Risk of bias was assessed using a specific checklist. A total of 540,630 titles were identified, with 323 full-text articles assessed. Five studies involving 347 participants aged 9 to 15 years were included. Twenty-four MVPA cut-points were reported across seven clinical conditions, 16 of which were developed for different models of ActiGraph, seven for Actical and one for Tritrac-R3D. Statistical approaches included mixed regression, machine learning and receiver operating characteristic analyses. Disease-specific MVPA cut-points ranged from 152 to 735 counts·15 s−1, with lower cut-points found for inherited muscle disease and higher cut-points associated with intellectual disabilities. The lower MVPA cut-points for diseases characterised by both ambulatory and metabolic impairments likely reflect the higher energetic demands associated with those conditions.

1. Introduction

Regular physical activity (PA) is recommended for children and adolescents to promote health and well-being (World Health Organisation, 2015), irrespective of disease status. However, PA plays a particularly potent role in youth with chronic conditions and is associated with slowing disease progression in conditions such as cerebral palsy (CP; Keawutan et al., 2017, Verschuren et al., 2016). A common issue for children and adolescents with chronic conditions is the tendency to become less physically active with age and disease progression, which can lead to deconditioning and the initiation of a vicious negative spiral involving subsequent reductions in the ability to engage in PA (Durstine et al., 2013, Torpy et al., 2018).

Careful consideration should be given when recommending PA to children and adolescents with some chronic conditions due to the enhanced nutritional, metabolic and energetic requirements associated with the condition or structural disability (West et al., 2019). Children and adolescents with chronic conditions would, therefore, benefit from a greater understanding of the dose–response relationship between PA and health benefits in order to balance this with the potential negative sequalae that could ensue (Riner and Sellhorst, 2013). However, the current recommendation that children aged 5 to 18 years should accumulate, on average, at least 60 min of moderate-to-vigorous physical activity (MVPA) per day across the week (Department of Health and Social Care, 2019) has been developed for non-clinical populations and are therefore likely to have limited applicability to clinical populations. Indeed, a specific clinical guideline would warrant a higher degree of specificity and a cautious assessment of particular risks and benefits for each condition. It is therefore imperative to account for condition-specific factors that could be associated with exercise intolerance and/or an altered physiological response to exercise/PA (Wells et al., 2019).

PA recommendations tailored for children and adolescents with clinical conditions, however, remain sparse (Morris, 2008).

Objective methods used to assess PA, such as accelerometers, are appropriate for clinical settings due to the low participant burden and relatively low cost (Trost and O'Neil, 2014). Accelerometers are capable of detecting patterns of PA accumulation, as well as information on PA frequency and intensity, such as sedentary time (SED), light physical activity (LPA) and MVPA (Welk, 2005). Specifically, accelerometry measures velocity over a specific period of time, which can be translated into intensities of PA by using cut-points (Welk, 2005). However, the generation of these cut-points is highly challenging, for example, even within one type of accelerometer, the MVPA cut-point in healthy youth varies from 400 to 3,600 counts.min−1 (Cain et al., 2013). Whilst the accurate assessment of PA levels is particularly important in chronic conditions, inaccurate cut-points can result in over- or under-estimated predictions. Additionally, it is also important to consider the limitations associated with the use of accelerometry. For example, while accelerometry can accurately assess SED, it is not able to differentiate between various sedentary activities (Hurter et al., 2018). Moreover, factors such as brand and placement are likely to have an impact on the prediction of both SED and time spent in different PA intensities (Godfrey et al., 2008).

Amongst the challenges of calibrating accelerometry are the different methods to translate (e.g., PA protocols and criterion method) and interpret (e.g., statistical approach) the accelerometer raw signals into biological and behavioural outcomes (e.g., cut-points). Indeed, a recent systematic review summarising different accelerometry calibration studies in healthy populations acknowledged the lack of cut-points that account for individual characteristics, such as demographic and physiological variations (de Almeida Mendes et al., 2018). A key limitation of generalising cut-points developed for healthy populations to clinical populations is that they will not consider the altered resting metabolic rate (RMR) and higher energy expenditure (EE) for a given activity often evident in youth with chronic conditions (Bandini et al., 1991, Epstein et al., 1989, Ramsey et al., 1992). Whilst some research has sought to calibrate accelerometry in paediatric clinical conditions (Stephens et al., 2016, Trost et al., 2015), the lack of standardisation, wide variability in protocol designs and lack of healthy matched controls limits interpretation (Logan et al., 2016). Indeed, this systematic review can contribute by providing recommendations regarding the most appropriate criterion references, types of activities and statistical analyses to calibrate and cross-validate the cut-points.

The aim of this systematic review was to provide a critical overview of the protocols used to calibrate and validate accelerometry-derived MVPA cut-points in children and adolescents with clinical conditions and identify key parameters and considerations for future research.

2. Methods

This review was performed in accordance with the Preferred Reporting items for Systematic Review and Meta-Analysis statement (Liberati et al., 2009, Moher et al., 2015) and is registered on the International Prospective Register of Systematic Review (PROSPERO registration ID: CRD42016053880).

2.1. Search methods

The search was performed between March and July of 2017 using six databases (PubMed, SPORTDiscus, ScienceDirect, Scopus, ISI Web of Knowledge, Wiley Online Library). A Population Intervention Comparison Outcome (PICO) framework was adopted to build and structure the search; a detailed description of the search protocol is available on the web-appendix. The protocol and search strategy were reviewed by an experienced librarian and a pilot was performed to ensure the suitability of the criteria and search terms. The search terms were in accordance with the 2017 Medical Subject Headings and were inserted as keywords to all the databases and platforms. The search terms were: acceleromet*; acceleromet* AND (validation OR calibration); acceleromet* AND physical activity; wearable monitors AND (calibration OR validation); physical activity AND (calibration OR validation); acceleromet* cut-points; acceleromet* cut-points; energy expenditure AND acceleromet*; and classification AND physical activity intensities. The reference lists of relevant reviews and of all the studies included therein were examined for studies matching the inclusion criteria.

2.2. Eligibility criteria

Studies published in English from the year 2000 which generated MVPA accelerometry cut-points for accelerometry in children and adolescents (5 to 18 years) with any chronic clinical condition (disease of long duration and slow progression; Goodman et al., 2013) were included. Only studies published after the year 2000 were included in order to avoid inclusion of outdated accelerometers. Non-English, non-human and unpublished studies, book chapters, theses, monographs, dissertations and abstracts were not included. Studies in adults, or calibrating for healthy populations, sedentary behaviour or wheelchair users were excluded. Thus, studies using accelerometers along with additional technologies such as a microcontroller were not included.

2.3. Data extraction and management

An EndNote X7 (Clarivate Analytics, US) database was created with potential studies, and the lead author screened all the titles and abstracts. All full-texts selected by the first author (MSB) were screened by two co-authors (MAM and KAM) according to the pre-established inclusion criteria. Supplementary information for each study was consulted when available. In the case of missing information or variables required for completion of the extraction sheet, study authors were contacted, however, no additional data was provided. Data was extracted from the included full-texts by MSB and reviewed by KAM and MAM (Table 1). Any discrepancies were discussed by the three authors until a consensus was reached.

Table 1.

Summary of the data extracted from the included studies.

Data extraction field Information extracted
Context and participants The author, year and sample size of the study; participant characteristics such as age, health status, height, weight, BMI, ethnicity; and covariates measured such as self-report questionnaire data and health scales related to disease assessments were extracted.
Study design and methods used Any information related to the accelerometer, such as accelerometer model (e.g., number of axes); accelerometer placement (e.g., wrist [dominant/non-dominant], hip, chest); accelerometer settings (e.g., epoch, sampling frequency, use of low frequency filter); and data processing decisions (e.g., wear-time criteria) were extracted. Additionally, any information related to the calibration protocol, such as protocol design (e.g., laboratory-based, field-based, daily-life protocol); duration of the protocol; adjustment of specific variables (e.g., age, body mass); performance of individual calibration; criterion measure (e.g., energy expenditure, direct observation, heart rate); resting metabolic rate assessment; statistical approach (e.g., ROC-curve analyses, linear regression, machine learning); validation method (e.g., validation, cross-validation leave-one-out, cross-validation k-fold); and assessment for agreement (e.g., Kappa, Bland-Altman) were also extracted.
Findings The extracted outcomes were protocol design and cut-points. All the extracted protocols were classified in four categories: laboratory-based (walking or running, over-ground or on a treadmill), free-living (assessment of participant routine), daily-life (daily-life activities performed at the research site), and mixed (at least two of laboratory-based, free-living and daily-life) protocols.
Quality of the study checklist sheet.
Open in a new tab

The risk of bias was assessed independently by MSB and MAM using a specific checklist (Table 2) created according to previous recommendations for calibration protocols (Bassett et al., 2012, Freedson et al., 2005, Welk, 2005). This checklist considers six elements of the calibration protocol (sample characteristics, accelerometry settings, criterion measure, statistical approach for calibration, and statistical approach for validation) to rate studies as good, fair or poor according to the criteria described in Table 2. The inter-rater reliability was calculated using Kappa scores with 0.8 as the minimum acceptable inter-rater agreement (McHugh, 2012). Where any discrepancies arose following the risk assessment, all three authors involved in the screening and data extraction (MSB, MAM and KAM) discussed these until a consensus was reached.

Table 2.

Quality and risk assessment criteria according to descriptive variables and study design.

Standard Poor Fair Good
1. Sample Characteristics Study did not include any descriptive variables other than age and sex. Study included height, weight, body mass index and variables specific to the clinical condition. Study included height, weight, body mass index, ethnicity, resting metabolic rate, maturity stages and variables specific to the clinical condition.
2. Accelerometry Settings Study described accelerometer model. Study included accelerometer model, number of axes and placement position. Study included accelerometer model, number of axes, placement, sampling frequency, epoch length and any filtering techniques.
3. Protocol Design Calibration protocol composed by walking or treadmill test. Calibration used a mixed protocol (daily-life activities and a treadmill test). Mixed protocol combining daily-life activities, laboratory protocol test on a treadmill and free-living assessments.
4. Criterion Speed or direct observation. Heart rate or metabolic equivalent. Energy expenditure (including resting metabolic rate estimation*).
5. Statistical Approach for Calibration Linear regression or Individual linear regression. ROC curve analyses. Machine learning techniques, hierarchical models or multilevel modelling, adjusting for factors related to participants characteristics and to the pathophysiology of the clinical condition to develop the cut-point.
6. Statistical Approach for Validation No validation assessment. Leave-one-out cross-validation and agreement assessment using Bland-Altman or kappa score. K-fold cross-validation using different samples and activities. Agreement assessment using Bland-Altman or Kapa score, and estimates the intraclass correlation coefficient, and/or limits of agreement.
Open in a new tab

ROC: receiver operating characteristic. *The criteria for a valid resting metabolic rate estimation was a minimum of 15 min of steady state, preferably adopting the formula of Weir (1949).

A narrative synthesis of the studies was performed due to the heterogeneity of calibration protocols encountered, covering the topics of the protocol design, description of, and adjustment for, disease-specific factors, accelerometry model and settings, criterion measure and the statistical approach for generating and validating the cut-points. All cut-points in counts·min−1 were reintegrated to counts·15 s−1 epochs, which is commonly used in youth, to allow inter-study comparability.

3. Results

A total of 543,741 titles were found across all databases, with 540,630 titles remaining following the removal of duplicates. Following initial screening, 619 articles were selected by the main author for full-text assessment. In total, 614 studies were subsequently excluded, primarily due to being in a healthy population (279 studies; Fig. 1). A list of all full-text studies that were excluded can be obtained from the correspondent author. Five studies (Clanchy et al., 2011, McGarty et al., 2016, Ryan et al., 2014, Stephens et al., 2016, Trost et al., 2015), including 347, 9–15 year old, participants, with a total of 24 generated MVPA cut-points for seven clinical conditions, were included in the final synthesis. The clinical conditions were: CP, intellectual disabilities, CF, congenital heart diseases (CHD), haemophilia (HE), inherited muscle disease (IMD), juvenile idiopathic arthritis (JIA; Table 3).

Fig. 1.

Fig. 1

Open in a new tab

PRISMA flow chart presenting the systematic literature search.

Table 3.

Summary of included studies calibrating accelerometry in paediatric clinical groups.

Studies Participants Accelerometer Calibration Protocol Statistical Approach Outcome
Author, year Sample size (n)
Health status
Control Group
Sex (boy/girl)
Age (range or mean ± SD)
Height (mean ± SD)
Weight (range or mean ± SD)
BMI (range or mean ± SD)
Ethnicity
Covariates		 Device Model
Number of axes
Placement
Sampling frequency
Filter
Epoch
Sampling duration
Wear time		 Physiological/
Observational
EE estimation
RMR estimation
Individual calibration
Protocol type
Duration		 Calibration
Validation
Agreement		 Cut-Points/
Equation
Trost et al., 2015 n = 51
Cerebral Palsy
GMFCS I (27)
GMFCS II (12)
GMFCS III (12)
Control: 0
28 girls
12 ± 3 years
147.0 ± 16.5 cm
46.8 ± 19.0 kg
GMFCS		 ActiGraph GT3X
Tri-axial
Right hip
30 Hz
Epoch: 1 s		 Physiological: VO2
Resting VO2: Schofield
Individual calibration: no
Protocol type: Mixed – daily-life and walking
Duration: 120 min		 Calibration: Binary DT
Validation: LOOCV
Agreement: Kappa and ROC		 Cut-points (counts·15 s−1)
All levels:
LPA: < 72
GMFCS I
MVPA: 724
GMFSC II
MVPA: 685
GMFCS III
MVPA: 669
Ryan et al., 2014 n = 18
Cerebral Palsy
Control: no
11.4 ± 3.2 years
147.0 ± 18.5 cm
44.6 ± 16.9 kg
BMI: 20 ± 4.5 kg·m−2
GMFCS		 RT3
Right hip
Epoch: 60 s		 Physiological: VO2
RMR: Oxford equation
Individual calibration: none
Protocol type: laboratory
Duration: 36 min		 Calibration:
ROC curve
Validation: none
Agreement: Kappa score		 Cut-points (counts·min−1):
LPA: 52
MVPA: 689.3
Clanchy et al., 2011 n = 29
Cerebral palsy
Control: no
13 girls
12.5 ± 2.0 years
156.6 ± 11.0 cm
47.7 ± 16.1 kg
GMFCS		 ActiGraph (7164)
Uniaxial
Least affected hip
10 Hz
Epoch: 1 s		 Physiological: VO2
RMR: Schofield equation
Individual calibration: none
Protocol type: laboratory
Duration: 60 min		 Calibration:
ROC curve
Validation: none
Agreement: none		 Cut-points (counts·min−1):
LPA: 1627.3
MVPA: 2942.1
VIG: 4683.6
McGarty et al., 2016 n = 50
Validation: 36
Intellectual disabilities
Control: no
37 girls
9.54 ± 1.09 years
143 ± 0.9 cm
39.33 ± 10.28 kg
BMI: 19.9 ± 3.8 kg·m−2 		ActiGraph Wgt3X+
Tri-axial
Right hip
30 Hz
Epoch: 10 s		 Physiological: Direct Observation
Individual calibration: none
Protocol type: Daily-life
Duration: 45 min		 Calibration: ROC
Validation: LOOCV
Agreement: Kappa score		 Cut-points (counts·min−1):
VA:
SED: 507
MPA: 1008–2300
VPA: 2301
MVPA: 1008
VM:
SED: 1863
MPA: 2610–4214
VPA: 4215
MVPA: 2610
Stephens et al., 2016 n = 195
Control: n = 29
13 girls
13.1 ± 2.8 years
162 ± 16 cm
57.6 ± 20 kg
Skinfold: 38 ± 17
Tanner stages: 30% (stages 1–2), 70% (stage 3)
CHAQ: 0.15 ± 0,26
PedsQL: 83 ± 9
Cystic fibrosis (n = 32)
14 girls
12.8 ± 2.9 years
156 ± 16 cm
45 ± 14 kg
Skinfold: 31 ± 13
Tanner stage: 19% (stages 1–2), 81% (stage 3)
CHAQ: 0.27 ± 0.3
PedsQL: 78 ± 12
Congenital heart disease (n = 15)
5 girls
13.6 ± 3.3 years
161 ± 17 cm
54 ± 17 kg
Skinfold: 42 ± 15.5
Tanner Stage: 38% (stages 1–2), 62% (stage 3)
CHAQ: 0.17 ± 0.3
PedsQL: 72 ± 12
Haemophilia (n = 28)
0 girls
12.4 ± 3.3 years
156 ± 19 cm
53 ± 20.7 kg
Skinfold: 40 ± 20
Tanner Stage: 27% (stages 1–2), 73% (stage 3)
CHAQ: 0.25 ± 0.4
PedsQL: 82 ± 16
Idiopathic muscular dystrophies (n = 30)
8 girls
12 ± 3.4 years
146 ± 22 cm
41 ± 14 kg
Skinfold: 41 ± 18
Tanner stage: 70% (stages 1–2) 30% (stage 3)
CHAQ: 0.8 ± 0.7
PedsQL: 68 ± 17
Juvenile dermatomyositis (n = 31)
20 girls
13.4 ± 2.3 years
159 ± 11 cm
52 ± 14 kg
Skinfold: 48 ± 17
Tanner stage: 27% (stages 1–2), 73% (stage 3)
CHAQ: 0.4 ± 0.6
PedsQL: 77 ± 15
Juvenile arthritis (n = 31)
23 girls
12.7 ± 2.6 years
154 ± 12 cm
47 ± 14 kg
Skinfold: 46 ± 22
Tanner stage: 32 (stages 1–2), 68% (stage 3)
CHAQ: 0.5 ± 0.5
PedQL: 72 ± 13		 ActiGraph (7164) and Actical
Uniaxial
Right hip
10 HZ / 32 Hz
Epoch: 15 s		 Physiological: VO2 and HR
RMR: 2 h fasting – 20 min in rest
Individual calibration: no
Protocol type: Mixed: laboratory and daily-life
Duration: 240 min		 Calibration: Mixed regression models for equation, ROC curve for cut-points.
Validation: LOOCV
Agreement: none		 Chronic disease (combined) – ActiGraph
SED: 10
LPA: 10–426
MVPA: 426–785
Chronic disease (combined) – Actical
SED: 10
LPA: 17–288
MVPA: 289–570
Cystic fibrosis - ActiGraph
SED: 10
LPA: 10–487
MVPA: 487–852
Cystic fibrosis - Actical
SED: 5
LPA: 5–368
MVPA: 368–1025
Congenital heart disease - ActiGraph
SED: 10
LPA: 10–349
MVPA: 349–785
Congenital heart disease - Actical
SED: 9
LPA: 9–349
MVPA: 349–633
Haemophilia - ActiGraph
SED: 17
LPA: 17–432
MVPA: 432–788
Haemophilia - Actical
SED: 19
LPA: 19–306
MVPA: 306–1114
Inherited muscle disease - ActiGraph
SED: 37
LPA: 37–663
MVPA: 663–972
Inherited muscle disease - Actical
SED: 14
LPA: 14–297
MVPA: 297–523
Juvenile dermatomyositis- ActiGraph
SED: 14
LPA: 14–172
MVPA: 172–543
Juvenile dermatomyositis - Actical
SED: 18
LPA: 10–166
MVPA: 166–601
Juvenile arthritis - ActiGraph
SED: 25
LPA: 25–255
MVPA: 255–771
Juvenile arthritis - Actical
SED: 19
LPA: 19–152
MVPA: 152–542
Open in a new tab

SD: standard deviation; BMI: body mass index; EE: energy expenditure; RMR: resting metabolic rate; GMFCS: gross motor function classification system; VO2: oxygen uptake, LOOV: leave-one-out cross-validation; ROC: receiver operating characteristic; SED: sedentary time; LPA: light physical activity; MVPA: moderate-to-vigorous physical activity; VIG: vigorous activity; CHAQ: childhood health assessment questionnaire; PedsQL: pediatric quality of life inventory.

The inter-rater Kappa score for risk of bias was 0.80, with authors disagreeing regarding ‘accelerometry settings’, and were resolved after MSB and MAM discussed each point, resulting in a Kappa score of 1. Most studies (n = 4) were classified as fair for sample characteristics, with only one study scoring as good. One study scored as fair, and four as good, for accelerometry settings, with three and two studies classified as fair and good, respectively, for protocol design. For criterion measure, one scored as good, three as fair and one as poor. The majority (n = 4) of the studies scored as fair for statistical approach for calibration, with only one scoring as good. Finally, regarding the statistical approach for validation, three studies scored as fair and two as poor (Table 4).

Table 4.

Checklist Risk of Bias Assessment Results.

Study Sample Characteristics Accelerometry Settings Protocol Design Criterion Statistical Approach for Calibrations Statistical Approach for Validations
Clanchy et al., 2011 Fair Fair Poor Fair Fair Poor
Ryan et al., 2014 Fair Good Fair Poor Fair Fair
Trost et al., 2015 Fair Good Fair Fair Good Fair
McGarty et al., 2016 Fair Good Poor Fair Fair Poor
Stephens et al., 2016 Good Good Fair Good Fair Fair
Open in a new tab

Quality of life (Varni et al., 2004), maturity status (Emmanuel and Bokor, 2017, Stephens et al., 2016) and a generic health questionnaire (Feldman et al., 1995, Huber et al., 2001) were used as co-variates. Additionally, three studies (Clanchy et al., 2011, Ryan et al., 2014, Trost et al., 2015) used the specific classification system for CP (Gross Motor Function Classification System - GMFCS). Whilst covariates were considered by most of the included studies, only one study (Stephens et al., 2016) adjusted for disease-specific factors when generating the cut-points, although no formal description was provided on the variables included in the model. None of the studies investigated whether the disease-specific factors and participant demographics impacted on the developed cut-points.

3.1. Accelerometers

Sixteen of the included MVPA cut-points were developed for different ActiGraph models (McGarty et al., 2016, Ryan et al., 2014, Stephens et al., 2016, Trost et al., 2015), seven for Actical (Stephens et al., 2016) and one for Tritrac-R3D (RT3; Table 5; (Ryan et al., 2014). This translates to 15 MVPA cut-points derived from the vertical axis (VA) (Clanchy et al., 2011, Stephens et al., 2016) and nine from the vector magnitude (VM; (McGarty et al., 2016, Ryan et al., 2014, Trost et al., 2015). Three studies utilised hip-worn accelerometry on the right side (McGarty et al., 2016, Stephens et al., 2016, Trost et al., 2015) and two studies calibrating for CP placed the accelerometer on the least affected side (Clanchy et al., 2011, Ryan et al., 2014). The sample frequency varied between 1 and 32 Hz, with one study (Clanchy et al., 2011) not specifying this information. Two studies used an epoch of 15 s (Stephens et al., 2016, Trost et al., 2015), with others using 1 s (Clanchy et al., 2011), 10 s (McGarty et al., 2016) and 60 s (Ryan et al., 2014).

Table 5.

Summary of the accelerometer models used by the included studies.

Name / Model Manufacturer Weight and Size Memory Capacity Axis Frequency Sampling
ActiGraph 7164 (CSA) ActiGraph LLC Pensacola, FL 45.5 g
5.1 × 4.1 × 1.5 cm		 22 days of data with 60 s epoch Uniaxial 10 Hz
ActiGraph GT3X ActiGraph LLC Pensacola, FL 27 g
3.8 × 3.7 × 1.8 cm		 378 days using 60 s epoch Tri-axial 30 Hz
ActiGraph wGT3X+ ActiGraph LLC Pensacola, FL 19 g
4.6 × 3.3 1.5 cm		 38 days 100 Hz Tri-axial 30–100 Hz
Actical Mini-Mitter Sunriver, OR 17.5 g
2.8 × 2.7 × 1.0 cm		 45d using 60 s epoch Uniaxial 32 Hz
Research Tracker accelerometer (RT3) StayHealthy, Inc; Monrovia, California 71.5 g
71 × 56 × 28 mm		 30 days Tri-axial 0.017–1 Hz
Open in a new tab

3.2. Calibration protocol settings

A daily-life calibration protocol was the most commonly used (n = 3), generating 22 MVPA cut-points, with only two studies utilising a laboratory-based protocol (Clanchy et al., 2011, Ryan et al., 2014). Indirect calorimetry was the most common physiological criterion used for calibration (Clanchy et al., 2011, Ryan et al., 2014, Stephens et al., 2016, Trost et al., 2015), with one study using direct observation (McGarty et al., 2016). The protocol duration varied from 35 to 240 min. Resting metabolic rate was estimated by Stephens et al. (2016) using the Weir equation, whereas Clanchy et al., 2011, Trost et al., 2015 used the Schofield equation and Ryan et al. (2014) the Oxford equation. As McGarty et al. (2016) developed cut-points through direct observation, a RMR estimation was not required. All included studies performed a group calibration rather than individual calibrations.

3.3. Statistical approach

Fourteen MVPA cut-points were developed through mixed regression models (Stephens et al., 2016), six using machine learning (regressing trees; Trost et al., 2015), and four through Receiver Operating Characteristic (ROC) analysis (Clanchy et al., 2011, McGarty et al., 2016, Ryan et al., 2014, Stephens et al., 2016). Only one study did not perform any kind of validation (Clanchy et al., 2011), with all other validations performed using leave-one-out cross-validations. No studies utilised independent samples or a different set of activities to cross-validate. Eighteen (Clanchy et al., 2011, Ryan et al., 2014, Stephens et al., 2016, Trost et al., 2015) of the generated cut-points were validated through comparison of previously established cut-points developed for healthy populations (Evenson et al., 2006, Puyau et al., 2002, Rowlands et al., 2004, Vanhelst et al., 2010). Three studies (McGarty et al., 2016, Ryan et al., 2014, Trost et al., 2015) utilised the Kappa score for agreement assessment, whereas two studies (Clanchy et al., 2011, Stephens et al., 2016) performed ANOVA.

3.4. Outcome

The disease-specific MVPA cut-points ranged from 152 to 735 counts·15 s−1, with 19 MVPA cut-points presented in counts·15 s−1, and four presented in counts·min−1 (Table 6). The sensitivity of the cut-points ranged from 37 to 91%, and the specificity ranged from 85 to 97%. Cerebral palsy was the mostly widely studied clinical condition, with eight cut-points developed across three studies (Clanchy et al., 2011, Ryan et al., 2014, Trost et al., 2015). Trost et al. (2015) generated cut-points for different degrees of CP severity, with fair to excellent accuracy, demonstrating better accuracy (lower rates of misclassification, particularly for GMFCS III and for LPA classification) than Evenson et al. (2006) cut-points. In contrast, Ryan et al., 2014, Clanchy et al., 2011 did not develop specific cut-points for different GMFCS levels or perform a leave-one-out cross validation, using specificity and sensitivity as a measure of validation. Clanchy et al. (2011) cut-points showed no significant improvement in PA classification accuracy compared to healthy population cut-points, whilst the MVPA cut-points of Ryan et al. (2014) demonstrated moderate classification agreement (Evenson et al., 2006, Rowlands et al., 2004, Vanhelst et al., 2010). Similarly, Stephens et al (2016) also applied healthy population cut-points (Evenson et al., 2006) to their participants with various chronic conditions (CF, IMD, JIA, HE and CHD), which resulted in poor-to-moderate sensitivity in PA classification. Most of the disease-specific cut-points developed were below the previously established MVPA cut-points for healthy populations (e.g., 2,020 to 8,199 counts·min−1).

Table 6.

Summary and validity of the clinical-specific moderate-to-vigorous cut-points.

Conditions (n) Study Reason for split Cut-points MVPA (original) Cut-points MVPA converted to counts·15 s−1 Criterion Validity
Cerebral palsy (7) Trost et al., 2015 GMFCS I / VA 535 (counts·15 s−1) N/A LOOCV – 81.1%
Trost et al., 2015 GMFCS II / VA 333 (counts·15 s−1) N/A LOOCV – 76.7%
Trost et al., 2015 GMFCS III/VA 200 (counts·15 s−1) N/A LOOCV – 82.9%
Trost et al., 2015 GMFCS I / VM 724 (counts·15 s−1) N/A LOOCV – 80.5%
Trost et al., 2015 GMFCS II / VM 685 (counts·15 s−1) N/A LOOCV – 75.6%
Trost et al., 2015 GMFCS III / VM 669 (counts·15 s−1) N/A LOOCV – 84.2%
Ryan et al., 2014 N/A 689.3 (counts·min−1) 172.3 Se – 86.7% / Sp – 91.9%
Clanchy et al., 2011 N/A 2942 (counts·min−1) 735.5 Se – 91.4% / Sp – 86.2%
Intellectual disability (2) McGarty et al., 2016 VA 1008 (counts·min−1) 252 LOOCV – 93%
Se – 91% / Sp – 95%
McGarty et al., 2016 VM 2610 (counts·min−1) 652 LOOCV – 87%
Se – 91% / 84%
Cystic fibrosis (2) Stephens et al., 2016 CF / ActiGraph 7164 487 (counts·15 s−1) N/A Se – 71% / Sp – 85%
Stephens et al., 2016 CF / Actical 7164 368 (counts·15 s−1) N/A Se – 51% / Sp – 91%
Chronic heart disease (2) Stephens et al., 2016 CHD / ActiGraph 7164 349 (counts·15 s−1) N/A Se – 42% / Sp – 85%
Stephens et al., 2016 CHD / Actical 349 (counts·15 s−1) N/A Se – 41 / Sp – 94%
Inherited muscle disease (2) Stephens et al., 2016 IMD / ActiGraph 7164 663 (counts·15 s−1) N/A Se – 81% / Sp – 90%
Stephens et al., 2016 IMD / Actical 297 (counts·15 s−1) N/A Se – 47% / Sp – 96%
Juvenile dermatomyositis (2) Stephens et al., 2016 JDM /ActiGraph 7164 172 (counts·15 s−1) N/A Se – 41% / Sp – 90%
Stephens et al., 2016 JDM / Actical 166 (counts·15 s−1) N/A Se – 37% / Sp – 94%
Haemophilia (2) Stephens et al., 2016 HE / Actical 306 (counts·15 s−1) N/A Se – 49% / Sp – 92%
Stephens et al., 2016 HE / ActiGraph 7164 432 (counts·15 s−1) N/A Se – 53% / Sp – 92%
Juvenile arthritis (2) Stephens et al., 2016 JIA / Actical 152 (counts·15 s−1) N/A Se – 49% / Sp – 94%
Stephens et al., 2016 JIA / ActiGraph 7164 255 (counts·15 s−1) N/A Se – 41% / Sp – 90%
Overall (CF, JA, HE, CHD, JDM, IMD) (2) Stephens et al., 2016 Overall Diseases / Actical 289 (counts·15 s−1) N/A Se – 77% / Se – 97%
Stephens et al., 2016 Overall Diseases / ActiGraph 7164 426 (counts·15 s−1) N/A Se – 78% / Sp – 94%
Open in a new tab

MVPA: moderate-to-vigorous physical activity; GMFCS: gross motor function classification system; VA: vector axial, VM: vector magnitude; LOOCV: leave-one-out cross-validation; Se: sensitivity; Sp: specificity; CF: cystic fibrosis; CHD: congenital heart disease; IMD: inherited muscle disease; JMD: juvenile dermatomyositis; HE: haemophilia; JA: juvenile arthritis.

4. Discussion

Twenty-four MVPA cut-points were extracted from five studies across seven different paediatric clinical groups. Overall, the review revealed little consensus with regards to MVPA cut-points, due to, at least in part, the relatively low number of calibration studies and broad range of protocol designs and accelerometer settings used in the studies, thereby limiting inter-study comparisons. Nonetheless, despite this, a thorough methodological quality assessment of the included studies was performed, which contributed to a higher transparency and aided the interpretation of the outcomes. Moreover, this review presented a critical analysis of the methodological challenges faced when developing cut-points for clinical paediatric populations, providing recommendations for future studies.

4.1. Calibration protocol for paediatric clinical populations

The majority of the included studies utilised daily-life (McGarty et al., 2016) or mixed (Stephens et al., 2016, Trost et al., 2015) protocols composed of daily-life and laboratory protocols. To accommodate different disease and disability levels, Stephens et al. (2016) adjusted their laboratory-based protocol by performing two different treadmill tests based on 6-min walking test performance. Whilst the protocol can greatly impact the PA classification, the physiological criterion adopted is equally important. For example, both Trost et al., 2015, Stephens et al., 2016 utilised indirect calorimetry as criterion, which therefore considers the higher energetic demand associated with a given activity in some chronic conditions (Walker et al., 2015). Specifically, diseases associated with chronic inflammation (e.g., CF, obesity) and musculoskeletal adaptations (e.g., CP, JIA, IMD) can reduce exercise tolerance, leading to chronic deconditioning and a higher EE demand for a given activity (Mehta, 2015).

It is well known that the majority of paediatric clinical conditions are associated with altered cardiometabolic demands (Bar-Or and Rowland, 2004). Thus, studies calibrating accelerometry for these populations should adopt EE as their criterion method. Another important consideration is that RMR changes dramatically according to maturity, disease and health parameters (McErlane et al., 2017), such as chronic inflammation and reductions in PA (Buchdahl et al., 1988, Eisenstein and Berkun, 2014). Specifically, individuals with CF often have a greater RMR, which can be explained to some extent by pulmonary impairment (Dorlochter et al., 2002) and increased cost of breathing (Bell et al., 1996, Frankenfield et al., 2017). Conversely, children with certain types of CP have a reduced RMR due to a lower energetic requirement at rest and altered body composition (e.g., reduced fat free mass and lean body mass; Bandini et al., 1995, Bandini et al., 1991, Stallings et al., 1993). Consequently, condition-specific calibration protocols adopting EE as the criterion should measure RMR. Despite using indirect calorimetry in their protocols, some of the included studies utilised Schofield and Oxford equations (Clanchy et al., 2011) to determine RMR. Whilst such equations may provide a low-cost estimation of RMR, they are based on chronological, rather than biological, age (McMurray et al., 2015), and do not account for sex or health status. This may lead to an inaccurate estimation of RMR, and consequently of EE, in clinical populations (De Wit et al., 2010, Fuster et al., 2007). Therefore, the measurement of oxygen uptake at rest should be utilised to provide a precise estimation of RMR, and consequently enhance the accuracy of the disease-specific cut-points in youth with chronic conditions (Stephens et al., 2016).

It is also important to consider the influence of disease severity within a condition, which is likely to affect the relative energetic demand, as might differences in the treatment and medication strategies between patients (Walker et al., 2015). Indeed, Ryan et al., 2014, Clanchy et al., 2011 did not stratify their sample by the GMFCS scale, resulting in large heterogeneity of CP-severity across participants, with some children not able to finish the protocol. In contrast, Trost et al. (2015) demonstrated that the relationship between EE and activity counts changed significantly according to GMFCS level, with children classified as level III having greater EE during locomotion when compared to levels I and II.

4.2. Statistical approach

The statistical approach chosen is highly influential in the translation of the physiological criterion into cut-points. Linear regression, which was initially one of the most commonly used methods for calibration, cannot account for the non-linear relationship between PA and EE (Freedson et al., 2005, Welk, 2005). Consequently, most of the studies included in this review utilised ROC analyses to develop their cut-points. Whilst ROC is more accurate than linear regression (Welk, 2005), it is dependent on the number of participants and does not allow adjustment of disease-specific factors (Staudenmayer et al., 2009).

Alternatively, mixed regression modelling is an exploratory analysis, particularly useful due to its flexible nature that allows the inclusion of disease-specific factors (Welk, 2005). Stephens et al. (2016) utilised mixed regression modelling to control for disease-specific factors to generate predictive equations for children and adolescents with CF, HE, JIA, CHD and IMD (Aadland and Steene-Johannessen, 2012, Lopes et al., 2009), reporting that heart rate improved the model and lowered the standard error associated with the prediction. These findings agree with those in healthy populations (Altini et al., 2014), with the improvements in standard error likely to be attributable to the reduction of the inter-individual variability caused by the adjustment of physiological signals. It is noteworthy that whilst a certain degree of accuracy can be achieved with cut-points, recent PA research has moved towards using machine learning. Indeed, more complex machine learning analysis have provided a higher degree of accuracy in comparison with traditional cut-points (Bonomi et al., 2009, Staudenmayer et al., 2015, Staudenmayer et al., 2009, Welk, 2005). Despite this, a calibration protocol is still required even when using those techniques. Indeed, machine learning can also be used to develop cut-points, for example, Trost et al. (2015) used Binary Decision Trees to generate CP-specific cut-points. Whilst machine learning provides high accuracy, evidence suggests that considerable bias can arise from using a small sample size (Combrisson and Jerbi, 2015). Alternatively, approaches such as using different testing and training data sets, and testing algorithm performance (i.e., nested cross-validation), can provide unbiased performance estimates even with small sample sizes (Vabalas et al., 2019).

A cross-validation analysis of the cut-points evaluates the predictive models to ensure validity and avoid over-fitting, and it can be performed through different methods such as the k-fold or leave-one-out cross-validation. Specifically, considering that the developed cut-points might be biased to the sample characteristics or to the calibration protocol design, the use of an independent sample with a different set of activities for cross-validating the cut-points is recommended (Welk, 2005). Stephens et al., 2016, Trost et al., 2015 applied a leave-one-out cross-validation, identified as the most appropriate approach when working with smaller samples (Welk et al., 2003), or to lessen the burden on the participants. It is further recommended that the disease-specific cut-points should also be validated against a healthy matched control group to ensure that potential cut-point discrepancies are a result of the pathophysiology rather than from the protocol design. Further to the cross-validation, agreement measures, such as Kappa score and Bland-Altman, indicate whether two methods can be used concomitantly or interchangeably, thereby facilitating inter-study comparisons (Bland and Altman, 1986). Alternatively, recent research has used a statistical equivalence test to measure agreement, which has been shown to be more appropriate for highlighting similarities between methods (Dixon et al., 2018, Kim et al., 2016). Particularly, the performance of agreement measures between activity counts and the criterion measures in a calibration protocol ensures that both measurements are comparable, avoiding further errors to the developed cut-points (Welk, 2005).

4.3. Outcome: cut-points

Cross-validation identified moderate to excellent accuracy for most of the disease-specific cut-points. Considerable inter-study discrepancies were found when comparisons were made between the disease-specific and previously established healthy population cut-points. For example, whilst Trost et al. (2015) found that applying cut-points developed for healthy populations (Evenson et al., 2006) to CP children resulted in poor accuracy and misclassification, Ryan et al., 2014, Clanchy et al., 2011 demonstrated fair to moderate accuracy (Rowlands et al., 2004, Vanhelst et al., 2010). Indeed, converse to Ryan et al., 2014, Clanchy et al., 2011, Trost et al., 2015 calibrated for each level of the GMFCS instead of performing an overall calibration, and applied machine learning techniques to generate the CP cut-points, presenting higher specificity than the cut-points developed for healthy populations. Furthermore, Stephens et al. (2016) also found that their disease-specific cut-points (CF, CHD, HE, JIA and IMD) had improved accuracy when compared with standard cut-points, thereby supporting the notion that specific cut-points are necessary for clinical populations.

Given that SED is mainly classified based on stationary activities and therefore does not consider musculoskeletal disabilities, it is unsurprising that some studies (Clanchy et al., 2011, Ryan et al., 2014, Trost et al., 2015) demonstrated fair to excellent accuracy when utilising healthy population-based SED cut-points for children with less severe CP. Despite this, poor classification of LPA may affect specific clinical populations, such as CP (Verschuren et al., 2014), who may not be able to engage in MVPA activities, and would therefore greatly benefit from a reduction in SED (Ryan et al., 2015). Specifically, considering that daily PA is a composite measure, an increase in LPA could be associated with a reduction in SED and enhancement on the total volume of PA (Bassett et al., 2017). Indeed, estimation of LPA for children with CP through standard cut-points, such as Evenson et al., 2006, Vanhelst et al., 2010, presented poor to fair classification accuracy (Clanchy et al., 2011, Ryan et al., 2014, Trost et al., 2015). Additionally, the lack of standardisation regarding protocol design and statistical approach hinders the applicability of the cut-points, which might explain the variability found between cut-points developed for the same clinical condition. Consequently, age- and sex-matched healthy control groups are essential to elucidate whether the differences observed in the disease-specific protocol are due to the disease severity or to protocol discrepancies. However, only one study (Stephens et al., 2016) included a control group although this was only used for baseline comparisons.

4.4. Strengths and limitations

The present systematic review is associated with numerous strengths. Firstly, an experienced librarian was consulted to revise the initial protocol and a pilot search was conducted to minimise errors, leading to changes in the eligibility of participants, outcomes, risk of bias assessment and analysis. Moreover, the initial search terms were adapted following advice from the librarian. The pilot search generated a large number of studies for participants across the lifespan and health continuum, therefore, the inclusion criteria for participants were limited to only children and adolescents with clinical conditions. Nevertheless, the literature was initially screened to capture all calibration studies for healthy and clinical populations. Whilst this strategy resulted in an extensive search, it also minimised the possibility of missing studies calibrating for a clinical condition. However, this strategy is not without limitations, as it required having only one author screen all the titles and abstracts. Nonetheless, different approaches were adopted to minimise error. Specifically, an EndNote library was created, and the same search strategy was used for all databases. Whilst double data entry was not performed, a data extraction sheet was created and checked by two co-authors (KAM, MAM), and subsequently made available to all authors during the extraction process.

A qualitative data synthesis was performed due the heterogeneity of calibration protocols and the calculation of cut-point effect sizes not being possible, thereby precluding a meta-analysis from being performed. The heterogeneity of the protocols can partially be explained by the inclusion of a broad range of clinical conditions. However, whilst the comparison of numerous clinical conditions of a different nature may be questioned, the primary aim of the review was to investigate the structure of different calibration protocols and how they accounted for the pathophysiology of the respective conditions. Despite the varying nature of the conditions included, only a small range of studies calibrated accelerometry in clinical populations, which hinders further conclusions regarding the optimal protocol.

5. Conclusion

Overall, this systematic review highlights the broad range of protocol designs and accelerometer settings of studies developing MVPA cut-points for children and adolescents with clinical conditions. Research seeking to develop disease-specific paediatric cut-points should consider the pathophysiology of the disease and seek to include a measure of EE, an accurately assessed RMR and a healthy comparison group. Moreover, all cut-points developed should be cross-validated. In summary, studies calibrating accelerometry in paediatric clinical populations are urgently required to establish an optimal calibration protocol. Subsequently, the enhancement in the assessment and surveillance of PA for clinical populations could lead to the development of more informed clinically specific PA guidelines.

Authors’ contributions

MSB made substantial contributions to conception, design, systematic search, data analysis and interpretation, and drafted of the manuscript. MAM and KAM made substantial contributions to conception, design, systematic search, data analysis and interpretation, manuscript writing and critically revised the manuscript for important intellectual content. LL supported the design of the search-protocol and critically revised the methodology and general content of the manuscript. AB and CW critically revised the manuscript for important intellectual content. All the authors approved the final manuscript.

Funding

This review summarises independent research funded by the Cystic Fibrosis Trust UK under its programme grant for Strategic Research Centres (grant reference No RP-PG-0108-10011). MSB is a funded PhD student by the Cystic Fibrosis Trust. The funder had no role in the conduct of the study, the writing of the manuscript, or the decision to submit it for publication.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

We would like to thank the librarian Philippa Price (Swansea University) for her advice on the initial protocol and input on the methods section.

Footnotes

Appendix A

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pmedr.2020.101142.

Contributor Information

Mayara S. Bianchim, Email: 890407@swansea.ac.uk.

Melitta A. McNarry, Email: M.McNarry@swansea.ac.uk.

Lillebeth Larun, Email: Lillebeth.Larun@fhi.no.

Alan R. Barker, Email: A.R.Barker@exeter.ac.uk.

Craig A. Williams, Email: C.A.Williams@exeter.ac.uk.

Kelly A. Mackintosh, Email: K.Mackintosh@swansea.ac.uk.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Supplementary Data 1
mmc1.docx (18.3KB, docx)

References

  1. Aadland E., Steene-Johannessen J. The use of individual cut points from treadmill walking to assess free-living moderate to vigorous physical activity in obese subjects by accelerometry: is it useful? BMC Med. Res. Methodol. 2012;12:172. doi: 10.1186/1471-2288-12-172. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Altini M., Penders J., Vullers R., Amft O. Personalizing energy expenditure estimation using physiological signals normalization during activities of daily living. Physiol. Meas. 2014;35:1797–1811. doi: 10.1088/0967-3334/35/9/1797. [DOI] [PubMed] [Google Scholar]
  3. Bandini L.G., Schoeller D.A., Fukagawa N.K., Wykes L.J., Dietz W.H. Body composition and energy expenditure in adolescents with cerebral palsy or myelodysplasia. Pediatr. Res. 1991;29:70–77. doi: 10.1203/00006450-199101000-00014. [DOI] [PubMed] [Google Scholar]
  4. Bandini L.G., Puelzl-Quinn H., Morelli J.A., Fukagawa N.K. Estimation of energy requirements in persons with severe central nervous system impairment. J. Pediatr. 1995;126:828–832. doi: 10.1016/s0022-3476(95)70423-x. [DOI] [PubMed] [Google Scholar]
  5. Bar-Or, O., Rowland, T.W., 2004. Pediatric Exercise Medicine: From Physiologic Principles to Health Care Application. Human Kinetics.
  6. Bassett D.R., Rowlands A.V., Trost S.G. Calibration and validation of wearable monitors. Med. Sci. Sports Exerc. 2012;44:S32–S38. doi: 10.1249/MSS.0b013e3182399cf7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bassett D.R., Toth L.P., LaMunion S.R., Crouter S.E. Step counting: a review of measurement considerations and health-related applications. Sports Med. 2017:1303–1315. doi: 10.1007/s40279-016-0663-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bell S.C., Saunders M.J., Elborn J.S., Shale D.J. Resting energy expenditure and oxygen cost of breathing in patients with cystic fibrosis. Thorax. 1996;51:126–131. doi: 10.1136/thx.51.2.126. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Bland J.M., Altman D.G. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–310. [PubMed] [Google Scholar]
  10. Bonomi A.G., Plasqui G., Goris A.H.C., Westerterp K.R. Improving assessment of daily energy expenditure by identifying types of physical activity with a single accelerometer. J. Appl. Physiol. 2009;107:655–661. doi: 10.1152/japplphysiol.00150.2009. [DOI] [PubMed] [Google Scholar]
  11. Buchdahl R.M., Cox M., Fulleylove C., Marchant J.L., Tomkins A.M., Brueton M.J., Warner J.O. Increased resting energy expenditure in cystic fibrosis. J. Appl. Physiol. 1988;1985(64):1810–1816. doi: 10.1152/jappl.1988.64.5.1810. [DOI] [PubMed] [Google Scholar]
  12. Cain K.L., Sallis J.F., Conway T.L., Van Dyck D., Calhoon L. Using accelerometers in youth physical activity studies: a review of methods. J. Phys. Act Health. 2013;10:437–450. doi: 10.1123/jpah.10.3.437. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Clanchy K.M., Tweedy S.M., Boyd R.N., Trost S.G. Validity of accelerometry in ambulatory children and adolescents with cerebral palsy. Eur. J. Appl. Physiol. 2011;111:2951–2959. doi: 10.1007/s00421-011-1915-2. [DOI] [PubMed] [Google Scholar]
  14. Combrisson E., Jerbi K. Exceeding chance level by chance: the caveat of theoretical chance levels in brain signal classification and statistical assessment of decoding accuracy. J. Neurosci. Methods. 2015;250:126–136. doi: 10.1016/j.jneumeth.2015.01.010. [DOI] [PubMed] [Google Scholar]
  15. de Almeida Mendes M., da Silva I.C.M., Ramires V.V., Reichert F.F., Martins R.C., Tomasi E. Calibration of raw accelerometer data to measure physical activity: a systematic review. Gait Posture. 2018;61:98–110. doi: 10.1016/j.gaitpost.2017.12.028. [DOI] [PubMed] [Google Scholar]
  16. De Wit B., Meyer R., Desai A., Macrae D., Pathan N. Challenge of predicting resting energy expenditure in children undergoing surgery for congenital heart disease. Pediatr. Crit. Care Med. 2010;11:496–501. doi: 10.1097/PCC.0b013e3181ce7465. [DOI] [PubMed] [Google Scholar]
  17. Department of Health and Social Care, 2019. Physical Activity Guidelines: UK Chief Medical Officers’ Report. Department of Health and Social Care; London, UK. [(accessed on 02 October 2019)]. Available online: https://www.gov.uk/government/publications/physical-activity-guidelines-uk-chief-medical-officers-report.
  18. Dixon P.M., Saint-Maurice P.F., Kim Y., Hibbing P., Bai Y., Welk G.J. A primer on the use of equivalence testing for evaluating measurement agreement. Med. Sci. Sports Exerc. 2018;50:837–845. doi: 10.1249/MSS.0000000000001481. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Dorlochter L., Roksund O., Helgheim V., Rosendahl K., Fluge G. Resting energy expenditure and lung disease in cystic fibrosis. J. Cyst. Fibros. 2002;1:131–136. doi: 10.1016/s1569-1993(02)00076-0. [DOI] [PubMed] [Google Scholar]
  20. Durstine J.L., Gordon B., Wang Z., Luo X. Chronic disease and the link to physical activity. J. Sport Health Sci. 2013;2:3–11. [Google Scholar]
  21. Eisenstein E.M., Berkun Y. Diagnosis and classification of juvenile idiopathic arthritis. J. Autoimmun. 2014;48–49:31–33. doi: 10.1016/j.jaut.2014.01.009. [DOI] [PubMed] [Google Scholar]
  22. Emmanuel, M., Bokor, B.R., 2017. Tanner Stages. [PubMed]
  23. Epstein L.H., Wing R.R., Cluss P., Fernstrom M.H., Penner B., Perkins K.A., Nudelman S., Marks B., Valoski Resting metabolic rate in lean and obese children: relationship to child and parent weight and percent-overweight change. Am. J. Clinical Nutrition. 1989;49:331–336. doi: 10.1093/ajcn/49.2.331. [DOI] [PubMed] [Google Scholar]
  24. Evenson, K.R., Catellier, D.J., Gill, K., Ondrak, K.S., McMurray, R.G., 2006. Calibration of two objective measures of physical activity for children, J. Sports Sci. Dec 2006, p. 1557. [DOI] [PubMed]
  25. Feldman B.M., Ayling-Campos A., Luy L., Stevens D., Silverman E.D., Laxer R.M. Measuring disability in juvenile dermatomyositis: validity of the childhood health assessment questionnaire. J. Rheumatol. 1995;22:326–331. [PubMed] [Google Scholar]
  26. Frankenfield D.C., Ashcraft C.M., Drasher T.L., Reid E.K., Vender R.L. Characteristics of resting metabolic rate in critically Ill, mechanically ventilated adults with cystic fibrosis. JPEN J. Parenteral Enteral Nutrition. 2017;41:601–606. doi: 10.1177/0148607115617152. [DOI] [PubMed] [Google Scholar]
  27. Freedson P., Pober D., Janz K.F. Calibration of accelerometer output for children. Med. Sci. Sports Exerc. 2005;37:S523–S530. doi: 10.1249/01.mss.0000185658.28284.ba. [DOI] [PubMed] [Google Scholar]
  28. Fuster C.O., Fuster G.O., Galindo A.D., Galo A.P., Verdugo J.M., Lozano F.M. Analysis of energy expenditure in adults with cystic fibrosis: comparison of indirect calorimetry and prediction equations. Arch. Bronconeumol. 2007;43:366–372. doi: 10.1016/s1579-2129(07)60087-1. [DOI] [PubMed] [Google Scholar]
  29. Godfrey, A., Conway, R., Meagher, D., G, O.L., 2008. Direct measurement of human movement by accelerometry. Med. Eng. Phys. 30, 1364–1386. [DOI] [PubMed]
  30. Goodman R.A., Posner S.F., Huang E.S., Parekh A.K., Koh H.K. Defining and measuring chronic conditions: imperatives for research, policy, program, and practice. Prev. Chronic Dis. 2013;10:E66. doi: 10.5888/pcd10.120239. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Huber A.M., Hicks J.E., Lachenbruch P.A., Perez M.D., Zemel L.S., Rennebohm R.M., Wallace C.A., Lindsley C.B., Passo M.H. Validation of the Childhood Health Assessment Questionnaire in the juvenile idiopathic myopathies. Juvenile Dermatomyositis Disease Activity Collaborative Study Group. J. Rheumatol. 2001;28:1106–1111. [PubMed] [Google Scholar]
  32. Hurter, L., Fairclough, S.J., Knowles, Z.R., Porcellato, L.A., Cooper-Ryan, A.M., Boddy, L.M., 2018. Establishing Raw Acceleration Thresholds to Classify Sedentary and Stationary Behaviour in Children. Children (Basel) 5. [DOI] [PMC free article] [PubMed]
  33. Keawutan, P., Bell, K.L., Oftedal, S., Ware, R.S., Stevenson, R.D., Davies, P.S.W., Boyd, R., 2017. Longitudinal Physical Activity and Sedentary Behaviour in Preschool-Aged Children with Cerebral Palsy Across All Functional Levels. Developmental medicine and child neurology, 59. [DOI] [PubMed]
  34. Kim Y., Crouter S.E., Lee J.M., Dixon P.M., Gaesser G.A., Welk G.J. Comparisons of prediction equations for estimating energy expenditure in youth. J. Sci. Med. Sport. 2016;19:35–40. doi: 10.1016/j.jsams.2014.10.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Liberati A., Altman D.G., Tetzlaff J., Mulrow C., Gotzsche P.C., Ioannidis J.P., Clarke M., Devereaux P.J., Kleijnen J. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLoS Med. 2009;6 doi: 10.1371/journal.pmed.1000100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Logan G.R., Duncan S., Harris N.K., Hinckson E.A., Schofield G. Adolescent physical activity levels: discrepancies with accelerometer data analysis. J. Sports Sci. 2016;34:2047–2053. doi: 10.1080/02640414.2016.1150599. [DOI] [PubMed] [Google Scholar]
  37. Lopes Vitor Pires, Magalhães Pedro, Bragada José, Vasques Catarina. Actigraph calibration in obese/overweight and type 2 diabetes mellitus middle-aged to old adult patients. J. Phys. Activity Health. 2009;6(s1):S133–S140. doi: 10.1123/jpah.6.s1.s133. [DOI] [PubMed] [Google Scholar]
  38. McErlane, F., Carrasco, R., Kearsley-Fleet, L., Baildam, E.M., Wedderburn, L.R., Foster, H.E., Ioannou, Y., Chieng, S.E.A., Davidson, J.E., et al., 2017. Growth patterns in early juvenile idiopathic arthritis: Results from the Childhood Arthritis Prospective Study (CAPS). Semin Arthritis Rheum. [DOI] [PMC free article] [PubMed]
  39. McGarty A.M., Penpraze V., Melville C.A. Calibration and Cross-Validation of the ActiGraph wGT3X+ Accelerometer for the Estimation of Physical Activity Intensity in Children with Intellectual Disabilities. PLoS ONE. 2016;11 doi: 10.1371/journal.pone.0164928. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. McHugh M.L. Interrater reliability: the kappa statistic. Biochem Med (Zagreb) 2012:276–282. [PMC free article] [PubMed] [Google Scholar]
  41. McMurray R.G., Butte N.F., Crouter S.E., Trost S.G., Pfeiffer K.A., Bassett D.R., Puyau M.R., Berrigan D., Watson K.B. Exploring metrics to express energy expenditure of physical activity in youth. PLoS ONE. 2015;10 doi: 10.1371/journal.pone.0130869. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Mehta N.M. Energy expenditure: how much does it matter in infant and pediatric chronic disorders? Pediatr. Res. 2015;77:168–172. doi: 10.1038/pr.2014.180. [DOI] [PubMed] [Google Scholar]
  43. Moher D., Shamseer L., Clarke M., Ghersi D., Liberati A., Petticrew M., Shekelle P., Stewart L.A. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Systematic Rev. 2015;4:1. doi: 10.1186/2046-4053-4-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Morris P.J. Physical activity recommendations for children and adolescents with chronic disease. Curr. Sports Med. Rep. 2008;7:353–358. doi: 10.1249/JSR.0b013e31818f0795. [DOI] [PubMed] [Google Scholar]
  45. Puyau M.R., Adolph A.L., Vohra F.A., Butte N.F. Validation and calibration of physical activity monitors in children. Obes. Res. 2002;10:150–157. doi: 10.1038/oby.2002.24. [DOI] [PubMed] [Google Scholar]
  46. Ramsey B.W., Farrell P.M., Pencharz P. Nutritional assessment and management in cystic fibrosis: a consensus report. The Consensus Committee. Am. J. Clin. Nutr. 1992;55:108–116. doi: 10.1093/ajcn/55.1.108. [DOI] [PubMed] [Google Scholar]
  47. Riner W.F., Sellhorst S.H. Physical activity and exercise in children with chronic health conditions. J. Sport Health Sci. 2013;2:12–20. [Google Scholar]
  48. Rowlands A.V., Thomas P.W.M., Eston R.G., Topping R. Validation of the RT3 triaxial accelerometer for the assessment of physical activity. Med. Sci. Sports Exerc. 2004;36:518–524. doi: 10.1249/01.mss.0000117158.14542.e7. [DOI] [PubMed] [Google Scholar]
  49. Ryan J.M., Hensey O., McLoughlin B., Lyons A., Gormley J. Associations of sedentary behaviour, physical activity, blood pressure and anthropometric measures with cardiorespiratory fitness in children with cerebral palsy. PLoS ONE. 2015 doi: 10.1371/journal.pone.0123267. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Ryan J., Walsh M., Gormley J. Ability of RT3 accelerometer cut points to detect physical activity intensity in ambulatory children with cerebral palsy. Adapted Phys. Activity Quarterly. 2014;31:310–324. doi: 10.1123/apaq.2013-0088. [DOI] [PubMed] [Google Scholar]
  51. Stallings V.A., Charney E.B., Davies J.C., Cronk C.E. Nutrition-related growth failure of children with quadriplegic cerebral palsy. Dev. Med. Child Neurol. 1993;35:126–138. doi: 10.1111/j.1469-8749.1993.tb11614.x. [DOI] [PubMed] [Google Scholar]
  52. Staudenmayer J., Pober D., Crouter S., Bassett D., Freedson P. An artificial neural network to estimate physical activity energy expenditure and identify physical activity type from an accelerometer. J. Appl. Physiol. 2009;107:1300–1307. doi: 10.1152/japplphysiol.00465.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Staudenmayer J., He S., Hickey A., Sasaki J., Freedson P. Methods to estimate aspects of physical activity and sedentary behavior from high-frequency wrist accelerometer measurements. J. Appl. Physiol. 2015;1985(119):396–403. doi: 10.1152/japplphysiol.00026.2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Stephens S., Pullenayegum E., Schneiderman J., McCrindle B., Abad A., Ignas D., Takken T., Beyene J., Biggar D. Validation of accelerometer prediction equations in children with chronic disease. Pediatric Exercise Sci. 2016;28:117–132. doi: 10.1123/pes.2014-0185. [DOI] [PubMed] [Google Scholar]
  55. Torpy J.M., Campbell A., Glass R.M. Chronic diseases of children. JAMA. 2018;303:682–782. doi: 10.1001/jama.303.7.682. [DOI] [PubMed] [Google Scholar]
  56. Trost S.G., O'Neil M. Clinical use of objective measures of physical activity. Br. J. Sports Med. 2014;48:178–181. doi: 10.1136/bjsports-2013-093173. [DOI] [PubMed] [Google Scholar]
  57. Trost S.G., Fragala-Pinkham M., Lennon N., O'Neil M.E. Decision trees for detection of activity intensity in youth with cerebral palsy. Med. Sci. Sports Exerc. 2015;48:958–966. doi: 10.1249/MSS.0000000000000842. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Vabalas A., Gowen E., Poliakoff E., Casson A.J. Machine learning algorithm validation with a limited sample size. PLoS ONE. 2019 doi: 10.1371/journal.pone.0224365. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Vanhelst Jeremy, Béghin Laurent, Rasoamanana Patrick, Theunynck Denis, Meskini Touffik, Iliescu Catalina, Duhamel Alain, Turck Dominique, Gottrand Frédéric. Calibration of the RT3 accelerometer for various patterns of physical activity in children and adolescents. J. Sports Sci. 2010;28(4):381–387. doi: 10.1080/02640410903508821. [DOI] [PubMed] [Google Scholar]
  60. Varni J.W., Burwinkle T.M., Szer I.S. The PedsQL multidimensional fatigue scale in pediatric rheumatology: reliability and validity. J. Rheumatol. 2004;31:2494–2500. [PubMed] [Google Scholar]
  61. Verschuren O., Darrah J., Novak I., Ketelaar M., Wiart L. Health-enhancing physical activity in children with cerebral palsy: more of the same is not enough. Phys. Ther. 2014;94:297–305. doi: 10.2522/ptj.20130214. [DOI] [PubMed] [Google Scholar]
  62. Verschuren O., Peterson M.D., Balemans A.C., Hurvitz E.A. Exercise and physical activity recommendations for people with cerebral palsy. Dev. Med. Child Neurol. 2016;58:798–808. doi: 10.1111/dmcn.13053. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Walker J.L., Bell K.L., Stevenson R.D., Weir K.A., Boyd R.N., Davies P.S. Differences in body composition according to functional ability in preschool-aged children with cerebral palsy. Clin. Nutr. 2015;34:140–145. doi: 10.1016/j.clnu.2014.02.007. [DOI] [PubMed] [Google Scholar]
  64. Weir J.B. New methods for calculating metabolic rate with special reference to protein metabolism. J. Physiol. 1949;109:1–9. doi: 10.1113/jphysiol.1949.sp004363. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Welk G.J. Principles of design and analyses for the calibration of accelerometry-based activity monitors. Med. Sci. Sports Exerc. 2005;37:S501–S511. doi: 10.1249/01.mss.0000185660.38335.de. [DOI] [PubMed] [Google Scholar]
  66. Welk G.J., Almeida J., Morss G. Laboratory calibration and validation of the Biotrainer and Actitrac activity monitors. Med. Sci. Sports Exerc. 2003;35:1057–1064. doi: 10.1249/01.MSS.0000069525.56078.22. [DOI] [PubMed] [Google Scholar]
  67. Wells S.L.W., Laura B., Jane E.S., Jessica E.C., Samantha S., Gillian W., Shilpa D., Greg D. Physical activity for children with chronic disease; a narrative review and practical applications. BMC Pediatrics. 2019;19:1–18. doi: 10.1186/s12887-018-1377-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. West S.L., Banks L., Schneiderman J.E., Caterini J.E., Stephens S., White G., Dogra S., Wells G.D. Physical activity for children with chronic disease; a narrative review and practical applications. BMC Pediatrics. 2019;19:12. doi: 10.1186/s12887-018-1377-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. World Health Organisation . Physical activity and young people. World Health Organisation; 2015. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Data 1
mmc1.docx (18.3KB, docx)

Articles from Preventive Medicine Reports are provided here courtesy of Elsevier

RESOURCES