Actigraphy of Wrist and Ankle for Measuring Sleep Duration in Altitude Travelers

Abstract

Latshang, Tsogyal Daniela, Daniela Juliana Mueller, Christian Maurizio Lo Cascio, Anne-Christin Stöwhas, Katrin Stadelmann, Noemi Tesler, Peter Achermann, Reto Huber, Malcolm Kohler, and Konrad Ernst Bloch. Actigraphy of wrist and ankle for measuring sleep duration in altitude travelers. High Alt Med Biol. 17:194–202, 2016— Aims: Actigraphy might be convenient to assess sleep disturbances in altitude field studies. Therefore, we evaluated whether actigraphy accurately measures sleep duration in healthy subjects traveling to altitude. Methods: Fifty-one healthy men, aged mean ± standard deviation (SD) 27 ± 9 years, were studied during one night at Zurich (490 m), two nights at Davos Wolfgang (1630 m), and two nights at Jakobshorn (2590 m), in randomized order. Sleep duration measured by actigraphy, using a one-axis device at the wrist (n = 51), a three-axis device at the other wrist, and a three-axis device at the ankle (n = 22), was compared with corresponding total sleep time (TST) measured by polysomnography. Results: During 255 polysomnographic overnight studies, 449 paired actigraphic recordings were obtained. The median polysomnographic-derived TST ranged from 397 to 408 minutes. Actigraphic mean TST from wrists with one-axis and three-axis devices, and from ankle agreed well with polysomnographic values with a bias of +1, −7, +6 minutes, respectively. Corresponding limits of agreement (±2 SD of bias) were ±51, ±60, and ±59 minutes. Limits of agreement of mean TST over five nights by actigraphy and polysomnography were similar to the coefficient of repeatability (2 SD of mean) of polysomnographic TST, that is, ±31, ±38, and ±36 minutes versus ±34 minutes. Conclusions: Actigraphy of the wrist or ankle by a one-axis or a three-axis device accurately estimates mean TST in groups of subjects and mean TST over several nights in individuals traveling to altitude. Therefore, actigraphy is valuable for assessing effects of altitude and other environmental influences on sleep duration during field studies over extended periods.

Introduction

Sleep may be disturbed by various environmental stimuli, including noise, extreme temperature, and hypoxia, among others. To obtain objective data on sleep duration and its temporal distribution during field studies evaluating environmental influences, robust and simple techniques are required. Polysomnography, the gold standard for measuring sleep, is often not practicable under such conditions, because it is technically demanding, requires fragile equipment, and cannot be performed over prolonged periods.

We have recently shown that actigraphy, a technique based on recordings of acceleration by a device of the size of a watch placed at the wrist, provides accurate estimates of total sleep time (TST) in mountaineers at high altitude (Nussbaumer-Ochsner et al., 2011). In that study, we used a single-axis accelerometer, which might have underestimated movements that were not parallel to the axis of the sensor. Whether actigraphy using sensors in several axes is superior in estimating sleep time compared with a single-axis accelerometer has not been conclusively studied (Kaminsky and Ozemek, 2012). Apart from estimating sleep duration, actigraphy is also used to assess physical activity in people undergoing rehabilitation or training programs (Talkowski et al., 2009; Dyrstad et al., 2014; Freene et al., 2014; Herman Hansen et al., 2014). For such applications, the accelerometer is usually not placed at the wrist but at the hip, ankle, or upper arm, to better detect physical activity and estimate active energy consumption (Dyrstad et al., 2014; Herman Hansen et al., 2014).

Whether actigraphy by sensors placed at the ankle and at the wrist are equally suitable for estimating sleep duration has not been conclusively studied (Belanger et al., 2013). To address these points, the purpose of the current study was to evaluate the hypothesis that sleep duration measured by a one-axis and a three-axis actigraph placed at the wrist is similar to sleep duration that is correspondingly measured by polysomnography and to investigate whether an accelerometer placed at the ankle estimates similar sleep duration as one placed at the wrist.

Materials and Methods

Study design

Participants were studied during five nights: one night at the University Hospital Zurich (490 m, mean barometric pressure [PB] 719 mmHg), two nights at the Hochgebirgsklinik Davos Wolfgang (1630 m, PB 630 mmHg), and two nights at the mountain hostel Davos Jakobshorn (2590 m, PB 562 mmHg). Subjects were randomly assigned to one of four sequences of altitude exposure: 490-1630-2590 m; 490-2590-1630 m; 1630-2590-490 m; and 2590-1630-490 m. The data were collected during a randomized cross-over study evaluating effects of altitude on sleep and breathing (ClinicalTrials.gov NTC01130948) (Latshang et al., 2013); the data on actigraphy presented here have not been published.

Participants

Healthy, non-smoking men, who were 18–70 years old, living below 800 m, and with a body mass index of 18–30 kg/m², were invited to participate. Any disease requiring medical treatment, alcohol or drug abuse, and any disorder of sleep or wakefulness as evaluated by questionnaires, and previous intolerance of altitude up to 2600 m were exclusion criteria. The study was approved by the institutional ethics committee, and subjects gave written informed consent. For details, see Latshang et al. (2013).

Measurements and outcomes

Actigraphy

Two different actigraph devices (MSR; Electronics GmbH, Henggart, Switzerland) that were specifically developed for altitude field studies were used: the MSR2005 that records acceleration in one axis, and the MSR2010 that records acceleration in three axes, along with PB and temperature. Acceleration signals were recorded at 256 Hz with a 12-bit resolution over the range of ±2 g. Data averaged over consecutive 1 minute intervals (epochs) were stored in the internal memory of the devices. All participants wore a one-axis actigraph (MSR2005) at the wrist during the study period; 22 participants additionally wore a three-axis actigraph (MSR2010) at the other wrist, and one three-axis actigraph at the ankle. Synchronization of actigraphic and polysomnographic recordings was achieved by recording common time stamps.

Actigraphic data were analyzed by dedicated software (Respironics Actiware 5; Philips Respironics, Murrysville, PA). For each night, the following measures were evaluated: time in bed (time from lights-off to lights-on); TST (sum of all epochs with activity below threshold); sleep efficiency (TST in percentage of time in bed); and sleep latency (time from lights-off to the beginning of the first three consecutive epochs with activity below the threshold).

Polysomnography

Polysomnography was performed according to standard techniques (Bloch, 1997) (Alice 5; Philips Respironics). Recorded signals included electroencephalogram (C3A2, C4A1) and two electro-oculogram leads, submental and bilateral anterior tibial electromyogram (EMG), pulse oximetry, calibrated respiratory inductive plethysmography (RespitracePT; Nims, Miami Beach, FL), nasal prong pressure recordings (Thurnheer et al., 2001), an oral thermistor, and bilateral diaphragmatic surface EMG (Maarsingh et al., 2000) to differentiate obstructive from central apnea. Analyzers were blinded to the symptoms scores of subjective sleep time estimates and other clinical data (Latshang et al., 2013).

Statistics

Data are summarized as medians (quartiles) and means (standard deviation [SD]), as appropriate. Corresponding variables derived from actigraphy and polysomnography over the course of the five individual study nights were compared graphically and by computing the bias (mean difference) and 95% confidence intervals (i.e., limits of agreement, ±2 SD of bias) (Bland and Altman, 1986). To assess the precision of actigraphy in estimating mean values of TST by polysomnography over the course of five nights, limits of agreement for five-night means of TST by actigraphy and polysomnography were additionally computed and compared with the coefficients of repeatability (±2 SD of mean) (Bland and Altman, 1986) for TST by polysomnography during the same five nights. Correlations were computed by the Spearman rank-order test.

Effects of altitude were evaluated by the Friedman analysis of variance (ANOVA) followed by Wilcoxon matched-pairs or Mann–Whitney U tests, as appropriate. Statistical significance was assumed at p < 0.05 by applying a Bonferroni correction as appropriate.

Results

Of 190 screened subjects, 51 met inclusion criteria and were randomized to one of the altitude exposure sequences. All 51 participating subjects completed the study. Their median age was 24 (quartiles 20–28 years), body mass index was 23.0 kg/m² (quartiles 21.0–24.8), and Epworth sleepiness scale score was 7 points (quartiles 4–9). Further details were previously published (Latshang et al., 2013). Actigraphic data were not available for technical or logistic reasons in six subjects at Zurich and in two subjects at altitude locations. Simultaneous recordings by a one-axis device on one wrist, a three-axis device on the other wrist, and a three-axis device at an ankle were available in 22 subjects. Thus, 241 of 255 simultaneous polysomnographic and actigraphic recordings (one-axis, wrist), 106 simultaneous polysomnographic and actigraphic recordings (three-axis, wrist), and 102 simultaneous polysomnographic and actigraphic recordings (three-axis, ankle) were available.

Results of polysomnographic- and actigraphic-derived sleep measures are summarized in Table 1. TST and sleep efficiency revealed only minimal changes over the course of the study, that is, mean TST by polysomnography was between 397 and 408 minutes at the three locations (Table 1). Compared with values recorded at 490 m, oxygen saturation was slightly but statistically significantly diminished at both 1630 m and 2590 m (Table 1). The apnea/hypopnoea index was increased on the first and second nights at 1630 and 2590 m compared with values at 490 m due to high-altitude periodic breathing (Latshang et al., 2013) (Table 1).

Table 1.

Sleep Variables Assessed by Polysomnography and Actigraphy

	Zurich, 490 m	Davos Wolfgang, 1630 m		Davos Jakobshorn, 2590 m
	1st night	1st night	2nd night	1st night	2nd night	p overall
Polysomnography	n = 45	n = 49	n = 49	n = 49	n = 49
TST, min	397 (381; 411)	400 (375; 411)	408^* (401; 414)	402 (384; 409)	406 (392; 413)	0.029
SE, %	94 (91; 97)	95 (88; 97)	96 (94; 97)	95 (91; 97)	96 (93; 98)	0.079
Sleep latency, min	10 (8; 18)	11 (7; 15)	10 (7; 12)	9 (7; 13)	9^* (7; 11)	0.020
SpO₂, %	96 (95; 96)	94^* (93; 95)	94^* (93; 95)	90^*Ώ¶ (89; 91)	91^*Ώ¶ (90; 92)	<0.001
Apnea/hypopnea index, 1/h	4.2 (2.3; 6.8)	7.0^* (4.1; 12.3)	5.4 (3.5; 10.5)	13.1^*Ώ¶ (7.0; 29.0)	8.0^* (4.4; 20.5)	<0.001
Sleep quality, visual analog score	6.2 (4.0; 7.4)	6.2 (5.0; 7.5)	7.1^* (6.0; 8.0)	5.5^¶ (3.8; 7.0)	6.3 (5.4; 7.8)	0.006
Actigraphy, one-axis, wrist	n = 45	n = 49	n = 49	n = 49	n = 49
TST, min	396 (375; 406)	404 (393; 409)	401 (392; 408)	401 (392; 406)	400 (393; 405)	0.218
SE, %	94 (90; 96)	96 (93; 97)	95 (93; 96)	95 (93; 96)	95 (93; 96)	0.549
Sleep latency, min	4 (2; 8)	4 (2; 6)	3 (1; 6)	3 (2; 6)	3 (2; 6)	0.625
Actigraphy, three-axis, wrist	n = 18	n = 22	n = 22	n = 22	n = 22
TST, min	397 (387; 401)	389 (373; 400)	391 (380; 395)	389 (380; 395)	388 (376; 399)	0.437
SE, %	94 (92; 95)	92 (89; 95)	93 (90; 94)	92 (91; 94)	92 (89; 95)	0.209
Sleep latency, min	1 (0; 2)	2 (0; 4)	2 (0; 5)	3 (0; 4)	1 (0; 5)	0.533
Actigraphy, three-axis, ankle	n = 18	n = 21	n = 21	n = 21	n = 21
TST, min	408 (398; 411)	399 (394; 408)	399 (393; 404)	402 (390; 405)	404 (394; 409)	0.339
SE, %	97 (95; 98)	95 (94; 97)	95 (94; 96)	96 (93; 97)	95 (93; 97)	0.267
Sleep latency, min	1 (0; 2)	1 (0; 2)	0 (0; 3)	2 (0; 3)	1 (0; 4)	0.652

Medians (quartiles).

p < 0.05 versus 490 m, ^Ώp < 0.05 versus 1630 m day 1, ^¶p < 0.05 versus 1630 m day 2.

SE, sleep efficiency; Spo₂, oxygen saturation as measured by pulse oximetry; TST, total sleep time (TST in % of time in bed).

TST and sleep efficiency measured by the one-axis and three-axis actigraph at the wrist and at the ankle, respectively, agreed closely with corresponding values measured by polysomnography; the bias was +1, −7, and +6 minutes, respectively (Tables 1 and 2). The precision of actigraphic versus polysomnographic TST quantified by the limits of agreement was only moderate (one-axis wrist ±51 minutes, three-axis wrist ±60 minutes, three-axis ankle ±59 minutes). Figure 1 shows identity plots and Bland–Altman plots of actigraphic versus polysomnographic sleep variables. Corresponding numerical values are presented in Table 2.

FIG. 1.

Bland–Altman analysis of agreement in TST recorded at the three different altitudes (490, 1630, and 2590 m) during five individual nights in the 51 study participants by a one-axis device at the wrist, a three-axis device at the other wrist, and a three-axis device at the ankle in comparison to polysomnography, respectively. The panels show identity plots (A, C, E), and differences in TST (ΔTST) versus mean TST (actigraphy and polysomnography) (B, D, F). The mean difference (bias) between actigraphic- and polysomnographic-derived values is represented by the full horizontal line; limits of agreement (bias ±2 SD) are represented by horizontal dashed lines. TST, total sleep time.

Table 2.

Agreement of Sleep Variables Measured by Actigraphy and Polysomnography

	Zurich, 490 m	Davos Wolfgang, 1630 m		Davos Jakobshorn, 2590 m
	1st night	1st night	2nd night	1st night	2nd night	All 5 individual nights	Means of all 5 nights
Actigraphy, one-axis at wrist vs. PSG
Numbers of comparisons	45	49	49	49	49	241	49
TST, min: bias ± limits of agreement	−4 ± 40	12 ± 69	−4 ± 31	3 ± 52	−3 ± 48	1 ± 51	2 ± 31
Correlation coefficient, R	0.48^#	0.40^#	0.30^#	0.27	−0.06	0.29^#	0.29^#
Sleep efficiency, %: bias ± limits of agreement	−0 ± 10	3 ± 17	−1 ± 8	2 ± 15	0 ± 11	1 ± 13	1 ± 7
Correlation coefficient, R	0.42^#	0.44^#	0.24	0.25	0.13	0.31	0.31
Sleep latency, min: bias ± limits of agreement	−7 ± 15	−0 ± 23	−6 ± 5	−6 ± 8	−4 ± 17^Ώ¶	−7 ± 16	−6 ± 8
Correlation coefficient, R	0.53^#	0.36^#	0.68^#	0.45^#	0.30^#	0.46^#	0.46^#
Actigraphy, three-axis at wrist vs. PSG
Numbers of comparisons	18	22	22	22	22	106	22
TST, min: bias ± limits of agreement	−6 ± 40	2 ± 71	−18 ± 50	−2 ± 75	−9 ± 47	−7 ± 60	−5 ± 38
Correlation coefficient, R	0.18	0.34	−0.15	0.24	0.20	0.19	0.19
Sleep efficiency, %: bias ± limits of agreement	−1 ± 9	1 ± 17	−3 ± 10	0 ± 18	−2 ± 11	−1 ± 14	−1 ± 9
Correlation coefficient, R	0.21	0.51^#	−0.10	0.26	0.19	0.27^#	0.27^#
Sleep latency, min: bias ± limits of agreement	−9 ± 10	−2 ± 20	−8 ± 11	−8 ± 13	−7 ± 8	−9 ± 13	−9 ± 7
Correlation coefficient, R	0.43	0.33	−0.01	0.03	0.07	0.16	0.16
Actigraphy, three-axis at ankle vs. PSG
Numbers of comparisons	18	21	21	21	21	102	21
TST, min: bias ± limits of agreement	2 ± 43	17 ± 76	−4 ± 35	9 ± 77	6 ± 43	6 ± 59	8 ± 36
Correlation coefficient, R	−0.08	0.19	0.15	−0.04	0.26	0.11	0.11
Sleep efficiency, %: bias ± limits of agreement	1 ± 10	4 ± 18	0 ± 6	3 ± 18	2 ± 10	2 ± 13	2 ± 9
Correlation coefficient, R	0.02	0.15	0.32	0.18	0.22	0.17	0.17
Sleep latency, min: bias ± limits of agreement	−10 ± 10	−14 ± 28	−9 ± 9	−9 ± 11	−7 ± 11	−10 ± 16	−10 ± 8
Correlation coefficient, R	0.38	0.57^#	0.21	0.17	−0.21	0.24^#	0.24^#

Values represent mean differences (bias) of values by actigraphy minus corresponding values by polysomnography ±2 SD (limits of agreement).

Bold values are results from the entire dataset including 5 nights. Wilcoxon matched pairs comparisons were performed when Friedman analysis of variance was significant (p < 0.05), ^*p < 0.05 versus 490m, ^Ωp < 0.05 versus 1630m day 1, ^¶p < 0.05 versus 1630m day 2. Correlation analysis was performed using Spearman rank-order testing with # denoting p < 0.05.

PSG, polysomnography; SD, standard deviation.

The bias of sleep efficiency by actigraphy versus polysomnography (i.e., the mean difference) was small: one-axis device at wrist: +1%, three-axis device at wrist: −7%, and three-axis device at ankle: +6%; the corresponding limits of agreement were ±13%, ±14%, and ±13%. The bias of actigraphy versus polysomnography for sleep latency was −7, −9, and −10 minutes for the one-axis device at the wrist, the three-axis device at the wrist, and the three-axis device at the ankle, respectively. The corresponding limits of agreement were ±16, ±13, and ±16 minutes.

To evaluate the precision of actigraphy in estimating polysomnographic-derived mean sleep variables over the course of five nights, limits of agreement between actigraphic and polysomnographic means of TST, sleep latency, and sleep efficiency over five nights were compared, respectively (Table 2 and Fig. 2). Analysis of the five-night means by one-axis devices at the wrist, three-axis devices at the wrist, and three-axis devices at the ankle revealed limits of agreement with polysomnography of ±31 to ±38 minutes for TST, of ±7 to ±8 minutes for sleep latency, and of ±7% to ±9% for sleep efficiency (Table 2). Thus, these measures of precision were close to the coefficients of repeatability (i.e., 2 SD of mean difference) of ±34 minutes for TST, of ±9 minutes for sleep latency, and of ±9% for sleep efficiency by polysomnography over the course of the same five nights.

FIG. 2.

Bland–Altman analysis of agreement in the individual five-night mean values of TST measured in the 51 study participants by a one-axis device at the wrist, a three-axis device at the other wrist, and a three-axis device at the ankle in comparison to corresponding mean values measured by polysomnography. The panels show identity plots (A, C, E), and differences in TST (ΔTST) versus mean values of TST (B, D, F) recorded at the three different altitudes (490, 1630, and 2590 m). The mean difference (bias) between actigraphic- and polysomnographic-derived values is represented by a horizontal full line; limits of agreement (bias ±2 SD) are represented by horizontal dashed lines. For comparison, the range enclosed by the repeatability coefficient of TST measured by polysomnography during the same five study nights is shown by vertical lines with error bars (B, D, F). These ranges are similar to the limits of agreement between actigraphic- and polysomnographic-derived TST.

Actigraphy derived by the one-axis and the three-axis devices worn at the wrist revealed similar values, and the respective deviations from polysomnography were also similar (bias of TST, 12 minutes; limits of agreements, ±36 minutes for all comparisons) (Tables 1 and 2 and Figs. 1 and 2). Figure 3 illustrates similar rest–activity patterns recorded over the course of 4 days/nights in one subject by the three individual sensors of a three-axis device, the mean value of these signals, and the signal of a one-axis device. Acceleration recorded in the different axes of the three-axis device revealed that they were closely correlated among themselves (R = 0.914), suggesting that movements of the wrist took place in all three axes and could, therefore, be picked up by any of the three individual sensors.

FIG. 3.

Time series of acceleration (in arbitrary relative units) recorded in one subject by a one-axis device at the wrist (A), corresponding acceleration recorded by a three-axis device at the other wrist in the x, y, and z axes (B–D), and the mean value of the three-axis device (E). The accelerations in the three axes (x, y, z) (shown in B, C, D) were closely correlated (x-axis: R = 0.921, y-axis: R = 0.904, and z-axis: R = 0.887; all axes: R = 0.914, p < 0.0001 in all instances). The recording extends over four nights (91.7 hours).

Comparisons of actigraphic sleep variables derived from a three-axis device at the wrist and at the ankle revealed similar results compared with polysomnography (Tables 1 and 2 and Figs. 1 and 2).

Discussion

The current study in healthy men staying at 3 different altitudes of 490 m, 1630, and 2590 m revealed that a one-axis or a three-axis actigraph worn at the wrist or a three-axis actigraph worn at the ankle provided accurate estimates of mean TST (within 1–8 minutes) and sleep efficiency (within 1%–2%), respectively, and slightly underestimated sleep latency (by 6–10 minutes) when compared with polysomnography. The precision of the actigraphic estimates of these variables for individual nights was modest; whereas the precision of actigraphy in estimating mean TST, sleep efficiency, and sleep latency over five nights was similar to that of polysomnography as illustrated by similar measures of precision, that is, the limits of agreement of actigraphy were similar to the coefficients of repeatability of corresponding polysomnographic variables. Since actigraphy is unobtrusive and easily portable, it is a promising tool for the investigation of the effects of altitude and other environmental influences on sleep in field studies over several weeks.

The validity of actigraphic assessment of sleep duration has been investigated by several studies in subjects both with and without sleep disturbances (Kushida et al., 2001; Morgenthaler et al., 2007; Martin and Hakim, 2011; McCall and McCall, 2012). There was generally a significant correlation between actigraphic and polysomnographic sleep variables, but accuracy in terms of bias and limits of agreement of actigraphy-derived variables in comparison to polysomnography was rarely reported (Morgenthaler et al., 2007). In 21 healthy volunteers, De Souza et al. (2003) measured TST by actigraphy using two different scoring algorithms that provided TST with a bias of 18.5 and 8.1 minutes, respectively, and limits of agreement of ±41.4 and ±42.0 minutes, compared with polysomnography. In 24 patients with obstructive sleep apnea syndrome discontinuing their continuous positive airway pressure therapy during a night in the sleep laboratory, Gagnadoux et al. (2004) reported a bias and limits of agreement of TST versus polysomnography of 2.5 minutes and ±70.6 minutes, respectively. Kushida et al. (2001) described even a larger bias and limits of agreements of actigraphic TST versus polysomnography (60 and ±88.2 minutes) in 100 patients with sleep disorders. Furthermore, in a total of 77 healthy adults and chronic insomnia patients, Marino et al. (2013) evaluated accuracy of actigraphy in detecting sleep and wakefulness compared with polysomnography. The bias and limits of agreement of the wakefulness after sleep onset time were −12.6 ± 68 minutes; values for TST were not reported. In studies of 54 depressed insomniacs, actigraphy revealed a bias and limits of agreement of TST versus polysomnography of +12.8 and ±120.5 minutes, respectively (McCall and McCall, 2012).

In the current study performed at moderate altitude up to 2590 m, the accuracy of wrist actigraphy with either one-axis or three-axis devices was similar to values reported in the cited studies performed at low altitude in terms of bias and limits of agreement of TST (Table 2). Moreover, accuracy of actigraphic-derived sleep efficiency in the current study was also similar to that reported for normal subjects in previous comparisons (de Souza et al., 2003). In a high-altitude study in 14 mountaineers staying in a hut at 4559 m (Capanna Regina Margherita), we found a slightly better precision of actigraphic TST versus polysomnography; that is, the limits of agreement were ±35 minutes, and the bias was +5 minutes (Nussbaumer-Ochsner et al., 2011). The lower precision of TST achieved in the current compared with the previous study might be related to a greater heterogeneity in the 51 participants compared with the 14 mountaineers in the previous study, differences in sleep structure and in the amount of periodic breathing at 4559 versus 2590 and 1630 m, and other, unknown factors.

Sleep estimates of the wrist actigraphy with one-axis and three-axis devices were similar (Tables 1 and 2 and Fig. 3). Apparently, movements of the arm that indicate wakefulness during the night were associated with acceleration in several axes, so that they were detected by each of the sensitive single-axis sensors in both the three-axis and the one-axis device (Fig. 3).

According to our knowledge, the current study is the first to compare sleep variables derived by actigraphy at the ankle and at the wrist with corresponding data by polysomnography (Israel et al., 2003; Littner et al., 2003). The results revealed a similar accuracy in estimation of polysomnographic TST by actimeters at the two positions (Table 2). Kos et al. (2007) investigated actigraphic recordings from wrist versus ankle in 19 patients suffering from multiple sclerosis and in 10 healthy subjects during 3 days. Both sensor positions provided similar data on daytime activity patterns, but the wrist device was preferred by the patients. Unfortunately, sleep was not assessed. Other studies have suggested that activity measurements during walking and running by a sensor at the ankle were underestimating movements of the upper part of the body, but no sleep data were available (Hendelman et al., 2000). The current data establish the potential of actigraphy of the ankle to evaluate both sleep duration and physical activity during daytime with one single device.

Although actigraphy had a limited precision in estimating individual TST and other sleep variables during one single night, the technique is still valuable to evaluate sleep changes in response to environmental exposures and other interventions in groups of study participants. Since actigraphy can be applied conveniently over multiple nights, mean estimates of TST can be obtained. Our analysis of mean TST, sleep latency, and sleep efficiency over five nights revealed an improved precision so that limits of agreement with polysomnography were similar to the coefficient of repeatability of the reference standard, polysomnography (Fig. 2). Thus, actigraphy is particularly well suited in settings where mean sleep variables over prolonged periods are of interest.

A limitation of actigraphy is its inability to detect changes in sleep structure. Polysomnography did not show any significant or clinically relevant changes in TST, sleep latency, or sleep efficiency at the higher altitudes of 1630 and 2590 m compared with the 490 m baseline studies (Table 1). In contrast, analysis of sleep stages and of spectral power of the electroencephalogram (previously reported) (Latshang et al., 2013; Stadelmann et al., 2013) revealed a slight reduction of non-rapid eye movement stages 3 and 4 and a reduced spectral power in the low-frequency range.

Conclusion

In summary, our data suggest that actigraphy of the wrist or ankle with one-axis or three-axis devices is well feasible and accurate at different moderate altitudes to measure TST, sleep efficiency, and sleep latency. Considering the simple and unobtrusive application and the potential for use over several weeks, the technique is highly valuable for investigations of sleep during altitude field studies.

Footnotes

Acknowledgments

Grants from the Swiss National Science Foundation and the Zurich Center of Integrative Human Physiology, the University of Zurich; Swiss Federal Accident Insurances (SUVA), the Lung League of Zurich are acknowledged.

Author Disclosure Statement

No competing financial interests exist.

References

Ancoli-Israel

, Cole

, Alessi

, Chambers

, Moorcroft

, and Pollak

. (2003). The role of actigraphy in the study of sleep and circadian rhythms. Sleep, 26:342–392.

Belanger

, Bernier

, Paquet

, Simard

, and Carrier

. (2013). Validating actigraphy as a measure of sleep for preschool children. J Clin Sleep Med, 9:701–706.

Bland

, and Altman

. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, 1:307–310.

Bloch

. (1997). Polysomnography: a systematic review. Technol Health Care, 5:285–305.

de Souza

, Benedito-Silva

, Pires

, Poyares

, Tufik

, and Calil

. (2003). Further validation of actigraphy for sleep studies. Sleep, 26:81–85.

Dyrstad

, Hansen

, Holme

, and Anderssen

. (2014). Comparison of self-reported versus accelerometer-measured physical activity. Med Sci Sports Exerc, 46:99–106.

Freene

, Waddington

, Chesworth

, Davey

, and Cochrane

. (2014). Validating two self-report physical activity measures in middle-aged adults completing a group exercise or home-based physical activity program. J Sci Med Sport, 17:611–616.

Gagnadoux

, Nguyen

, Rakotonanahary

, Vidal

, and Fleury

. (2004). Wrist-actigraphic estimation of sleep time under nCPAP treatment in sleep apnoea patients. Eur Respir J, 23:891–895.

Hendelman

, Miller

, Baggett

, Debold

, and Freedson

. (2000). Validity of accelerometry for the assessment of moderate intensity physical activity in the field. Med Sci Sports Exerc, 32:S442–S449.

10.

Herman Hansen

, Bortnes

, Hildebrand

, Holme

, Kolle

, and Anderssen

. (2014). Validity of the ActiGraph GT1M during walking and cycling. J Sports Sci, 32:510–516.

11.

Kaminsky

, and Ozemek

. (2012). A comparison of the Actigraph GT1M and GT3X accelerometers under standardized and free-living conditions. Physiol Meas, 33:1869–1876.

12.

Kos

, Nagels

, D'Hooghe

, Duquet

, Ilsbroukx

, Delbeke

, and Kerckhofs

. (2007). Measuring activity patterns using actigraphy in multiple sclerosis. Chronobiol Int, 24:345–356.

13.

Kushida

, Chang

, Gadkary

, Guilleminault

, Carrillo

, and Dement

. (2001). Comparison of actigraphic, polysomnographic, and subjective assessment of sleep parameters in sleep-disordered patients. Sleep Med, 2:389–396.

14.

Latshang

, Lo Cascio

, Stowhas

, Grimm

, Stadelmann

, Tesler

, Achermann

, Huber

, Kohler

, and Bloch

. (2013). Are nocturnal breathing, sleep, and cognitive performance impaired at moderate altitude (1,630–632,590 m)?. Sleep, 36:1969–1976.

15.

Littner

, Kushida

, Anderson

, Bailey

, Berry

, Davila

, Hirshkowitz

, Kapen

, Kramer

, Loube

, et al. (2003). Practice parameters for the role of actigraphy in the study of sleep and circadian rhythms: an update for 2002. Sleep, 26:337–341.

16.

Maarsingh

, van Eykern

, Sprikkelman

, Hoekstra

, and van Aalderen

. (2000). Respiratory muscle activity measured with a noninvasive EMG technique: technical aspects and reproducibility. J Appl Physiol (1985), 88:1955–1961.

17.

Marino

, Li

, Rueschman

, Winkelman

, Ellenbogen

, Solet

, Dulin

, Berkman

, and Buxton

. (2013). Measuring sleep: accuracy, sensitivity, and specificity of wrist actigraphy compared to polysomnography. Sleep, 36:1747–1755.

18.

Martin

, and Hakim

. (2011). Wrist actigraphy. Chest, 139:1514–1527.

19.

McCall

, and McCall

. (2012). Comparison of actigraphy with polysomnography and sleep logs in depressed insomniacs. J Sleep Res, 21:122–127.

20.

Morgenthaler

, Alessi

, Friedman

, Owens

, Kapur

, Boehlecke

, Brown

, Chesson

Jr. , Coleman

, Lee-Chiong

, et al. (2007). Practice parameters for the use of actigraphy in the assessment of sleep and sleep disorders: an update for 2007. Sleep, 30:519–529.

21.

Nussbaumer-Ochsner

, Schuepfer

, Siebenmann

, Maggiorini

, and Bloch

. (2011). High altitude sleep disturbances monitored by actigraphy and polysomnography. High Alt Med Biol, 12:229–236.

22.

Stadelmann

, Latshang

, Lo Cascio

, Tesler

, Stoewhas

, Kohler

, Bloch

, Huber

, and Achermann

. (2013). Quantitative changes in the sleep EEG at moderate altitude (1630 m and 2590 m). PLoS One, 8:e76945.

23.

Talkowski

, Lenze

, Munin

, Harrison

, and Brach

. (2009). Patient participation and physical activity during rehabilitation and future functional outcomes in patients after hip fracture. Arch Phys Med Rehabil, 90:618–622.

24.

Thurnheer

, Xie

, and Bloch

. (2001). Accuracy of nasal cannula pressure recordings for assessment of ventilation during sleep. Am J Respir Crit Care Med, 164:1914–1919.