Research in Veterinary Science 93 (2012) 412–416

Contents lists available at SciVerse ScienceDirect

Research in Veterinary Science

journal homepage: www.elsevier .com/locate / rvsc

Accelerometer validity and placement for detection of changes in physical activityin dogs under controlled conditions on a treadmill

Timothy Preston a,1, Wendy Baltzer a,⇑, Stewart Trost b,2

a Oregon State University College of Veterinary Medicine, Dept. Clinical Sciences, 105 Magruder Hall, Corvallis, OR 97331, United Statesb Oregon State University, College of Public Health and Human Sciences, Dept. Exercise Physiology, 203D Women’s Building, Corvallis, OR 97331, United States

a r t i c l e i n f o

Article history:Received 28 April 2011Accepted 12 August 2011

Keywords:DogAccelerometryTreadmillExercise intensity

0034-5288/$ - see front matter Published by Elsevierdoi:10.1016/j.rvsc.2011.08.005

⇑ Corresponding author. Tel./fax: +1 541 737 4812.E-mail addresses: timothy.prestion@oregonstate

baltzer@oregonstate.edu (W. Baltzer), stewart.trost@o1 Tel.: +1 541 737 4835; fax: +1 541 737 4818.2 Tel.: +1 541 737 5941.

a b s t r a c t

The objective of the research was to determine the optimal location and method of attachment for accel-erometer-based motion sensors, and to validate their ability to differentiate rest and increases in speed inhealthy dogs moving on a treadmill. Two accelerometers were placed on a harness between the scapulaeof dogs with one in a pouch and one directly attached to the harness. Two additional accelerometers wereplaced (pouched and not pouched) ventrally on the dog’s collar. Data were recorded in 1 s epochs withdogs moving in stages lasting 3 min each on a treadmill: (1) at rest, lateral recumbency, (2) treadmillat 0% slope, 3 km/h, (3) treadmill at 0% slope, 5 km/h, (4) treadmill at 0% slope, 7 km/h, (5) treadmill at5% slope, 5 km/h, and (6) treadmill at 5% slope, 7 km/h. Only the harness with the accelerometer in apouch along the dorsal midline yielded statistically significant increases (P < 0.05) in vector magnitudeas walking speed of the dogs increased (5–7 km/h) while on the treadmill. Statistically significantincreases in vector magnitude were detected in the dogs as the walking speed increased from 5 to7 km/h, however, changes in vector magnitude were not detected when activity intensity was increasedas a result of walking up a 5% grade. Accelerometers are a valid and objective tool able to discriminatebetween and monitor different levels of activity in dogs in terms of speed of movement but not in energyexpenditure that occurs with movement up hill.

Published by Elsevier Ltd.

1. Introduction

An accelerometer measures accelerations and decelerations in1–3 planes. Triaxial accelerometers measure movement in threeplanes, vertical, mediolateral, and craniocaudal. The use of acceler-ometers has been validated as an objective and representativemeasure of spontaneous activity in humans and dogs (Welket al., 2004; Hansen et al., 2007; Chan et al., 2005; Brown et al.,2010a; Welk, 2002). The potential benefits of accelerometrycompared with traditional monitoring of activity in companionanimals include excellent owner and dog compliance as well asthe ability to obtain objective data unbiased by owner and re-searcher observations. Accelerometry data has been determinedto be a consistent estimate of routine activity in dogs, and the abil-ity to monitor activity counts and changes in activity intensity overlong periods of time in a home environment makes accelerometermonitoring an invaluable research tool in veterinary medicine(Hansen et al., 2007; Brown et al., 2010a; Dow et al., 2009).

Ltd.

.edu (T. Preston), wendy.regonstate.edu (S. Trost).

In human research, accelerometry has been a useful tool inevaluating energy expenditure and physical activity in healthyand diseased people of varying ages (Hansen et al., 2007; Copelandand Esliger, 2009; Kodama et al., 2002). Validation of accelerometeruse in human research has relied primarily on comparison to mea-surement of energy expenditure using oximetry or calorimetry, butalso has validated as a method to detect changes in the speed ofwalking or running (Yokoyama et al., 2002; Kumahara et al.,2004). The vector magnitude recorded by triaxial accelerometershas been found to have a direct linear relationship to the speed ofthe treadmill at speeds less than 10 km/h (Rowlands et al., 2007;Brage et al., 2003). In human studies, placement of the accelerome-ter is most commonly on the trunk (hip) rather than an extremitysuch as the ankle or wrist, in order to detect changes in the subject’sactivity rather than position of the extremity (Ward et al., 2005).

Accelerometers in companion animal research have been usedto objectively determine responses to therapy by monitoring adog’s or cat’s activity, either at home or in a laboratory setting(Yamada and Tokuriki, 2000; Lascelles et al., 2007, 2008). Bychanging the threshold of the accelerometer, differences in thetype of movement recorded, such as one body part or movementof the entire dog in a cage, can be detected (Yamada and Tokuriki,2000). More recent work in companion animals has investigatedthe optimal sampling intervals for at home monitoring in dogs

T. Preston et al. / Research in Veterinary Science 93 (2012) 412–416 413

and at least 7 days of continuous monitoring was found to berequired for the least variable estimate of a dog’s activity (Dowet al., 2009). The effect of body conformation and age of the dogon the accelerometer measurements that are recorded has alsobeen studied (Brown et al., 2010b). The age of the animal and bodyconformation did not affect accelerometer counts when the dogswere resting, walking, or trotting laps, however, as the dog in-creased in body weight and age, there was a decrease in countswhen the dogs were trotted up and down stairs. This study didnot specify, however, the exact rate at which the dogs were movingand whether the accelerometers could detect a change in intensityof the activity of the dogs (Dow et al., 2009). The general intensityof activity has been investigated in dogs with the activity charac-terized by the dogs’ owners as sedentary, light intensity physicalactivity indoors, light to moderate intensity physical activity out-doors and vigorous physical activity outdoors (Yam et al., 2011).The accelerometer was able to detect differences in those catego-ries in dogs with the reliability of detection in changes in activitygreatest with a 7-day interval of monitoring (Yam et al., 2011).Clearly, general activity of a dog can be assessed using an acceler-ometer, however, the ability of an accelerometer to detect changesin the speed of dogs moving in one direction or up an incline hasnot been previously investigated.

The establishment of a gold standard methodology with param-eters for the use of accelerometers in companion animals is essen-tial for the development of clinically relevant research withreproducible and meaningful results. While previous reports havebeen able to detect differences in activity such as resting, walking,and trotting in dogs with accelerometers, establishment of theability of accelerometers to detect differences in the speed ofmovement of dogs under controlled conditions has yet to bereported. The objective of this study was to validate the placementof accelerometers on dogs during controlled speeds of movementand to determine whether an accelerometer could detect differ-ences in intensity of activity of dogs walking, trotting, and runningon a treadmill with and without an uphill grade or incline. To theauthors’ knowledge, monitoring of treadmill activity in dogs withaccelerometry has not been previously reported.

2. Materials and methods

2.1. Dog population

The study protocol was approved by the institution’s AnimalCare and Use Committee. The study population included sixhealthy dogs aged from 2 to 9 years old (mean ± SD, 4.8 ± 2.4) withbody weights ranging from 14.7 to 32 kg (mean ± SD, 24.9 ± 6.7).Four of the dogs were spayed females and two were neuteredmales and consisted of the following breeds: Labrador retriever(Hansen et al., 2007), Goldendoodle, Weimeraner, Border collie,and mixed breed. All of the dogs were examined by a boardedsurgeon and determined to be free from orthopedic disease. Allof the dogs were owned by staff at our institution.

2.2. Accelerometer

The GT3X is a solid-state triaxial accelerometer measuring3.8 cm � 3.7 cm � 1.8 cm and weighs just 27 g. It is able to collectmotion data on three axes for analysis and includes an inclinometer,which is designed to be able to determine the subject’s position inspace. Data is recorded in user-defined epochs using the proprietarysoftware, ActiLife5 v5.4.0, Actigraph, Pensacola, FL. Acceleration isrecorded with precision and consistency, measuring and recordingtime varying accelerations ranging in magnitude from approxi-mately 0.05 to 2.5 G’s. The data is digitized using a 12-bit analog-

to-digital converter at a rate of 30 Hertz (Hz). The digitized data isthen limited by a filter that narrows the frequency range to between0.25 and 2.5 Hz in an effort to reduce noise and increase accuracy ofthe device, and then stored on 16 MB of solid state memory capableof storing approximately 400 days’ worth of data. The GT3X has alithium-ion battery which limits its consistent use to 20 days.

2.3. Data collection

All dogs were fitted with the same standard adjustable nylonharness, which was adjusted to size of the dog accordingly. Eachdog was individually fitted with 8 tri-axial accelerometers (Acti-graph GTX3, Actigraph LLC, Pensacola, FL) prior to initiation ofexercise, with each accelerometer assigned a unique number toensure the same placement on each dog after each trial’s data weredownloaded. Each accelerometer was programmed to record 1-sepochs for a period of 3-min. The accelerometer used in this studyhas been previously validated in dogs and is manufactured for useon dogs and humans (Yam et al., 2011). No recommendations aremade by the manufacturer as to whether the clip or pouch shouldbe used on dogs and both are made by the manufacturer of theaccelerometer (Actigraph LLC, Pensacola, FL). One-second epochsallowed for greater resolution of activity monitoring throughouteach stage of trial and level of activity.

Two accelerometers (Actigraph GTX3, Actigraph LLC, Pensacola,FL) were placed on the harness between the scapulae along thedorsal midline adjacent to one another and equidistant from themidline approximately 2 cm apart. Of these two accelerometersalong the dorsal midline, one was placed within a pouch (BP305,Actigraph, Pensacola, FL). The pouch was fitted with a belt-loopand threaded onto the harness to sit along the dorsal midline.The adjacent dorsal midline accelerometer that was not pouchedwas attached to the harness using the plastic clip supplied fromthe manufacturer (Actigraph, Pensacola, FL). Two additional accel-erometers (Actigraph GTX3, Actigraph LLC, Pensacola, FL) werethen placed in a similar fashion on the dog’s collar – positionedadjacent to one another and equidistant from the midline (approx-imately 1 cm from midline) so as to sit symmetrically straddlingthe trachea ventrally on the dog’s neck.

Epochs were defined as 1-s and recording was continuousthroughout each stage once activated. Once the accelerometerswere all attached to the harness and activated, the dogs wereplaced through a series of trial stages, each 3-min in length andalways in the same order as follows:

Stage 1: During this stage the dogs were required to lie in rightlateral recumbency for 3-min to simulate sleep or restphases of daily activity. Minimal to no restraint wasused to ensure accuracy. If the animal broke the proto-col, the stage was repeated in its entirety.

Stage 2: Dogs were placed onto a custom treadmill for dogs(SM3701-X2, Fit Fur Life Ltd., Surrey, UK) at 0% slopeand walked at 3 km/h for 3 min.

Stage 3: Dogs were walked on the treadmill at 0% slope at5 km/h for 3 min.

Stage 4: Dogs were trotted on the treadmill at 0% slope at 7 km/h for 3 min.

Stage 5: Dogs were walked on the treadmill at 5% slope at5 km/h for 3 min.

Stage 6: Dogs were trotted on the treadmill at 5% slope at 7 km/h for 3 min.

At the successful completion of all stages, the harness wasremoved from the subject and the accelerometers connected toa computer via the supplied USB cable. Proprietary software

Fig. 1. Collar and clip configuration (left) and collar and pouch configuration (right) data results. Data were recorded in 1-s epochs with dogs moving in stages lasting 3 mineach on a treadmill: (1) at rest, lateral recumbency, (2) treadmill at 0% slope, 3 km/h, (3) treadmill at 0% slope, 5 km/h, (4) treadmill at 0% slope, 7 km/h, (5) treadmill at 5%slope, 5 km/h, and (6) treadmill at 5% slope, 7 km/h. Vector magnitude did not change significantly with increase in the dogs’ speed on the treadmill when the accelerometerswere clipped or pouched on the dogs’ collars (P > 0.05). Increasing the dogs’ work by increasing the grade of the treadmill from 0% to 5% slope did not significantly alter thevector magnitude recorded by the accelerometers (P > 0.05). The line inside the box represents the median value with the upper and lower limits of the box representing the25th and 75th percentiles for the data. The upper and lower whiskers represent the highest and lowest datum within 1.5 interquartile ranges above the 75th percentile orbelow the 25th percentile, respectively. The diamond symbol represents the mean vector magnitude and the circles represent datum 1.5 interquartile ranges above the 75thpercentile and below the 25th percentile at each stage.

414 T. Preston et al. / Research in Veterinary Science 93 (2012) 412–416

(ActiLife5 v5.5.0, Actigraph, Pensacola, FL) was used to downloadall information from the accelerometers for analysis and pro-gramming. After data from each accelerometer were downloadedand secured, a numerical suffix was added to correlate the datasets to each subject’s trials. The accelerometers were then re-placed at their assigned site onto the harness and reused forthe next subject.

2.4. Statistics

All statistical analyses were performed using SAS (Version 9.2,SAS Institute Inc, Cary, NC). Descriptive statistics and box plotswere generated to qualitatively examine the impact of monitorplacement and attachment on GT3X output. For each placementand attachment combination, differences in GT3X vector magni-tude across the six activity trials were evaluated for statisticalsignificance using one-way repeated measures ANOVA. Pre-planned single degree freedom contrasts were used to determineif GT3X vector magnitude increased significantly from the previ-ous trial. Statistical significance was set at an alpha level of 0.05.

3. Results

Box plots depicting the distribution of the vector magnituderecorded from the accelerometers for the six activity trials areshown in Figs. 1 and 2. Only the harness and pouch configura-tion provided accelerometer data that was not dramaticallyskewed and increased systematically with increasing speed oflocomotion. While all four placement/attachment configurationsdemonstrated some increase in vector magnitude as activity le-vel transitioned from rest, to walking, to trotting; only the har-ness with the accelerometer in a pouch yielded statisticallysignificant increases in vector magnitude as level walking speedincreased from 5 to 7 km/h, (P < 0.05). Regardless of placement/attachment configuration, vector magnitude did not increase sig-nificantly when activity intensity was increased as a result ofwalking on a 5% grade.

4. Discussion

Physical activity measurement in the dog’s home environmenthas the potential to objectively determine the activity of dogs clin-ically, without bias from the owner, attending veterinarian, orresearcher. Previously, the GT3X has been investigated to deter-mine if changes in habitual activity could be detected in dogs intheir home environment. This investigation found that the acceler-ometer could detect differences in the dogs’ behavior ranging fromsedentary to moderate activity to vigorous physical activity (Yamet al., 2011). Other studies have also been able to detect changesin activity in dogs over time, however, none have used a treadmillwith the speed, direction, and grade or incline of movementcontrolled prior to the results reported here (Hansen et al., 2007;Dow et al., 2009; Yamada and Tokuriki, 2000). In addition, theeffect of location and attachment of accelerometers on data valid-ity using standardized speeds and inclines on a treadmill has notbeen previously reported. Recent studies have shown thatalthough treadmill activity has been shown to be kinematicallydifferent in humans walking on treadmills versus the ground, thereis no significant difference between accelerometer readings withthe subject on a treadmill or on the ground (Riley et al., 2007;Vanhelst et al., 2009). Accelerometers are sensitive in determininglevels of activity in dogs and humans, and are also specific to thetype of activity such as sleep, vigor of physical activity, and evenmotor vehicle travel (Welk et al., 2004; Brown et al., 2010b; Woodet al., 2008; Weiss et al., 2010; Munneke et al., 2001; Pate et al.,2006; Maddocks et al., 2010).

Although all four placement/attachment configurations testedin this study demonstrated some increase in vector magnitude asactivity level transitioned from rest to 5–7 km/h; only the har-ness/pouch yielded statistically significant increases in vectormagnitude as level walking speed increased. Changes in activityhave been detected while dogs were wearing accelerometers invarious locations including their collar dorsally, the collar ven-trally, laterally on the thorax in a vest, the axilla in a vest, the ster-num in a vest, the ventral abdomen in a vest, around the humerusin a legging, and on the antebrachium in a legging (Hansen et al.,

Fig. 2. Harness and clip configuration (left), and harness and pouch configuration (right) data results. The stages are the same as for Fig. 1. Vector magnitude increasedsignificantly as the speed of the treadmill increased with the accelerometer on the dogs’ harness in a pouch from stages 1–4, P < 0.05, however, the accelerometer was unableto detect changes in the intensity of work performed by the dogs by increasing the grade of the treadmill from 0% to 5% in stages 5 and 6 (P > 0.05). See Fig. 1 for explanation ofsymbols.

T. Preston et al. / Research in Veterinary Science 93 (2012) 412–416 415

2007; Brown et al., 2010a; Yam et al., 2011). Spearman correlationsof 0.71–0.93 were reported for the detection of movement with theaccelerometer on any of the locations list above, however, onlyvisual assessment by an observer was used to assess the dogs’speed and direction of movement making the findings qualitativerather than objective (Hansen et al., 2007). Placement of the mon-itor on the dog’s collar is the most convenient location and mostfrequently used in previous research, but could result in falselyelevated activity counts if the dog’s leash is attached to the collaror if the dog shakes his head frequently while moving (Hansenet al., 2007; Dow et al., 2009; Yamada and Tokuriki, 2000; Yamet al., 2011). The results of the study reported here indicate thatchanges in speed on a treadmill are only detected when the accel-erometer is located in a pouch dorsally on a harness around thedog’s thorax. Based upon the data reported here, the use of a har-ness with the accelerometer located dorsally near the dog’s scapulawith the accelerometer in a pouch instead of tied to the harness isthe recommended method of placement and attachment in studieswhere the researcher requires as little as 2 km/h change in speedwithout any change in direction to be detected.

The difference between the pouch versus clip on the dog’sharness for detection of changes in speed was a surprising findingin this study. The pouch allowed the accelerometer to be placed ina pocket that closed tightly around the monitor and had a belt-looptype slot on the back that the harness was threaded through. Thispouch configuration allowed us to place the harness with thepouch snugly on the dog so that there was less movement of theharness, pouch and accelerometer when the dog was moving atfaster speeds. In addition, in the dorsal position on the harness,the pouch loop fit tightly around the harness strap so that, in thedorsal position between the dogs’ shoulders, the accelerometerdid not slide around the strap as the dog was moving. The clipwas plastic and simply attached to the accelerometer with adhe-sive tape, then was clipped to the harness. The clip did not fittightly to the harness strap and so did not prevent the accelerom-eter from sliding along the harness or collar when the dog movedat faster speeds, especially in positions other than between thedog’s shoulders dorsally, which may have explained why the clipconfiguration did not perform as well as the pouch. The harnesswas placed on each dog and individually adjusted so that therewas little movement of the harness strap dorsally on the dog, butit was not placed too tightly that it would obstruct movement ofthe dog when walking, trotting or running. The collar around the

dog’s neck had to be placed loose enough that the dog couldbreathe and move freely, especially during running. This may havebeen the reason that the harness-accelerometer provided statisti-cally different results with changes in speed of 2 km/h and thecollar-accelerometer configuration did not.

The triaxial accelerometer used in this study could not detectchanges in incline of the treadmill up to a 5% grade at speeds upto 7 km/h in dogs. The vector magnitude did not increase signifi-cantly when the dogs’ activity intensity was increased as a resultof walking or trotting on a 5% grade. Similar findings have beenreported in humans with accelerometers displaying a wide rangeof vector magnitudes to subjects walking on an incline (Terrieret al., 2001; Montoye et al., 1983). Researchers have begun to usealtimeters, barometers, and differential GPS along with accelerom-eters in an effort to determine alterations in energy expenditureand movement on an incline and decline that an accelerometercannot detect alone (Perrin et al., 2000; Yamazaki et al., 2009). Fur-ther investigations using these methods to detect changes inincline in dogs may be warranted.

Limitations of this study include the inherent size and confor-mational differences between breeds of dogs, limited points ofattachment of the accelerometer and small sample size. Furtherstudies with larger pooled data and individualized tailoring ofaccelerometer parameters are required to establish a veterinarygold standard methodology. Following the human developmentof accelerometry standards, we have begun by first determiningthat accelerometers in dogs were able to differentiate changes inthe vigor of activity and rest, and a useful position for placementof accelerometers in dogs to enable detection of changes in speedin one direction. Future studies will need to determine if acceler-ometers are able detect changes in energy expenditure, not justspeed and general level of activity. Once this is achieved, research-ers may be able to develop appropriate algorithms and data cut offpoints to accurately determine energy expenditure using acceler-ometers for a patient’s individual physical parameters. Althoughprevious studies have already used accelerometers to look atchanges in activity in response to therapy, none have correlatedthe activity to a known speed of movement prior to this report(Brown et al., 2010a; Lascelles et al., 2007).

Activity monitor use in veterinary research has shown a strongassociation between a variety of conditions and changes in levels ofactivity, but none, to the author’s knowledge, have determined thevalidity of accelerometers to detect changes in levels of activity

416 T. Preston et al. / Research in Veterinary Science 93 (2012) 412–416

under controlled conditions (Hansen et al., 2007; Brown et al.,2010a,b; Lascelles et al., 2008). In order to detect changes in speedin one direction in dogs, the recommended placement of the accel-erometer is in a pouch attached dorsally to a harness. Accelerom-eters are a useful and objective tool for the measurement ofactivity in dogs and they are able to distinguish between differentspeeds of walking and running up to 7 km/h under the controlledconditions of a treadmill.

References

Brage, S., Wedderkopp, N., Franks, P.W., Anderson, L.B., Frogerg, K., 2003.Reexamination of the validity and reliability of the CSA monitor in walkingand running. Medicine and Science in Sports and Exercise 35, 1447–1454.

Brown, D.C., Boston, R.C., Farrar, J.T., 2010a. Use of an activity monitor to detectresponse to treatment in dogs with osteoarthritis. Journal of the AmericanVeterinary Medical Association 237 (1), 66–70.

Brown, D.C., Michel, K.E., Love, M., Dow, C., 2010b. Evaluation of the effect ofsignalment and body conformation on activity monitoring in companion dogs.American Journal of Veterinary Research 71 (3), 322–325.

Chan, C.B., Spierenburg, M., Ihle, S.L., Tudor-Locke, C., 2005. Use of pedometers tomeasure physical activity in dogs. Journal of the American Veterinary MedicalAssociation 226 (12), 2010–2015.

Copeland, J.L., Esliger, D.W., 2009. Accelerometer assessment of physical activity inactive healthy older adults. Journal of Aging and Physical Activity 17 (1), 17–30.

Dow, C., Michel, K.E., Love, M., Brown, D.C., 2009. Evaluation of optimal samplinginterval for activity monitoring in companion dogs. American Journal ofVeterinary Research 70 (4), 444–448.

Hansen, B.D., Lascelles, B.D., Keene, B.W., Adams, A.K., Thomson, A.E., 2007.Evaluation of an accelerometer for at-home monitoring of spontaneousactivity in dogs. American Journal of Veterinary Research 68 (5), 468–475.

Kodama, Y., Tamakoshi, A., Nishizuka, T., Hirano, N., Kawamura, T., Ohno, Y., et al.,2002. Validity of an accelerometer for estimating daily physical activity –Comparison with a diary method. Nippon Koshu Eisei Zasshi 49 (7), 643–647.

Kumahara, H., Schutz, Y., Ayabe, M., Yoshioka, M., Yoshitake, Y., Shindo, M., et al.,2004. The use of uniaxial accelerometry for the assessment of physical-activity-related energy expenditure: a validation study against whole-body indirectcalorimetry. British Journal of Nutrition 91 (2), 235–243.

Lascelles, B.D., Hansen, B.D., Roe, S., DePuy, V., Thomson, A., Pierce, C.C., et al., 2007.Evaluation of client-specific outcome measures and activity monitoring tomeasure pain relief in cats with osteoarthritis. Journal of Veterinary InternalMedicine 21 (3), 410–416.

Lascelles, B.D., Hansen, B.D., Thomson, A., Pierce, C.C., Boland, E., Smith, E.S., 2008.Evaluation of a digitally integrated accelerometer-based activity monitor for themeasurement of activity in cats. Veterinary Anaesthesia and Analgesia 35 (2),173–183.

Maddocks, M., Petrou, A., Skipper, L., Wilcock, A., 2010. Validity of threeaccelerometers during treadmill walking and motor vehicle travel. BritishJournal of Sports Medicine 44 (8), 606–608.

Montoye, H.J., Washburn, R., Servais, S., Ertl, A., Wbster, J.G., Nagle, F.J., 1983.Estimation of energy expenditure by a portable accelerometer. Medicine andScience in Sports and Exercise 15, 403–407.

Munneke, M., de Jong, Z., Zwinderman, A.H., Tijhuis, G.J., Hazes, J.M., Vliet Vlieland,T.P., 2001. The value of a continuous ambulatory activity monitor to quantifythe amount and intensity of daily activity in patients with rheumatoid arthritis.Journal of Rheumatology 28 (4), 745–750.

Pate, R.R., Almeida, M.J., McIver, K.L., Pfeiffer, K.A., Dowda, M., 2006. Validation andcalibration of an accelerometer in preschool children. Obesity (Silver Spring) 14(11), 2000–2006.

Perrin, O., Terrier, P., Ladetto, Q., Merminod, B., Schutz, Y., 2000. Improvement inwalking speed prediction by accelerometry and altimetry, validated by DGPS.Medical and Biological Engineering and Computing 38 (2), 164–168.

Riley, P.O., Paolini, G., Della Croce, U., Paylo, K.W., Kerrigan, D.C., 2007. A kinematicand kinetic comparison of overground and treadmill walking in healthysubjects. Gait Posture 26 (1), 17–24.

Rowlands, A.V., Stone, M.R., Eston, R.G., 2007. Influence of speed and step frequencyduring walking and running on motion sensor output. Medicine and Science inSports and Exercise 39 (4), 716–727.

Terrier, P., Aminian, K., Schutz, Y., 2001. Can accelerometry accurately predict theenergy cost of uphill/downhill walking? Ergonomics 44 (1), 48–62.

Vanhelst, J., Zunquin, G., Theunynck, D., Mikulovic, J., Bui-Xuan, G., Beghin, L., 2009.Equivalence of accelerometer data for walking and running: treadmill versus onland. Journal of Sports Sciences 27 (7), 669–675.

Ward, D.S., Evenson, K.R., Vaughan, A., Rodgers, A.B., Troiano, R.P., 2005.Accelerometer use in physical activity research: best practices and researchrecommendations. Medicine and Science in Sports and Exercise 37,S582–S588.

Weiss, A.R., Johnson, N.L., Berger, N.A., Redline, S., 2010. Validity of activity-baseddevices to estimate sleep. J Clin Sleep Med 6 (4), 336–342.

Welk, G., 2002. Physical Activity Assessments for Health-Related Research, First ed.Human Kinetics Inc., Champaign, IL.

Welk, G.J., Schaben, J.A., Morrow Jr, J.R., 2004. Reliability of accelerometry-basedactivity monitors: a generalizability study. Medicine and Science in Sports andExercise 36 (9), 1637–1645.

Wood, A.C., Kuntsi, J., Asherson, P., Saudino, K.J., 2008. Actigraph data are reliable,with functional reliability increasing with aggregation. Behavior ResearchMethods 40 (3), 873–878.

Yam, P.S., Penpraze, V., Young, D., Todd, M.S., Cloney, K.A., Houston-Callaghan, K.A.,et al., 2011. Validity, practical utility and reliability of Actigraph accelerometryfor the measurement of habitual physical activity in dogs. Journal of SmallAnimal Practice 52 (2), 86–92.

Yamada, M., Tokuriki, M., 2000. Spontaneous activities measured continuously byan accelerometer in beagle dogs housed in a cage. Journal of Veterinary MedicalScience 62 (4), 443–447.

Yamazaki, T., Gen-No, H., Kamijo, Y., Okazaki, K., Masuki, S., Nose, H., 2009. A newdevice to estimate VO2 during incline walking by accelerometry and barometry.Medicine and Science in Sports and Exercise 41 (12), 2213–2219.

Yokoyama, Y., Kawamura, T., Tamakoshi, A., Noda, A., Hirai, M., Saito, H., et al., 2002.Comparison of accelerometry and oxymetry for measuring daily physicalactivity. Circulation Journal 66 (8), 751–754.