Share this article on:

A Novel Tool to Quantify Physical Activities: Ambulatory Accelerometry in Psychopharmacology

Tulen, Joke H.M. PhD; Bussmann, Hans B.J. MSc; van Steenis, Hugo G. MSc; Pepplinkhuizen, Lolke MD, PhD; Man't Veld, Arie J. in MD, PhD

Journal of Clinical Psychopharmacology: June 1997 - Volume 17 - Issue 3 - p 202-207

Accelerometry by means of body-mounted piezoresistive sensors was evaluated as a new method to quantify physical activities (body posture and physical and locomotor activity) in relation to the sedative and cardiovascular effects of benzodiazepines, in an ambulatory study. In a double-blind, randomized, crossover study, 12 healthy men received either an oral dose of 2 mg lorazepam, 0.5 mg alprazolam, 1 mg alprazolam, or a placebo on 4 different days. By means of a portable digital recorder, each day 4 hours of continuous measurements of accelerometer signals and heart rate were performed in a living room in the hospital. Changes in subjective sleepiness were assessed at the beginning, halfway, and at the end of the recording period. A separate validation study of the ambulatory environment was performed in three subjects, in which computer classification of activities based on accelerometry was compared with visual evaluation of simultaneously recorded videotapes. In our validation study, comparison of the computer classification with visual analysis based on videotapes revealed an overall agreement for spontaneous and standardized activities of 88% and 96%, respectively. In our pharmacological study, the subjects spent more time in the lying position (p < 0.01) and less time in the sitting position (p < 0.01) after benzodiazepine administration; the effects were strongest for lorazepam. Motility during static activities was reduced (p < 0.025), with motility after lorazepam administration being lowest. Both lorazepam and alprazolam (0.5 and 1 mg) increased subjective sleepiness (p < 0.01). On average, lorazepam induced an overall increase in mean heart rate of about 6%, whereas alprazolam reduced mean heart rate by 2% versus placebo (p <0.01); the effects were not dependent on posture. The validation study showed that accelerometry forms a reliable method to quantify aspects of normal daily activities. Our pharmacological study revealed that quantification of body postures, physical activity, and motility by means of ambulatory accelerometry proves to be an objective and promising tool to evaluate the psychological and cardiovascular effects of (psycho) pharmaca in relation to the postural and mobility activities of normal daily life. (J Clin Psychopharmacol 1997;17:202-207).

Departments of (Tulen, van Steenis, Pepplinkhuizen) Psychiatry, (Bussmann) Rehabilitation Medicine, and (Man in't Veld) Internal Medicine I, Erasmus University Rotterdam, Rotterdam, The Netherlands.

Received July 23, 1996; accepted after revision December 26, 1996.

Address requests for reprints to: J.H.M. Tulen, PhD, Department of Psychiatry, University Hospital Rotterdam-Dijkzigt. Dr. Molewaterplein 40, 3015 GD Rotterdam, The Netherlands.

Currently available portable digital recorders for mass storage of physiological signals make it feasible to study details of psychophysiological processes in real-life ambulatory situations. These technical developments open new avenues for psychopharmacological research: apart from obtaining information in a controlled laboratory setting, physiological measurements in a natural environment may provide additional insights when evaluating the effectiveness and side effects of pharmacological interventions in relation to normal daily functioning. Within this context, reliable monitoring of physical activities is essential for several reasons. From a cardiovascular perspective, interpretation of physiological data is often hampered by the profound effects that changes in body posture and locomotor activity have on ambulatory recorded cardiovascular signals. [1,2] Yet, these effects are relevant to quantify for reliable evaluation of the posture-dependent cardiovascular side effects of some types of drugs (e.g., orthostatic hypotension after tricyclic antidepressants). Furthermore, apart from functioning as a control for the interpretation of physiological signals, a detailed quantification of physical activities offers the possibility to study the effects of drugs on daily behavioral processes. Current subjective methods to assess aspects of daily activities (e.g., daily logs, questionnaires) often lack reliability and accuracy, [3] and, although quantification of overall 24-hour motor activity patterns by means of wrist actigraphy has proved to be useful in clinical research, [4] wrist actigraphy cannot be used to study details of posture-related behavioral activities.

Recent developments based on physical activity monitoring by means of body-mounted piezoresistive accelerometers show that it is possible to obtain objective and reliable estimates of body postures and physical activity in ambulatory situations. [5,6] By combining the signals of piezoresistive accelerometers on the trunk and the upper legs, it is feasible to discriminate both static (e.g., postures lying, sitting, standing) and dynamic (e.g., walking, cycling) activities. [5,6] An Activity Monitor has been developed that comprises four accelerometers, a portable data recorder, and additional computer analysis software. With this monitor, several validation studies have been performed, showing high percentages of agreement (85-90%) between the automatic computer classification of the ambulatory accelerometry signals and visual analysis of simultaneously recorded videotapes. [7]

The aim of this study was twofold: first, to investigate accelerometry as a method to evaluate drug-induced physiological effects while controlling for body posture and physical activity and second, to explore if activity monitoring by means of accelerometry is sufficiently sensitive to reflect effects of pharmacological manipulations on daily physical activities. Special attention was paid to the validation of the method for our "ambulatory" psychophysiological setting. Because there are presently no quantitative data available on the anxiolytic and sedative effects of benzodiazepines on daily behavioral functioning, we used accelerometry to assess the effects of alprazolam and lorazepam on subjective sleepiness and cardiovascular functioning, in relation to daily locomotor activities. Effects of two oral doses of alprazolam (0.5 and 1 mg) were compared with one oral dose of lorazepam (2 mg), in a placebo-controlled design. Alprazolam was chosen because several studies suggest that alprazolam specifically may attenuate adrenomedullary activity, [8,9] but does not seem to influence cardiac vagal (parasympathetic) tone. [10] In addition, it has been suggested that alprazolam may induce less sedation in comparison with clinically equivalent anxiolytic doses of other benzodiazepines. [11,12] Lorazepam was used as an active control: lorazepam may attenuate sympathetic nervous system activity and influence cardiac vagal tone, [13] and at an oral dose of 2 mg, lorazepam clearly induces sedation. [14]

Back to Top | Article Outline


Subjects, procedures, and design

In a double-blind, randomized, placebo-controlled, crossover study, 12 healthy male volunteers (mean age: 22 years, range 19-25) received either an oral dose of 2 mg lorazepam, 0.5 mg alprazolam, 1 mg alprazolam, or a placebo on 4 different days, each 5 days apart. The procedures were approved by the Medical Ethical Committee of the University Hospital Rotterdam Dijkzigt. The subjects were paid for their participation and gave written, informed consent before the study. The experimental days lasted from 8:15 a.m. to 5:00 p.m. At 8:15 a.m., the subjects received the oral doses of alprazolam, lorazepam, or placebo. During the morning part of the experiment, cardiovascular and catecholaminergic responses were studied during a standardized laboratory schedule of supine rest, active standing, and mental challenge. [15] In this article, we present data for the afternoon sessions of the study, which comprised ambulatory measurements of heart rate and accelerometry in a living room in the hospital where the subjects could move around freely, study, relax, or sleep. Thus, spontaneous body movements, postural changes, and heart rate were measured for a period of 4 consecutive hours (from 1:00 p.m. to 5:00 p.m.). At 1:00 p.m., 3:00 p.m., and 5:00 p.m., a questionnaire regarding subjective sedation was completed.

Back to Top | Article Outline

Measurements and analysis

Ambulatory measurements of accelerometer signals and electrocardiogram (ECG). Data acquisition was done by means of a portable digital recorder (Vitaport System; TEMEC Instruments, Kerkrade, The Netherlands), which was carried on a belt around the waist. To quantify static and dynamic activities, four uniaxial 3-g piezoresistive accelerometers were used (IC-3031). The accelerometer signals are a mixture of a component of the gravitational force (giving absolute angle information) and a component of the actual acceleration of the sensor. [7] Two sensors were mounted with double-sided tape on the skin over the sternum: in the upright standing position, one sensor being sensitive parallel to the field of gravity (the longitudinal or Y-axis) and one sensor attached perpendicular to the Y-axis sensor, sensitive along the sagittal or X-axis. The two sensors on the upper legs, placed approximately halfway between the spina iliaca anterior superior and the upper side of the patella, were sensitive along the sagittal or X-axis. The sensors are small sized (2 x 1.5 x 0.5 cm), require low power consumption, and are mechanically robust. The accelerometer signals were stored digitally on the Vitaport recorder at a sample frequency of 16 Hz. Heart rate (beats per minute) was derived from a precordial ECG lead. The ECG signal was preprocessed in the Vitaport recorder (R-wave triggering, 2-ms accuracy), transposed to heart rate time series, and stored at a sample frequency of 4 Hz.

Analysis of accelerometer signals. From each of the four accelerometer signals, two additional signals were derived: a low-pass filtered (LPF) (0.5 Hz) signal and a high-pass filtered (0.5 Hz), rectified, and smoothed (HRF) signal. A range of preset values for each sensor was employed to determine each static activity (lying back, lying side, lying prone, sitting, standing) on the basis of the four LPF signals, whereas the HRF signals of the legs served to discriminate between static and dynamic activities. Typical examples of the raw accelerometer signals during different body postures are depicted in Figure 1. The minimal duration of the static and dynamic activities was set at 5 seconds. The following output parameters were computed for each 4-hour period: the total time spent in each static activity (lying back, lying side, lying prone, sitting, or standing) or dynamic activity (undifferentiated) and the total number of activity (static or dynamic) periods. Motility during each static activity was defined as the mean of the HRF acceleration signal of all episodes per static activity: this signal reflects the acceleration variability; the sum of the motility values of the four sensors was used to reflect total body motility. The time spent in each activity was computed as a percentage of the total time of the recording.

Figure 1

Figure 1

Analysis of heart rate. Mean heart rate (beats per minute) was computed per static and dynamic activity category. Movement artifacts, R-wave detection failures, and technical quality of the recordings were evaluated both by means of visual analysis and predefined computer settings. Because of R-wave triggering problems, data of one subject were discarded.

Subjective sedation. Changes in sleepiness were assessed by means of the Dutch translation of the Stanford Sleepiness Scale [16] at 1:00 p.m., 3:00 p.m., and 5:00 p.m.

Back to Top | Article Outline

Accelerometry: Validation study

Three healthy male volunteers (age range: 19-24 years) participated in a separate study to evaluate the validity of the Activity Monitor for our specific "ambulatory" environment (the living room in the hospital). The subjects were studied during two sessions each (a placebo and a lorazepam session, double-blind administration) while measuring physical activities and body postures by means of accelerometry, in combination with video monitoring. The timing of the measurements (afternoon) and the procedures were similar to the ones used in the pharmacological study. Thus, spontaneous activities were monitored for a period of 4 hours, but a 15- to 20-minute standardized activity protocol (in which the subjects were instructed to perform various postural changes and locomotor activities according to a predefined schedule) was added to each session. Per second, the classification of activities according to the Activity Monitor was compared with the classification of the visual analysis of the videotapes, for the spontaneous (total: 3 x 2 x 4 = 24 hours) and standardized parts (total: 3 x 20 x 2 = 120 minutes = 2 hours) of the recordings. Details of this validation study are presented elsewhere (Bussmann JBJ, Tulen JHM, van Herel ECG, Stam HJ. Quantification of physical activities by means of ambulatory accelerometry: a validation study, submitted).

Statistics. Multivariate analyses of variance for repeated measurements were used to establish the effects of the benzodiazepines on the activity parameters, mean heart rate, and subjective sedation. A p value of < 0.02 was considered to indicate a significant effect. For the validation study, apart from the overall percentage of agreement between the visual and computer classifications, sensitivity percentage was calculated as the number of agreements between video and activity monitor (per video category) divided by the total number of video observations (per video category) * 100, and the predictive value percentage was computed as the number of agreements between video and activity monitor (per activity monitor category) divided by the total number of computer observations (per activity monitor category) * 100.

Back to Top | Article Outline


Activity monitoring: Validation study

Comparison of the computer classification with the analysis based on videotapes revealed a 96% agreement for the standardized activities and a 88% agreement for the spontaneous activities (Table 1). The sensitivity per activity category was for the standardized part of the recordings always higher than or equal to 95%; for the spontaneous part of the recordings, the static activities showed a sensitivity of at least 85% and for movement this was 58%. The predictive value per activity category ranged, for the standardized part, from 82 to 100%. For the spontaneous part of the recording, these values ranged from 69 to 99% for the static activities, whereas for movement a predictive value of 64% was observed. The predictive value of lying back was influenced by the classification of the Activity Monitor of sitting flopped in an easy chair with feet on a coffee Table aslying back; this occurred in one subject. When, for this subject, sitting flopped in an easy chair was redefined for the visual analysis as lying back, the overall predictive value for lying back increased from 69 to 100% (range 100-100), the sensitivity of sitting increased from 85 to 99% (range 99-100), and the overall percentage of agreement for spontaneous activities increased from 88 to 99% (range 97-100).

Table 1

Table 1

Back to Top | Article Outline

Pharmacological study

Subjective sedation. Sleepiness was significantly increased after alprazolam (0.5 and 1 mg) and lorazepam administration compared with placebo (p < 0.01); the effect seemed strongest at 1:00 p.m. and showed a significant time-dependent decline from 1:00 p.m. to 5:00 p.m. (Table 2, p < 0.001). Responses to 1 mg alprazolam seemed only marginally larger than after 0.5 mg alprazolam, whereas no clear differences were observed between 1 mg alprazolam and 2 mg lorazepam.

Table 2

Table 2

Body postures and locomotor activity. Analysis of spontaneous activities during the 4-hour recording period during placebo revealed that the subjects spent most of their time in the sitting position (76%; Table 2), with 11% of the time lying and 6% of the time standing. Dynamic movements occurred only during 7% of the total recorded time period. Five to 9 hours after oral benzodiazepine administration, the subjects spent significantly more time in the lying position (p < 0.01; lorazepam highest: 35%) and significantly less time in the sitting position (p <0.01; lorazepam lowest: 53%) (Table 2). Lorazepam and alprazolam had no effect on the time spent in static standing, general dynamic activities, or the total number of activity periods. After benzodiazepine administration, total body motility during static activities was reduced (p < 0.025), with motility after lorazepam administration being lowest (Table 2). Dose-related effects of alprazolam were not observed.

Ambulatory heart rate and physical activities. Mean heart rate showed a significant increase from supine to sitting (+9.6%), to standing (+18.5%), and to dynamic body movements (+6.6%) (p < 0.001) (Table 2). On average, lorazepam induced an overall increase in mean heart rate of about 6% and alprazolam a minor decrease in mean heart rate of 2%, versus placebo (p < 0.003); the effects were not dependent on posture. A clear dose-related effect of alprazolam was not observed.

Back to Top | Article Outline


Our data indicate that body posture and mobility during normal daily life can be quantified reliably by means of ambulatory accelerometry: in the spontaneous and standardized part of our validation study, the overall agreement between automatic computer classification and visual analysis of videotapes was 88% and 96%, respectively. Other validation studies in different research settings have provided similar (85-90%) high percentages of agreement. [7] Particularly, the (static) postural positions were accurately detected; further refinement of the Activity Monitor regarding differentiation and validation of dynamic activities is under study.

Back to Top | Article Outline

Pharmacological study: Effects of benzodiazepines

Sedation and physical activities. Subjective sleepiness increased to a similar extent after 1 mg alprazolam and 2 mg lorazepam, whereas smaller effects were observed after 0.5 mg alprazolam. The effects were strongest at 1:00 p.m.; a gradual recovery occurred during the recording period. The increase in subjective sedation did not reflect effects of a postlunch dip, because this effect was already present during the morning part of the study. [15] Increases in subjective sedation after 0.5 and 1 mg alprazolam have been reported before [11,12]; similarly, subjective sedation has been found to be increased after 2 mg lorazepam. [14] Although 1 mg alprazolam and 2 mg lorazepam equally increased subjective sedation, the effects of the benzodiazepines on body posture and motility were, on average, strongest for lorazepam. After lorazepam administration, the subjects spent more time in the lying position and had lower total body motility than after alprazolam administration or placebo, although this last effect was only marginally significant. Nevertheless, these data indicate that objective indices of postural and mobility activities and subjective indices of sedation may provide discrepant findings. The muscle-relaxant effects of the benzodiazepines, which may occur independently of the sedation-inducing effects, [17] probably contributed to the findings. This illustrates the necessity to combine both types of measurements when evaluating the effects of benzodiazepines in relation to daily functioning.

Ambulatory heart rate and physical activities Five to 9 hours after oral benzodiazepine administration, mean heart rate, on average, increased slightly after lorazepam and decreased slightly after alprazolam (both dose levels); both effects were not dependent on posture. Our findings are in agreement with studies reporting a lack of effect of alprazolam on heart rate during supine rest [9] or a decrease in heart rate when the subjects were in a sitting position. [18] Increased heart rate after (1-2.5 mg) oral lorazepam administration has been reported. [19] Further ambulatory studies with combined measurements of heart rate and blood pressure are needed to make conclusive statements regarding the possible differences in cardiovascular regulatory mechanisms between alprazolam and lorazepam.

Back to Top | Article Outline

Summary and conclusion

Ambulatory monitoring of physiological and behavioral processes is an intriguing option for both clinical and research purposes in a variety of psychological and medical disciplines. As a first application and validation, the present study substantiates the potential advantages of combining physiological measurements with accelerometry in ambulatory pharmacological research. Although our application was restricted to evaluation of the acute effects of benzodiazepines in a "controlled" ambulatory environment, the monitor offers more extensive possibilities for pharmacological research because the accelerometer sensors and the compact recording system with the extensive storage device can be worn without any discomfort for several days and nights in the natural environment of the subjects. Details of spontaneous dynamic activities and postural transitions, whether or not in relation to cardiovascular processes, may be informative to analyze after acute and chronic administration of drugs that are known to induce postural hypotension or tachycardia (antidepressants; neuroleptics). More than a mere quantification and inventory of potential side effects, the study of these processes provides objective indices of the patient's actual behavior before and after therapeutic interventions. Objective and accurate assessment of the duration and intensity of daily physical activities can help to describe relationships between mobility and acute or chronic disease states (cardiovascular, psychopathological, movement disorders) [20] or behavioral processes [21] so that a more reliable evaluation of the effects of (pharmaco)therapeutic interventions in relation to normal daily functioning can be made.

Back to Top | Article Outline


This study was supported by Upjohn-Nederland. The initial phase of development of the Activity Monitor occurred within the EUREKA project DYNAPORT, which was partly financed by the Dutch Ministry of Economic Affairs. Frans van den Berg assisted with the experiment.

Back to Top | Article Outline


1. Clarck LA, Denby L, Pregibon D, Harshfield GA, Pickering TG, Blank S, Laragh JH. A quantitative analysis of the effects of activity and time of day on the diurnal variations of blood pressure. J Chronic Dis 1987;40:671-81.
2. Pickering TG. Ambulatory monitoring and blood pressure variability. London: Science Press Ltd, 1991.
3. Patterson SM, Krantz DS, Montgomery LC, Deuster PA, Hedges SM, Nebel LE. Automated physical activity monitoring: validation and comparison with physiological and self-report measures. Psychophysiology 1993;30:296-305.
4. Raoux N, Benoit O, Dantchev N, Denise P, Franc B, Allilaire JF, Widlocher D. Circadian pattern of motor activity in major depressed patients undergoing antidepressant therapy: relationship between actigraphy measures and clinical course. Psychiatry Res 1994;52:85-98.
5. Veltink PH, Bussmann JBJ, de Vries W, Martens WLJ, van Lummel RC. Discrimination of dynamic activities using accelerometry. In: Veltink PH, van Lummel RC, eds. Dynamic analysis using body fixed sensors. Amsterdam, 2nd World Congress of Biomechanics, 1994:3-7.
6. Bussmann JBJ, Veltink PH, Koelma F, van Lummel RC, Stam HJ. Ambulatory monitoring of mobility-related activities: the initial phase of the development of an activity monitor. Eur J Phys Med Rehabil 1995;5:2-7.
7. Bussmann JBJ, Reuvekamp PJ, Veltink PH, Martens WLJ, Stam HJ. Validity of an instrument for ambulatory measurement of mobility activities. Submitted.
8. Breier A, Davis O, Buchanan R, Listwak GJ, Holmes C, Pickar D, Goldstein DS. Effects of alprazolam on pituitary-adrenal and catecholaminergic responses to metabolic stress in humans. Biol Psychiatry 1992;32:880-90.
9. Stratton JR, Halter JB. Effect of a benzodiazepine (alprazolam) on plasma epinephrine and norepinephrine levels during excercise stress. Am J Cardiol 1985;56:136-9.
10. McLeod DR, Hoehn-Saric R, Porges SW, Zimmerli WD. Effects of alprazolam and imipramine on parasympathetic cardiac control in patients with generalized anxiety disorder. Psychopharmacology 1992;107:535-40.
11. Dawson ED, Jue SG, Brodgen RN. Alprazolam, a review of its pharmacological properties and efficacy in the treatment of anxiety and depression. Drugs 1984;27:132-47.
12. Greenblatt DJ, Harmatz JS, Dorsey C, Shader RI. Comparative single-dose kinetics and dynamics of lorazepam, alprazolam, prazepam, and placebo. Clin Pharmacol Ther 1988;44:326-34.
13. Tulen J, Mulder G, Pepplinkhuizen L, Man in 't Veld A, van Steenis H, Moleman P. Effects of lorazepam on cardiac vagal tone during rest and mental stress: assessment by means of spectral analysis. Psychopharmacology 1994;114:81-9.
14. Preston GC, Ward CE, Broks P, Traub M, Stahl SM. Effects of lorazepam on memory, attention and sedation in man: antagonism by Ro 15-1788. Psychopharmacology 1989;97:222-7.
15. van den Berg F, Tulen JHM, Boomsma F, Noten JBGM, Moleman P, Pepplinkhuizen L. Effects of alprazolam and lorazepam on catecholaminergic and cardiovascular activity during supine rest, mental load, and orthostatic challenge. Psychopharmacology 1996;128:21-30.
16. Hoddes E, Zarcone V, Smythe H, Phillips R, Dement W. Quantification of sleepiness: a new approach. Psychophysiology 1973;10:431-6.
17. Costa E, ed. The benzodiazepines. From molecular biology to clinical practice. New York: Raven Press, 1983.
18. Rohrer T, von Richthofen V, Schulz C, Beyer J, Lehnert H. The stress-, but not corticotropin-releasing hormone-induced activation of the pituitary-adrenal axis in man is blocked by alprazolam. Horm Metab Res 1994;26:200-6.
19. File SE, Lister RG. A comparison of the effects of lorazepam with those of propranolol on experimentally-induced anxiety and performance. Br J Clin Pharmacol 1985;19:445-51.
20. Tyron WW. Activity measurement in psychology and medicine. New York: Plenum Press, 1991.
21. Montoye HJ, Taylor HL. Measurement of physical activity in population studies: a review. Hum Biol 1984;56:195-216.
© Williams & Wilkins 1997. All Rights Reserved.