Evaluation of 2 different wrist actigraphy devices in the adult population

Abstract

Study Objectives:

Actigraphy devices are used in sleep medicine. The Actiwatch 2 (Philips Respironics) was an example of a frequently used device in this field. Nevertheless, the discontinuation of this device has led to an increased necessity for the implementation of other available actigraphy methods capable of providing objective information. The objective of this study was to assess the performance of the new ActTrust 2 compared to the Actiwatch 2 model.

Methods:

This observational prospective study included 9 participants (77.760 activity logs) who were monitored for 7 days using 2 actigraphy wrist devices (ActTrust 2 and Actiwatch 2) simultaneously. The following variables were evaluated: midline estimating statistic of rhythm, amplitude, and acrophase; intradaily variability; interdaily stability; relative amplitude; and the mean of 5 consecutive hours with the lowest activity and the 10 consecutive hours with the highest activity. Furthermore, total sleep time, time in bed, sleep efficiency, sleep onset latency, wake after sleep onset, and awakenings were also included.

Results:

Actigraphy models indicated statistically significant differences in activity levels. Regarding the analysis of circadian rest–activity rhythms, 10 consecutive hours with the highest activity, midline estimating statistic of rhythm, and amplitude also exhibited these differences. Furthermore, the analysis of sleep–wakefulness revealed significant differences in the sleep onset latency and the number of awakenings.

Conclusions:

The ActTrust 2 and Actiwatch 2 models showed equivalent results in measuring circadian rest–activity rhythm and sleep. However, caution is advised when interpreting parameters such as midline estimating statistic of rhythm, amplitude, sleep onset latency, awakenings, and 10 consecutive hours with the highest activity variables.

Citation:

Henríquez-Beltrán M, Benítez ID, Juez-Garcia I, et al. Evaluation of 2 different wrist actigraphy devices in the adult population. J Clin Sleep Med. 2025;21(5):805–812.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Actigraphy devices, such as the Actiwatch 2, have been extensively used in sleep medicine to provide objective data on sleep patterns. With the discontinuation of Actiwatch 2, there is an increasing necessity to investigate alternative actigraphy options to ensure the continued availability of reliable sleep-related data.

Study Impact: The ActTrust 2 and Actiwatch 2 models showed equivalent results in measuring circadian rest–activity rhythm and sleep. However, caution is advised when interpreting parameters such as midline estimating statistic of rhythm, amplitude, sleep onset latency, awakenings, and 10 consecutive hours with the highest activity variables.

INTRODUCTION

For decades, wrist actigraphy devices have been valuable portable devices for recording information related to sleep–wakefulness and circadian rest–activity rhythm.^1–3 Actigraphs are based on data collected through integrated accelerometers, which capture movement on 3 axes: the x axis (forward and backward), the y axis (left and right), and the z axis (up and down).^2,4

Actigraphy infers sleep or wakefulness based on activity counts using scoring algorithms that have been validated using polysomnography.¹ According to the International Classification of Sleep Disorders, third edition, actigraphy devices are essential in the diagnosis of circadian rhythm sleep–wake disorders,^5,6 providing effective data for the diagnosis, monitoring, and treatment of sleep disorders and circadian rhythm sleep–wake disorders.¹ Actigraphy data results have been correlated with melatonin and body temperature evaluations.⁷ Moreover, actigraphy wrist devices have been shown as a valid and useful tool to assess the continuity of sleep, especially in patients with insomnia, both for diagnosis and during follow-up treatment.¹ In clinical contexts, actigraphs’ affordability and use of at least 7 days provide an accessible and objective measure of rest–activity patterns and parameters associated with sleep in both children and adult populations, particularly useful in patients with cognitive changes and those unable to provide sleep records.¹ Currently, a wide variety of accurate wrist actigraphy devices are available to monitor sleep–wakefulness and provide information about sleep and rest–activity rhythm. The Actiwatch 2 model from Philips is a device that is extensively used by expert sleep clinicians and sleep and chronobiology researchers. However, this model has been discontinued and is unavailable. Consequently, the necessity has arisen for the implementation of an alternative actigraphy method that is capable of providing objective information about circadian rest–activity rhythms and sleep–wakefulness. In this context, The ActTrust 2 model from Condor Instruments facilitates circadian rest–activity rhythm and sleep–wake studies using customizable data acquisition, an extended battery life, and a light sensor capable of capturing detailed lux and multispectral light data. The software generates automated reports comprising parametric and nonparametric analyses, as well as actograms integrating peripheral temperature and light exposure data. The streamlined connectivity interface facilitates data transfer, thereby supporting efficient and comprehensive research workflows. The primary objective of this study was to assess the performance of the ActTrust 2 model (Condor Instruments) and the Actiwatch 2 model (Philips Respironics) on circadian rest–activity rhythm and sleep–wakefulness.

METHODS

This observational prospective study was designed with an exploratory approach to evaluate the performance of sleep–wakefulness and circadian rest–activity rhythms between 2 different actigraphy wrist devices, the ActTrust 2 model (Condor Instruments) and the Actiwatch 2 model (Philips Respironics).

The study recruited 10 predominantly healthy daytime workers (5 male and 5 female) from a private university community in Los Angeles, Chile. The inclusion criteria were as follows: participants had to be over the age of 18 and under the age of 60 years and had to sign the informed consent form. Individuals with chronic respiratory disease, mental or physical disability, or any type of cardiometabolic disease not under control were excluded from the study. Those exhibiting circadian misalignment at baseline, as determined by actigraphy and self-reported sleep patterns, and those with inaccuracies in the raw data were also excluded. This study received approval from the Scientific Ethics Committee of the Universidad Santo Tomás in Santiago, Chile.

Study design

The actigraphy protocol entailed a clinical evaluation, which encompassed sociodemographic and anthropometric data collection. Obstructive sleep apnea was evaluated using a home sleep apnea test with the ApneaLink Air (Resmed, Australia) device. In accordance with the American Academy of Sleep Medicine guidelines,⁸ 1 researcher manually scored the test, maintaining complete blinding to the clinical data. The respiratory disturbance index was calculated, with obstructive sleep apnea defined as a respiratory disturbance index of ≥ 5 events/h.⁸ Moreover, a self-reported sleep assessment was conducted to ascertain the likelihood of obstructive sleep apnea and chronotype. Subsequently, the participants were asked to wear 2 different actigraphy wrist devices on the same extremity for 7 days, in conjunction with keeping a sleep diary. The sleep diary was used to record their sleep patterns, including bedtimes, wake times, and naps, providing self-reported data on their sleep habits.

Data collection

The data set included sociodemographic and anthropometric information, including age, weight, sex, height, hip circumference, alcohol consumption, and current smoking status. Furthermore, the study included their specific geographic location, level of higher education, and whether they were daytime workers.

Sleep questionnaire

The study employed the STOP-Bang questionnaire to determine the risk of obstructive sleep apnea.^8,9 Additionally, the study used the Morningness-Eveningness Questionnaire to evaluate the cohort’s chronotypes.¹⁰ Scores of 0–2 indicate a low risk, 3–4 indicate an intermediate risk, and 5–8 indicate a high risk of obstructive sleep apnea. Scores of 16–30 indicated an extreme evening chronotype, 31–41 a moderate evening chronotype, 42–58 an intermediate chronotype, 59–69 a moderate morning chronotype, and 70–86 an extreme morning chronotype.

Actigraphy variables

Following clinical evaluation and self-reported sleep measurement, individuals in this cohort received 2 actigraphy wrist devices ActTrust 2 model (Condor Instruments, Brazil) and the Actiwatch 2 model (Philips Respironics, United States). Both devices were worn by each participant on the same extremity for 7 days. Additionally, participants completed a sleep diary during which the patients wore the actigraphy device. This study considered variables related to circadian rest–activity rhythm and sleep–wakefulness.^11,12

Circadian rest–activity rhythm

We considered the nonparametric measures based on activity counts in 60-second epochs. Intradaily variability (IV) represents the fragmentation of the rest–activity rhythm within each 24 hours. Interdaily stability (IS) represents the similarity between one 24-hour period and the next. In addition, the relative amplitude (RA) was measured. This describes the robustness of the rest–activity rhythm and indicates whether there is a difference in the magnitude of activity between the active and rest phases. For the Actiwatch 2 model from Philips Respironics, the relative amplitude was calculated using the mean activity of the 5 consecutive hours with the lowest activity (L5) and the mean activity of the 10 consecutive hours with the highest activity (M10). The RA formula is as follows: M10 − L5/M10 + L5.^11–13 Moreover, cosinor variables were employed. The midline estimating statistic of rhythm (mesor) is defined as the mean value of the cosine curve; the amplitude is defined as the distance between the mean value and the maximum or minimum peak of a curve. The amplitude represents the magnitude of rhythm oscillations and, in physiological terms, indicates the intensity of change in the evaluated parameter throughout the cycle. The acrophase is obtained through the application of the cosinor method, which entails the fitting of a cosine curve to a time series of data. The acrophase is defined as the moment at which the maximum value of the cosine curve is reached.^2,11,12

Sleep–wakefulness

In this study, we considered the following sleep variables: time in bed; total sleep time; sleep efficiency, defined as the ratio between total sleep time and the time spent in bed; sleep onset latency (SOL), defined as the time spent awake until the first sleep episode while in bed; awakenings; and time spent wake after sleep onset.^1,13

Statistical analysis

Descriptive statistics were employed to summarize the characteristics of the study population. Qualitative variables were reported as absolute and relative frequencies, whereas quantitative variables were presented as medians with interquartile ranges (25th–75th percentiles).

Nonparametric methods were employed to assess the comparability of actigraphy parameters across devices. The Wilcoxon signed-rank test was used for paired comparisons, and the Kolmogorov–Smirnov test was employed to evaluate discrepancies in the distribution of activity logs. These methods were selected due to the limited sample size and the nonnormal distribution of certain variables.

To model circadian rhythms, the mean activity for each time point over a 24-hour period was calculated for each participant and averaged across the cohort. Subsequently, a moving average filter was applied to the activity signal in order to achieve a more refined and continuous output. The cosinor model was employed to estimate circadian rhythm parameters (mesor, amplitude, and acrophase) for each device, because it is a well-established method for chronobiological analysis using the aforementioned data.¹⁴ Furthermore, a population-mean cosinor analysis was conducted by averaging the linearized parameters from the individual models.¹⁵ All statistical analyses were conducted using R software (version 4.0.1; R Foundation for Statistical Computing, Vienna, Austria). A P value threshold of < .05 was established to define statistical significance.

RESULTS

Cohort characteristics

The cohort initially constituted 10 daytime workers residing in urban areas. However, 1 patient was excluded from the study due to the presence of inaccuracies in the raw data. The median (p₂₅; p₇₅) age was 31.0 (29.0; 53.0) years, and 55.6% were female. The anthropometric evaluation revealed a median body mass index of 26.1 (24.3; 27.9), and hip circumference exhibited a median of 101 cm (98.3; 105). A total of 66.7% had completed higher education. Smoking and alcohol consumption were reported as 11.1% and 44.4% of the study cohort, respectively. Concerning comorbidities, insulin resistance and arterial hypertension were observed in 11.1% of the cohort. Additionally, the cohort demonstrated a median respiratory disturbance index of 1.0 (0.1–4.0), a low prevalence of obstructive sleep apnea (88.9%), and a predominantly moderate morning chronotype (55.6%).

Activity logs

A total of 77.760 activity logs were recorded by each device (ActTrust 2 and Actiwatch 2). Concerning the distribution of activity, the Actiwatch 2 device exhibited higher values in the interquartile range compared to the ActTrust 2 device, with a median (first quartile; third quartile) of 52.0 (0.00; 254) compared to 62.0 (0.00; 321), respectively. However, both activity distributions exhibited similar values in the first and second quartiles (see Figure 1). Furthermore, the correlation between the activity measurements of the 2 devices was found to be strong, with a cross-correlation of 0.856 (0.848; 0.921).

Figure 1. The activity logs of each actigraphy model (red, ActTrust 2 model and green, Actiwatch 2 model).

The y axis is divided into 4 quartiles, ranging from 0–1, which represents the cumulative probability of both devices. The x axis, in turn, represents the distribution of the activity logs of both devices. The y axis of the figure on the right represents the distribution of activity values in both devices represented on the x axis.

Circadian rest–activity rhythm

The circadian rest–activity rhythm was evaluated using nonparametric measures, including IV, IS, RA, L5, and M10 (see Figure 2). No significant differences were observed when relative measures such as IV, IS, and RA were compared between ActTrust 2 and Actiwatch 2. As anticipated, M10 was significantly lower in ActTrust 2 compared to Actiwatch 2, with a median (first quartile; third quartile) of activity of 231.9 (223.2; 267.2) and 328.8 (307.2; 347.8), respectively (see Table 1). Conversely, the values of L5 did not demonstrate any significant differences between the 2 devices. Moreover, with regard to the parameters of the cosinor model, the ActTrust 2 model exhibited lower values for mesor and higher values for amplitude when compared to the Actiwatch 2. However, no significant differences were observed in acrophase between the devices (see Table 1). The estimation of the population cosinor showed coordinated rhythmicity between the devices (see Figure 3).

Figure 2. A comparative analysis of circadian rest–activity rhythm data obtained from ActTrust 2 (Condor Instruments) and the Actiwatch 2 (Philips Respironics).

Table 1.

Circadian rest–activity rhythm and cosinor parameters.

Actigraphy Variable	ActTrust 2 Model	Actiwatch 2 Model	P
	Median (Q1; Q3)	Median (Q1; Q3)
Circadian rest–activity rhythm
IV	0.8 (0.5; 0.9)	0.7 (0.6; 1.0)	.895
IS	0.6 (0.5; 0.6)	0.6 (0.5; 0.6)	.689
RA	0.9 (0.9; 1.0)	0.9 (0.9; 0.9)	.788
L5	6.8 (5.9; 9.8)	10.3 (9.1; 11.6)	.070
M10	231.9 (223.2; 267.2)	328.8 (307.2; 347.8)	.005*
Cosinor
Mesor	147.9 (124.9; 168.4)	191.6 (175.0; 214.8)	.007*
Amplitude	167.9 (159.1; 175.7)	127.8 (106.1; 145.8)	.012*
Acrophase	0.8 (0.6; 1.1)	0.7 (0.7; 1.1)	.895

^*Statistically significant differences. Acrophase = moment at which the maximum value of the cosine curve is reached, amplitude = half of the vertical distance between its highest and lowest points, IS = interdaily stability, IV = intradaily variability, L5 = mean activity of the 5 consecutive hours with the lowest activity, M10 = mean activity of the 10 consecutive hours with the highest activity, mesor = midline estimating statistic of rhythm, Q = quartile, RA = relative amplitude.

Figure 3. The moving average of the mean activity values and the cosinor analysis of the ActTrust 2 (Condor Instruments) and the Actiwatch 2 (Philips Respironics).

Sleep analysis

Finally, the rest periods were evaluated using sleep parameters estimated for each device. The parameters of sleep time estimation, including time in bed, total sleep time, sleep efficiency, and wake after sleep onset, demonstrated no significant differences between devices. However, the ActTrust 2 model exhibited lower values compared to the Actiwatch 2 in sleep onset latency and awakenings (see Table 2 and Figure S1 in the supplemental material).

Table 2.

Sleep parameters.

Actigraphy Variable	ActTrust 2 Model	Actiwatch 2 Model	P
	Median (Q1; Q3)	Median (Q1; Q3)
Sleep
TST	451.9 (434.0; 476.6)	412.1 (397.3; 473.6)	.233
TIB	523.9 (468.7; 535.9)	479.7 (445.3; 529.9)	.508
SE	91.6 (86.4; 92.8)	87.1 (85.7; 88.9)	.171
WASO	40.6 (26.0; 69.1)	44.3 (39.4; 47.4)	.825
Awakenings	8.7 (6.7; 9.1)	22.3 (21.3; 26.9)	<.001*
SOL	1.6 (1.3; 2.1)	7.0 (5.3; 15.3)	.015*

^*Statistically significant differences. SE = sleep efficiency, SOL = sleep onset latency, TIB = time in bed, TST = total sleep time, WASO = wake after sleep onset.

DISCUSSION

The findings of this study demonstrated statistically significant differences between ActTrust 2 and Actiwatch 2 actigraphy devices. The ActTrust 2 model exhibited a lower level of activity than that recorded by the Actiwatch 2 model. The analysis of circadian rest–activity rhythms revealed significant differences only in the M10 variable. In contrast, cosinor analysis revealed significant differences only in the mesor and amplitude variables. Furthermore, concerning the sleep variables, significant differences were observed in awakenings and SOL.

Currently, the actigraphy wrist device is an essential clinical tool for the diagnosis and follow-up stage in treatments of sleep–wake disorders (circadian rhythm sleep disorders).¹⁶ It is also a useful monitoring device for sleep disorders such as insomnia, excessive daytime sleepiness, and isolated rapid eye movement sleep behavior disorder.^17,18 Some studies have examined and contrasted the characteristics of various actigraphs.^19–21 The purpose of this study was to assess the performance of the ActTrust 2 model and the Actiwatch 2 model on circadian rest–activity rhythm and sleep–wakefulness. The Actiwatch 2 model is one of the most frequently employed accelerometer devices in the field of sleep research²² with efficacy substantiated through comparison with polysomnography, demonstrating higher sensitivity and specificity.²³ The ActTrust 2 model is an actigraphy device manufactured by the Condor Instruments Company. In the validation with polysomnography, this device demonstrated satisfactory performance with excellent sensitivity of 95.6% and good precision of 80.2% but low specificity in the evaluation of patients with sleep-disordered breathing.²⁴

At present, several studies investigating sleep–wakefulness and circadian rest–activity rhythm have incorporated the use of the ActTrust 2 model device in their methodologies.^25–28

According to our findings, the distribution of activity values between the ActTrust 2 model and the Actiwatch 2 model exhibited a similar trajectory during the initial stages and up to the second quartile (see Figure 1). However, as activity levels increase, the distribution begins to diverge, resulting in statistically significant differences. These results suggest an equivalent capacity for rest periods, although these equivalences diminish during periods of elevated activity, leading to substantial differences. Differences in activity levels and circadian metrics between the devices may lead to misinterpretations, causing clinicians to underestimate daytime activity or misinterpret sleep quality, which can result in inappropriate treatment recommendations.

Regarding the differences found in activity recordings, we hypothesize that they are due to the different thresholds set by the different actigraphs for the time above threshold calculation. Time above threshold is a measure of the number of tenths of a second spent above a certain activity level, which serves as an activity threshold.^29–32 This implies that, for a similar motion pattern, a different threshold would result in different time above threshold values. This idea is supported by the similarities observed in the activity recordings of both devices.

In a clinical context, discrepancies in activity thresholds may influence decision-making by affecting the assessment of circadian rest–activity patterns and physical activity levels. For instance, lower M10 values in ActTrust 2 compared to Actiwatch 2 could alter the identification of peak activity. Such differences may affect interpretations in contexts requiring precise activity monitoring, including rehabilitation tracking and sleep disorder evaluations. Future research could investigate the establishment of standardized reference values or calibration protocols to enhance the comparability between devices, thereby improving the reliability of actigraphy in both research and clinical contexts when different actigraphs are used in a single study.

Concerning the analysis of circadian rest–activity rhythms, significant differences were noted in the M10 variable between the 2 actigraphy models. Additionally, there was a similar pattern in the patients, indicating that both actigraphy models are similarly capable of detecting IV, IS, and RA variables. A study conducted by Bellone et al¹⁹ comparing 5 actigraphs, including the ActTrust 2 model, found no significant differences related to circadian rest–activity rhythm.

Moreover, our findings indicated statistically significant differences in both the mesor and amplitude variables. These results can be explained by the higher average registered activity levels found in the Actiwatch 2 model, which exhibited higher average registered activity levels and amplitude compared to the ActTrust 2 model, in alignment with the results of the previous distribution of activity values. Nonetheless, the acrophase variable did not exhibit significant differences, suggesting that the ActTrust 2 model and the Actiwatch 2 model are equivalent in their calculation of the zenith of the cosine curve value. This result is consistent with previous results,¹⁹ where significant differences were seen in mesor and amplitude, but not in the acrophase, where the ActTrust 2 model was equivalent to other actigraphy devices.

The ActTrust 2 model and the Actiwatch 2 model did not demonstrate statistically significant differences in sleep variables including total sleep time, time in bed, sleep efficiency, and wake after sleep onset. Nevertheless, the data revealed statistically significant differences in the number of awakenings and SOL. Despite the absence of significant differences in other sleep metrics, it is important to note that the observed patterns in sleep efficiency and wake after sleep onset indicate that the 2 models should not be considered entirely analogous.

Previous studies that compared Actigraphs models, including the ActTrust 2, demonstrated a high degree of correlation and no significant differences in variables such as sleep onset, total sleep time, and sleep efficiency.¹⁹ However, despite seeming similarities, research indicates notable differences in the algorithms used and the methods of data processing, influenced by the thresholds applied in movement measurements.

The systematic review by Scott et al that showed discrepancies between Philips ACTiwatch devices and other wearables diapositives described an overestimation of 2–3 minutes and a subestimation up to 14 minutes in sleep variables including sleep onset latency and time awake.³³ They also indicated that the precision of wearable devices for measuring sleep onset is largely determined by 3 key factors: the methodology employed, the characteristics of the population being studied, and the characteristics of the algorithms used.

In order to gain a comprehensive understanding of the current data, it is essential to acknowledge the limitations of the study. First, the sample size was relatively modest, which may have diminished the statistical power and limited the generalizability of the results. Although the extensive activity logs collected by the actigraphy devices provide a comprehensive account of the participants’ daily routines, the relatively small sample size may limit the ability to detect subtle circadian variations that could be evident in a larger cohort. Second, the study population was exclusively composed of Hispanic/Latino participants from a single geographic location, which limits the generalizability of the findings to other populations with different cultural, environmental, or genetic influences on circadian rhythms. Furthermore, lifestyle factors such as physical activity and sedentary behavior were not taken into account, despite the potential for these to exert a significant influence on circadian rhythms and sleep quality. Third, although actigraphy is a valuable tool for assessing sleep–wake patterns and circadian rest activity rhythms over extended periods, it does not offer the same level of detail as polysomnography, which is the gold standard for sleep evaluation. Nevertheless, actigraphy’s capacity to continuously monitor sleep and activity over several days offers distinctive insights into daily life. In addition, the study did not consider psychological factors, such as stress and anxiety, which could influence circadian rest activity rhythms, or interindividual variability in device adherence, which could introduce bias.

The ActTrust 2 is currently constrained by regional availability due to its recent market introduction, particularly in comparison to Philips devices. Additionally, it requires specific training, which may limit its broader adoption in certain contexts. However, ActTrust 2 offers a relatively affordable cost, providing a viable option while delivering objective and equivalent data on cosinor, sleep–wakefulness, and circadian rest–activity rhythm. Our preliminary data suggest that the ActTrust 2 model from Condor Instruments provides equivalent results compared with the Philips Actiwatch 2 model, which has been discontinued. Both devices yielded consistent measurements for circadian rest–activity rhythm and sleep–wakefulness, with parameters such as acrophase, total sleep time, time in bed, sleep efficiency, and wake after sleep onset showing equivalent values. Additional parameters, including IV, IS, RA, and L5, also demonstrated equivalence between the 2 models. However, careful interpretation is warranted for variables such as mesor, amplitude, SOL, awakenings, and M10.

CONCLUSIONS

The ActTrust 2 and Actiwatch 2 models yielded equivalent results in measuring circadian rest–activity rhythm and sleep–wakefulness. Nevertheless, it is advisable to exercise caution when interpreting certain parameters, including mesor, amplitude, SOL, awakenings, and M10 variables. Future research including populations with different sleep disorders would validate our findings.

DISCLOSURE STATEMENT

All authors have reviewed and approved the manuscript. Work for this study was performed at the Universidad Santo Tomás, Los Angeles, Chile. The authors report no conflicts of interest.

ABBREVIATIONS

IS

interdaily stability

IV

interdaily variability

L5

mean of 5 consecutive hours with the lowest activity

M10

the 10 consecutive hours with the highest activity

mesor

midline estimating statistic of rhythm

RA

relative amplitude

SOL

sleep onset latency