Introduction
Research evidence supports the benefits of physical activity (PA) for improving health-related outcomes among people with rheumatoid arthritis (RA) [
1]. More recently, studies also suggest sedentary behaviour (waking behaviour ≤ 1.5 metabolic equivalents [METs], whilst in a sitting, reclining or lying posture) [
2,
3] is adversely associated with RA outcomes [
4]. However, most evidence regarding the role of sedentary time and PA in RA is based on studies employing self-report methods to quantify engagement in these behaviours [
4,
5].
Device-based assessments of sedentary time and PA offer a more objective measure of behaviour, and have demonstrated higher validity and reliability relative to self-report instruments [
6-
8]. Consequently, devices are being more readily used to measure sedentary time and PA in different populations, including in RA [
4,
9]. Currently, hip-worn accelerometers (e.g., ActiGraph [Florida, USA]) are the most commonly employed device in RA studies to estimate the frequency, intensity and duration of free-living behaviour. The accelerometer records and stores raw acceleration data (g), which is subsequently processed to provide estimates of sedentary behaviour and PA. Currently, several processing methods can be applied to raw accelerometer data, with the dominant approach being the use of thresholds or ‘cut-points’ that classify behaviour as sedentary, light-intensity PA (LPA), moderate-intensity PA (MPA) or vigorous-intensity PA. There is an absence of a consensus on the ‘best’ method, with this decision dependent on the research question, study resources and research team expertise [
10].
A popular and widely accessible data processing method generates sedentary time and PA estimates by applying cut-points to accelerometer activity counts (‘count-based cut-points’) that have been derived from raw accelerometer data using the device manufacturer’s proprietary software. These count-based cut-points are commonly employed, largely due to intuitive and easy-to-use software platforms that facilitate straightforward processing and analysis of complex raw accelerometer data, thus, making the application of accelerometry accessible to researchers from a wide range of disciplines (e.g., clinical/medicine, exercise science, behavioural science). However, whilst the advantages of accelerometry (and specifically, count-based cut-points) to measure sedentary time and PA are being increasingly recognised by RA researchers, several limitations exist regarding their application in this patient group.
First, few accelerometers have been specifically validated for measurement of sedentary time and PA in RA (e.g., against indirect calorimetry). Consequently, existing RA studies employing accelerometers have largely employed count-based cut-points developed in validation studies of healthy participants [
11,
12] to quantify sedentary time and PA in RA. However, as RA patients differ markedly to people without RA in terms of physiology, physical function and associated activity patterns (e.g., RA patients demand a relatively higher basal metabolic rate compared to the general population [
13]), such sedentary time and PA estimates should be interpreted with caution.
Second, most existing count-based cut-points are uniaxial, generating sedentary time and PA estimates using data captured by a single axis of movement. Technological advancements are such that triaxial accelerometers are now common place, and can capture data across three axes (
Y,
X and
Z) to provide a more valid assessment of behaviour [
14]. Thus, given the increasing popularity of applying count-based cut-points to examine sedentary time and PA in RA studies, there is a critical need to develop RA-specific triaxial accelerometer count-based cut-points to provide a valid and accessible accelerometer data processing method for RA researchers.
Still, a key limitation of accelerometers is their inability to distinguish sitting (sedentary behaviour) from standing without movement (LPA). Specifically, accelerometers work on the basis that all movements registered below a ‘sedentary time cut-point’ are by default, classed as sedentary [
15]. However, low-movement behaviours may occur in a sitting or standing posture, but both may record accelerations that register below the ‘sedentary time cut-point’. Thus, accelerometers may lead to an overestimation of sedentary time by misclassifying low-movement standing behaviours as sitting (sedentary). The activPAL™ (PAL Technologies, Glasgow, UK) addresses this limitation, and is able to accurately classify behaviours as sitting/lying (sedentary), standing or stepping. This device is currently considered the gold standard for measurement of free-living sedentary time [
6]. Thus, the activPAL™ primarily offers a measure of sedentary behaviour, rather than frequency, intensity and duration of PA. Consequently, few RA studies have employed the activPAL™, with extant research employing this device focusing specifically on the role of sedentary behaviour [
16].
Considering exponential growth in research centred on the role of sedentary behaviour and PA for improving RA disease outcomes, it is critical that device-based measures are properly validated for use in this population. Therefore, the overarching aim of the current study was to validate the commonly employed ActiGraph GT3X+ and the activPAL3
μ™, for measurement of sedentary time and PA in RA. In a laboratory-validation (objective 1), this study aimed to: (a) validate the GT3X+ against indirect calorimetry to generate RA-specific triaxial (vector magnitude [VM]) accelerometer count-based cut-points for sedentary time, LPA and MPA; (b) validate the activPAL3
μ™ against direct observation for measurement of sedentary, standing and stepping time. Then, using these data, conduct a field-validation (objective 2) to compare the validity of the new RA-specific triaxial sedentary time count-based cut-point vs. a widely used non-RA uniaxial sedentary time count-based cut-point (< 100 counts/min) [
11,
12] for measurement of free-living sedentary time in RA, against the gold standard (activPAL3
μ™).
Discussion
The current study validated the ActiGraph GT3X+ and activPAL3
μ™—two devices commonly used in sedentary behaviour and PA research—for measurement of sedentary time and PA in people living with RA. Whilst there are several options for processing raw accelerometer data to quantify sedentary time and PA in healthy populations, count-based cut-points offer an accessible means of accelerometer data processing for researchers and health professionals working in rheumatology. To date, RA studies employing accelerometers have largely relied on the application of non-RA count-based cut-points to quantify free-living sedentary time and PA in this population [
29,
30], which are limited in their validity when we consider the unique physiology and associated movement patterns of people living with RA [
21,
22,
24]. Thus, there exists a critical need for the development of RA-specific count-based cut-points, which can be easily and consistently employed across RA studies.
In response, this is the first study to calibrate the commonly employed GT3X+ and define RA-specific triaxial accelerometer count-based cut-points, for valid measurement of sedentary time, LPA and MPA in RA. Our RA-specific count-based cut-points were derived according to energy requirements of behaviour among people with RA, and demonstrated high sensitivity and specificity for classification of sedentary time, LPA and MPA. Thus, the application of our novel RA-specific triaxial count-based cut-points are likely to provide more valid assessments of sedentary time and PA in RA, relative to employing non-RA uniaxial count-based cut-points developed in validation studies of healthy adults. We therefore recommend using the RA-specific count-based cut-points proposed herein, in future RA research.
This study also assessed the accuracy of the activPAL3
μ™ for measurement of sedentary, standing and stepping time in RA. Only one study has examined the ability of the activPAL™ to validly assess posture in RA [
31]. Larkin et al. [
31] employed regression analysis and observed strong
associations between activPAL™-assessed sedentary, standing and stepping time with directly observed behaviour. However, it would be surprising to find a non-significant relationship between two methods designed to measure the same variables [
27,
28]. Thus, we employed Bland–Altman analysis to determine
agreement between activPAL3
μ™-assessed vs. directly observed behaviours [
27,
32], and reported high classification accuracy (> 98%) between the two measures for all behaviours, in our sample of RA patients. This is in line with past research in non-RA populations [
26,
33] and further supports the recommendation that the activPAL™ be considered the gold standard for assessment of free-living sedentary time [
6], including in RA.
On the basis of this recommendation, we examined the validity of the RA-specific sedentary time count-based cut-point, using the activPAL3
μ™ as the criterion standard. Results revealed a mean difference of 2.3 h/day between sedentary time quantified using the RA-specific count-based cut-point vs. the activPAL3
μ™. Bland–Altman plots demonstrated most data points to fall above zero, suggesting overestimation of sedentary time using the RA-specific count-based cut-point, compared to the activPAL3
μ™. Still, when compared to the activPAL3
μ™, our RA-specific count-based cut-point produced a smaller mean difference, and narrower 95% LOA, relative to the commonly used non-RA count-based cut-point (< 100 counts/min) [
11,
12].
It is possible that the observed lack of agreement between sedentary time quantified using the RA-specific count-based cut-point vs. activPAL3
μ™-assessed sedentary time in this study reflects the inability of accelerometers to differentiate between sitting and standing, rather than relatively compromised validity of the RA-specific count-based cut-point described herein. Our data support this as a plausible explanation for two reasons. First, participants’ average MET value during ‘standing’ in the laboratory protocol was 0.8 METs (< the 1.5 METs used to define sedentary behaviour). Second, the downward trend observed in Bland–Altman plots suggests agreement between GT3X+- and activPAL3
μ™-assessed sedentary time improves at higher levels of sedentary time, where lower levels of PA (including standing) are likely to occur. That is, for people engaging in high levels of sedentary time, standing may occupy less of daily waking behaviour and, therefore, there is less opportunity to misclassify standing time as sedentary time. In a recent study comparing accelerometer- and activPAL™-assessed sedentary time in older adults, Aguilar-Farías et al. [
24] demonstrated that their population-specific sedentary time VM count-based cut-points (e.g., < 60 counts/min) were better able to detect combined activPAL™-assessed sedentary and standing time (AUC = 0.82), compared to activPAL™-assessed sedentary time alone (AUC = 0.73).
In summary, results suggest that future studies should employ the activPAL3
μ™ for valid assessment of sedentary time in people living with RA. When this is not possible, the RA-specific sedentary time count-based cut-point represents a more valid alternative, relative to the non-RA count-based cut-point of < 100 counts/min [
11,
12] in this population. However, these recommendations should be considered in the context of study limitations. First, the nature of the laboratory-validation meant that a free-living environment could not be wholly achieved, only replicated. Still, the laboratory protocol was informed by similar validation studies conducted in RA and non-RA populations, and included several activities typically undertaken in a free-living environment [
21,
31,
34]. Second, participants not reaching steady-state
VO
2 during laboratory-validation activities were excluded from ROC curve analysis, which reduced the number of data points available for cut-point calibration (out of a possible 199: sedentary time = 82; LPA = 87; MPA = 30). Nevertheless, the number of data points for each activity intensity are comparable to other studies that have developed accelerometer count-based cut-points for measuring sedentary time, LPA and MPA in populations with reduced physical function [
14]. Third, participants included in both laboratory- and field-based protocols were mostly females with moderate RA disease activity. Thus, findings may be less generalisable to male RA patients and those with more/less active disease. Future research should, therefore, confirm the validity of the RA-specific count-based cut-points and activPAL3
μ™ in different populations of RA patients (e.g., males, higher/lower disease activity). The current study has provided a ‘first step’ towards further work in this area.
Finally, the primary aim of the current study was to develop RA-specific triaxial accelerometer count-based cut-points to allow researchers to easily and consistently apply these criteria to accelerometer data in the RA population with heightened accuracy, compared to non-RA (and uniaxial) count-based cut-points. Indeed, the development of RA-specific count-based cut-points fills an important gap in the literature, providing an accessible tool for the growing number of rheumatology professionals (e.g., consultants, nurses, physiotherapists) conducting research to understand the role of sedentary time and PA in RA. However, due to a rapidly evolving field and technological advancements in the measurement of sedentary time and PA, it is important that future research examines the validity of other emerging analytical approaches that involve the development of complex data processing algorithms, to compliment the count-based cut-point validation model employed herein.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.