Validation study of Polar V800 accelerometer

Introduction

Since the middle of the last century the relationship between health, energy expenditure (EE) and physical activity (PA) has been widely studied (1). Regular PA is associated with numerous physical and mental health benefits (2-5). In addition, there is a dose-response relationship between life expectancy and the amount of PA accumulated (6), showing that those who engage weekly in ≥450 min of PA have a higher life expectancy than those who do not (7).

However, it is not enough to meet PA recommendations to guarantee a good metabolic health (8) and also sedentary behavior (SB) has to be taken into consideration (9,10). Researchers and health professionals have different methods to estimate EE, PA and SB levels of their patients/clients. These methods can be classified as subjective and objective (11). The subjective methods (questionnaires, activity diaries, and interviews) are easily applicable but they have a limited validity because their results are determined by the interpretation and the memory of the subject (12). Instead, the objective methods provide information of subject’s PA through the quantification of movement. Pedometers and accelerometers are the most frequently used objective instruments in free-living situations (13). Pedometers are characterized by simplicity, comfort and low cost but their validity estimating PA or SB level is limited as they only register the number of steps by time. However some of their limitations could be solved by accelerometers, which register intensity, duration and frequency of PA (14).

Accelerometers also have limitations since they do not identify the movement of all body parts, static exercises, movement on grading terrain or carrying a load (15). However they allow to estimate EE as well as patterns and levels of PA and SB during free-living situations with a high intra- and inter-instrument reliability (16), even slightly higher than the majority of questionnaires (test-retest reliability intraclass correlation coefficients: 0.77–0.90 for the ActiGraph 7164 vs. 0.59–0.84 for the questionnaires) (17). Moreover, accelerometers have a moderate validity compared to doubly labelled water estimating EE [0.39 energy expenditure due to PA (AEE) and 0.52 total energy expenditure (TEE) for uni-axial devices and 0.59 AEE and 0.61 TEE for tri-axial devices] (18,19). On average, uni-axial accelerometers underestimated AEE by 24% and TEE by 12%, while tri-axial devices underestimated AEE by 21% and TEE by 7% (19). Average absolute values of the differences between questionnaires and doubly labelled water estimates of AEE and TEE were 32% and 23% respectively (20).

Wearable technology has been identified as the first fitness trend in 2016, so the importance of these devices are increasing in the recent years (21). One example that hit the market was the Polar V800, capable of recording accelerometry, heart rate, speed and geolocation (GPS). This device belongs to the renowned Finnish brand Polar Electro, which is one of the leaders in the sector since 1983 launched to the world the first wireless heart rate monitor (22). Although this monitor is used by thousands of researchers, health professionals and athletes, there is no scientific study nowadays evaluating its validity to estimate EE, PA or SB patterns in free-living situations. So, it is necessary do it (4,23,24). For these reasons the aim of the present study was to compare the validity of Polar V800 against a validated ActiGraph accelerometer (ActiTrainer) quantifying EE, PA and SB in young adults.

Methods

Subjects

Twenty-two Caucasian healthy subjects (50% women) ranged between 19 and 23 years old volunteered to participate in the study. The protocol was conducted according to ethical standards derived from the Spanish Medical Ethics Code and the Declaration of Helsinki for research in Human 1974 (last modified in 2013) (25). The study was approved by the University’s Human Ethics Committee (University of Zaragoza). Prior to conducting, the study a written informed consent was obtained from each subject.

Experimental design

The subjects were given written and verbal instructions and required to wear both devices simultaneously from the moment they woke up in the morning until bedtime at night, so that a full day could be registered. In addition, both devices had to be removed for aquatic activities to prevent damage to the ActiTrainer because it is not waterproof. Subjects were asked to complete a provided diary and note the time of putting on the devices, of removing them at the end of the day, and any time when the devices were removed and reattached during the day.

Measurement

Anthropometric measurements

Body mass (BM) was measured in light clothing using an electronic balance (to 0.1 kg). Height was measured without shoes to the nearest 0.1 cm using a portable squad stadiometer for all subjects following the procedures defined by International Society for the Advancement in Kinanthropometry (ISAK). Body mass index (BMI) was calculated as weight (kg) divided by height squared (m²).

PA assessment

Subjects wore simultaneously the Polar V800 and the ActiTrainer for 7 consecutive days to observe PA and SB (26).

ActiTrainer: reference accelerometer

The ActiTrainer, an ActiGraph’s accelerometer, is surrounded by a metal shield and packed into a plastic enclosure measuring 8.6 cm × 3.3 cm × 1.5 cm, weighs around 53 gr and includes a 3 V coin lithium battery. It has a dynamic range of 0.25–2.5 G and a sampling frequency of 30 Hz, measuring acceleration in the Y axis (omnidirectional) (27). The device was mounted on the right hip with an adjusted elastic belt to ensure close contact with the body. Monitors were set to record PA in a 15 s epoch and the “Step count” and “Dual axis” modes were selected; the firmware version used was the 7.1.0.

The results obtained were processed with Actilife version 6.5.4 software (Actigraph, LLC, Pensacola, FL, USA). Registered counts/15 s were reintegrated to 60 s epoch. The classification of wear and non-wear intervals was done by the Choi algorithm (Zero-count threshold during a non-wear time interval; 90 min time window for consecutive zero/nonzero counts; and allowance of 2 min interval of nonzero counts with the up/downstream 30 min consecutive zero counts window for detection of artifactual movements) (28). Also, 10 hours were established as the minimum necessary to be considered a valid day (29) and a minimum of five valid days, one of them on the weekend (30,31), to include the subject in the analysis (26,30). The previous published cut-off points proposed by Freedson (32) were used to estimate PA levels respect to PA intensity: sedentary [0–99 counts per minute (CPM)], light (100–759 CPM), lifestyle (760–1,951 CPM), moderate (1,952–5,724 CPM), vigorous (5,725–9,498 CPM) and very vigorous (>9,499 CPM) (33). Freedson’s cut-off points were chosen because the sample used to validate it were similar (25 males, 24.8±4.2 yrs; and 25 females, age 22.9±3.8 yrs). Finally, <100 CPM was defined as SB (34) and the sedentary bouts were calculated for 10 minutes as for 1 hour.

Predicted EE in kcal·min⁻¹ of the ActiTrainer monitor were calculated using the combined equation specified in the Actilife’s software manual (35). Work-energy theorem was used when activity didn’t exceed 1,952 CPM threshold [EE (kcal·min⁻¹) = 0.0000191 × CPM × BM (kg)]; and the Freedson equation was used when activity counts exceed 1,952 CPM {EE (kcal·min⁻¹) = 0.00094 × CPM + [0.1346 × BM (kg) − 7.37418]}.

Polar V800

The Polar V800 measures 37 mm × 56 mm × 12.7 mm and weighs 79 gr, is operated by a 350 mAh Li-pol rechargeable battery. It registers activity by an internal tri-axial accelerometer that records wrist movements, giving several outcomes as EE (Kcal), sedentary alerts, PA (lying time, sitting time and active time -standing, walking and running-) and number of steps. The wrist placement was defined by the subject’s preference.

Epoch length, definition of non-wearing time, cut-off points and the formula used for the estimation of EE are pre-fixed by the manufacturer of the Polar V800 and are not indicated in the manual (36).

Statistical analysis

All statistical analyses were performed using IBM SPSS (v.20 for Windows, Chicago, IL, USA) and the significance level was set at P≤0.05.

Paired t-tests were used to analyze significant differences in variables between devices. Pearson’s correlation coefficients (r_p) were calculated to examine the correlation between outcomes, defined 0.9–1 as being very high correlation, 0.7–0.9 as high correlation, 0.5–0.7 as moderate correlation, 0.3–0.5 as low correlation and 0–0.3 as negligible correlation (37).

The agreement between monitors was examined using the analysis proposed by Bland-Altman; BIAS, standard deviation (SD) of BIAS and limits of agreement at 95% (LOA) were calculated to evaluate agreement among variables obtained by both devices (38). Also, the association between difference and magnitude of the measurement (e.g., heteroscedasticity) was examined, inserting the difference between the value of the measurement for the Polar V800 and the value for the ActiTrainer as dependent variable, and the average value [(value for the Polar V800 + value for ActiTrainer)/2] as independent variable.

Finally, sensitivity, specificity and area under the receiver operating characteristic curve (ROC-AUC value) were calculated to evaluate the ability of the devices to accurately define a person who fulfills the recommendation of 10,000 daily steps (39). Perfect classification is represented by an ROC-AUC of 1, whereas a complete absence of classification is represented by an area of 0.5 ROC-AUC values, of ≥0.90 are considered to be excellent, values between 0.80–0.90 are considered to be good, values between 0.70–0.80 are fair and values <0.70 are considered to be poor (40).

Results

The descriptive characteristics of the subjects are presented in Table 1. Eighteen of the 22 subjects (50% women) ranged between 19 and 23 years old (mean 21.00±1.24, see Table 1) met inclusion criteria for analysis (5 valid days, 1 of them on the weekend). There were no significant differences in age, gender, height, weight and BMI between the 18 subjects included in the study and the 4 with incomplete data that were excluded. Subjects wore the ActiTrainer and the Polar V800 for an average of 874.65±10.76 min·day⁻¹ and range of 600–1,222 min·day⁻¹. Variables obtained by both devices are shown in Table 2.

Table 1 Descriptive characteristics of subjects
Full table

Table 2 Variables compared between devices
Full table

Paired t-test results are showed in Figure 1. Polar V800’s alerts vs. ActiTrainer’s 1 hour sedentary bouts (P=0.595) and Polar V800’s walking time vs. ActiTrainer’s lifestyle time (P=0.484) weren’t statistically different, although the other variables differed significantly (P<0.05) between methods. Polar V800’s alerts vs. ActiTrainer’s 1 hour sedentary bouts (P=0.456) and Polar V800’s walking time vs. ActiTrainer’s lifestyle time (P=0.920) were also the variables that did not differ significantly when they were compared on weekdays (Figure 2). However, when weekend days were compared, Polar V800’s alerts vs. ActiTrainer’s 1 hour sedentary bouts (P=0.875), Polar V800’s walking time vs. ActiTrainer’s lifestyle time (P=0.296), Polar V800’s walking time vs. ActiTrainer’s moderate time (P=0.839), non-vigorous active time (P=0.219) and Vigorous active time (P=0.077) weren’t significantly different (Figure 3).

Figure 1 Polar V800 and ActiTrainer outcomes recorded during all week (mean ± SD). Values are means ± standard deviation (SD). The leftmost P is for the paired-samples t-test (*, P<0.05). r_p, Pearson’s correlation coefficient (**, P<0.05). TEE, total energy expenditure; RMR, resting metabolic rate for the Polar V800.

Figure 2 Polar V800 and ActiTrainer outcomes recorded during the week days (mean ± SD). Values are means ± standard deviation (SD). The leftmost P is for the paired-samples t-test (*, P<0.05). r_p, Pearson’s correlation coefficient (**, P<0.05). TEE, total energy expenditure; RMR, resting metabolic rate for the Polar V800.

Figure 3 Polar V800 and ActiTrainer outcomes recorded during the weekend days (mean ± SD). Values are means ± standard deviation (SD). The leftmost P is for the paired-samples t-test (*, P<0.05). r_p, Pearson’s correlation coefficient (**, P<0.05). TEE, total energy expenditure; RMR, resting metabolic rate for the Polar V800.

The r_p of the whole week (Figure 1) shows that there was a very high correlation (0.9–1) between steps measured by both accelerometers; there was a high correlation (0.7–0.9) for active time, Polar V800’s walking time vs. ActiTrainer’s moderate time and non-vigorous active time; and a moderate correlation (0.5–0.7) for Polar V800’s RMR Kcal vs. ActiTrainer’s Kcal, Sitting time and Polar V800’s walking time vs. ActiTrainer’s lifestyle time. Results separately for weekday and weekend days show similar trends (Figures 2,3).

Bland-Altman results (BIAS, SD of BIAS and LOA) are shown in Figures 4-6 and Tables 3-5. Heteroscedasticity analysis showed a significantly positive correlation between the difference and the mean value for all outcomes, except for Polar V800’s Kcal vs. ActiTrainer’s Kcal (P=0.868) and sitting time (P=0.616) (Table 3). As shown in Tables 4 and 5 results were similar when only weekdays or weekend days were compared, with the only exception that at the weekend no significant correlation was also found in the non-vigorous active time (P=0.059). So except for the variables just mentioned, evidence of “heteroscedasticity” between the devices was found, in other words, the width of the random scatter increased as the general size of measured.

Figure 4 Bland-Altman results during all week.

Figure 5 Bland-Altman results during the week days.

Figure 6 Bland-Altman results during the weekend days.

Table 3 Bland-Altman results during all week
Full table

Table 4 Bland-Altman results during the week days
Full table

Table 5 Bland-Altman results during the weekend days
Full table

Regarding ActiTrainer, Polar V800’s ability to accurately define a person who fulfills the recommendation of 10,000 daily steps was fair (ROC-AUC: 0.781±0.048; P<0.001; LOA =0.687–0.874) and the sensitivity and specificity was 98% and 58%, respectively.

Conclusions

The results of this study show that 1 hour sedentary periods estimated by the Polar V800 and the ActiGraph ActiTrainer were similar. Also, the time defined as walking by the Polar V800 was similar to the time dedicated in activities falling just below 3METs (defined by Freedson as Lifestyle intensities) in young adults (33). However, the Polar V800 did not estimate accurately the rest of the variables [e.g., TEE (Kcal), sitting time, active time, Polar V800’s walking time vs. ActiTrainer’s moderate time, non-vigorous active time, vigorous active time and steps].

The outcomes provided by an accelerometer depend on several variables as brand, generation, wearing position, epoch length, definition of non-wearing time, definition of a valid day, required number of days of monitoring, cut-off points, use and definition of activity bouts, or accelerometer firmware version (20). Some of the aforementioned variables (e.g., epoch length, definition of non-wearing time, cut-off points or the formula used for the estimation of EE) are pre-fixed by manufacturers of the Polar V800; moreover these settings are unknown by the user/researcher. This fact increases the complexity of the Polar V800’s outcome comparisons.

EE estimated by the Polar V800 was significantly higher than estimated by the ActiTrainer (P<0.001), the origin of this difference could be multifactorial. Firstly, the average wear time between both accelerometers was significantly different (P<0.001), the ActiTrainer was worn for 218.89 minutes/day less than the Polar V800. This difference in wear time may be due to the fact that the Polar V800 is more comfortable (41). In addition, some subjects did not follow the instructions and wore the Polar V800 during sleep and water activities; moreover the Polar V800 overestimated SB because it classifies some non-used time as lying because it does not have galvanic sensors which can detect if it is worn. Secondly, literature shows that tri-axial accelerometers underestimate TEE 5% less compared to the uni-axial accelerometers (18,19), which is in line with our results as the Polar V800 is tri-axial and the ActiTrainer is uni-axial (omnidirectional). Lastly, for the subjects who recorded training sessions, the Polar V800 used GPS and heart rate data (if the band is used) to estimate EE, thereby it could be more accurate. These reasons could explain why the correlation of this variable (r_p =0.483) was found to be lower than the one found by a recent study between other daily use monitors (Fitbit Flex, Jawbone UP 24, Misfit Shine, Nike + Fuelband SE and Polar Loop) (r_p =0.71–0.9) or research devices (Actigraph GT3X+ and BodyMedia Core) compared with indirect calorimetry (42). Even though the correlation was low for this variable, homoscedasticity was observed (P=0.868); so the difference between methods remains constant, which is probably explained because each device estimates the EE from its own formula.

In regards to sedentary alerts, no significant difference was found between the Polar V800’s alerts and ActiTrainer’s 1 hour sedentary bouts (P=0.595), so both devices could be valid for measuring these important indicators of SB. The Polar V800 can be an useful tool to modify these long periods sitting, such negative for health, since vibration and a screen message alerts the user when he/she spends too much time on SB (9,43). Such real-time feedback is suggested to be important to improve weight loss in obese and overweight adults (44).

Concerning the time spent in each zone of PA intensity, significant differences were observed for all variables except for Polar V800’s walking time vs. ActiTrainer’s lifestyle time (P=0.484). However at a theoretical level, even this seems coincidental, since: (I) both devices use different cutpoints, e.g., walking activity according to Freedson’s cutpoints could be “Moderate” (1,952–5,724 CPM) or even “Vigorous” (5,725–9,498 CPM) (33,45,46); (II) as it has been previously discussed, wearing time of the two devices was not similar; (III) the correlation shown in non-vigorous active time (r_p =0.847) is similar to the correlation found in other studies between daily use devices (Fitbit One, Fitbit Zip, Jawbone UP, Misfit Shine, Nike Fuelband, Striiv Smart and Withings Pulse) and research accelerometers (BodyMedia SenseWear and ActiGraph GT3X+) (r_p =0.52–0.91) (47).

Steps registered by each monitor were significantly different (BIAS: 2,487±2,293 steps/day; P<0.001). The ActiTrainer has been validated for steps recording in a similar study population (48), so it may be that the Polar V800 is not valid for registering steps, but it cannot be forgotten that wearing time of the devices was not the same. Besides, although it showed heteroscedasticity (P<0.001), the correlation between the steps of both methods was very high (r_p =0.904), which is similar to those between other daily devices (Fitbit One, Fitbit Zip, Jawbone UP, Misfit Shine, Nike Fuelband, Striiv Smart and Withings Pulse) and research accelerometers (BodyMedia SenseWear,and ActiGraph GT3X+) (r_p >0.8) (47).

Finally, the ability of the Polar V800 to define if a person meet the recommendation of 10,000 daily steps compared with the ActiTrainer has been studied. The 98% of the cases who met the recommendations, was classified as they fulfilled them (sensitivity), 58% of the subjects not meeting the recommendations was classified as they did not comply (specificity), and the ROC-AUC value was fair (0.781±0.048). Results presented can probably be explained by the wear time difference, because Polar V800 has been measuring more time, so more people can fulfill the recommendations according to this device.

In the last years it has been repeatedly pointed out how the use of new technologies can help to improve health and prevent future diseases (49). Some studies even suggest that the new wearables (e.g., Fitbit) are more effective to increase PA than those used a few years ago, such as pedometers (50). However, as said by some authors, most companies that manufacture wearable devices do not provide information nor guarantee the reliability, effectiveness or accuracy of their services and devices (51).

The main limitation of the present study was that the wear time of both devices was not the same, a possible solution would be to fragment both recordings in bouts (27), but Polar V800 only gives overall results of all day, so the best solution would be to compare with an waterproof accelerometer. Secondly, as the intensity of activities has been calculated with different cutpoints, the number of intensity levels obtained is different (6 vs. 5) (33). Moreover, even though the comparison has been made regarding an accelerometer that has been widely used in research and has shown its reliability, validity and comparability (29,34,52), in the future, validation of Polar V800’s PA and EE outcomes should be evaluated against gold standard (such as direct calorimetry and/or double labelled water). Among the strengths it is found that the study was conducted during free-living so there is a greater ecological external validity and therefore the results are more generalizable than in laboratory studies; besides, although having an inclusion criteria (5 valid days, one of them in the weekend) more demanding than the usual (4 valid days), the number of lost subjects is similar to other studies (18% vs. 17%); in addition there were no significant differences in the characteristics of valid subjects and those who don’t met inclusion criteria. Likewise, among subjects who met the inclusion criteria, the percentage that did not have valid data for all days of measurement (55%), and the average wearing time (14.6 h) are comparable to those observed in literature: (27–74%) and (13–15 h) respectively (20).

This study shows that the tri-axial accelerometer of Polar V800 has a comparable validity to the use of the ActiTrainer’s accelerometer as a standard reference assessing “1 hour sedentary bouts” and “lifestyle time” in young adults in free-living.

Acknowledgements

We would like to thank the collaboration of all subjects who participated voluntarily in the study.

Footnote

Conflicts of Interest: The authors have no conflicts of interest to declare.

Ethical Statement: The study was approved by the University’s Human Ethics Committee (University of Zaragoza) and written informed consent was obtained from all patients.

References

1. Morris JN, Heady JA, Raffle PA, et al. Coronary heart-disease and physical activity of work. Lancet 1953;265:1053-7. contd. [Crossref] [PubMed]
2. Physical Activity Guidelines Advisory Committee report, 2008. To the Secretary of Health and Human Services. Part A: executive summary. Nutr Rev 2009;67:114-20. [PubMed]
3. Wegner M, Helmich I, Machado S. Effects of exercise on anxiety and depression disorders: review of meta- analyses and neurobiological mechanisms. CNS Neurol Disord Drug Targets 2014;13:1002-14. [Crossref] [PubMed]
4. Garber CE, Blissmer B, Deschenes MR, et al. American College of Sports Medicine position stand. Quantity and quality of exercise for developing and maintaining cardiorespiratory, musculoskeletal, and neuromotor fitness in apparently healthy adults: guidance for prescribing exercise. Med Sci Sports Exerc 2011;43:1334-59. [Crossref] [PubMed]
5. American College of Sports M, Chodzko-Zajko WJ, Proctor DN, et al. American College of Sports Medicine position stand. Exercise and physical activity for older adults. Med Sci Sports Exerc 2009;41:1510-30. [Crossref] [PubMed]
6. Haskell WL, Lee IM, Pate RR, et al. Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Med Sci Sports Exerc 2007;39:1423-34. [Crossref] [PubMed]
7. Moore SC, Patel AV, Matthews CE, et al. Leisure time physical activity of moderate to vigorous intensity and mortality: a large pooled cohort analysis. PLoS Med 2012;9:e1001335. [Crossref] [PubMed]
8. Owen N, Healy GN, Matthews CE, et al. Too much sitting: the population health science of sedentary behavior. Exerc Sport Sci Rev 2010;38:105-13. [Crossref] [PubMed]
9. Brocklebank LA, Falconer CL, Page AS, et al. Accelerometer-measured sedentary time and cardiometabolic biomarkers: A systematic review. Prev Med 2015;76:92-102. [Crossref] [PubMed]
10. Biswas A, Oh PI, Faulkner GE, et al. Sedentary time and its association with risk for disease incidence, mortality, and hospitalization in adults: a systematic review and meta-analysis. Ann Int Med 2015;162:123-32. [Crossref] [PubMed]
11. Ainsworth B, Cahalin L, Buman M, et al. The current state of physical activity assessment tools. Prog Cardiovasc Dis 2015;57:387-95. [Crossref] [PubMed]
12. Hills AP, Mokhtar N, Byrne NM. Assessment of physical activity and energy expenditure: an overview of objective measures. Front Nutr 2014;1:5. [Crossref] [PubMed]
13. Strath SJ, Kaminsky LA, Ainsworth BE, et al. Guide to the assessment of physical activity: Clinical and research applications: a scientific statement from the American Heart Association. Circulation 2013;128:2259-79. [Crossref] [PubMed]
14. Aparicio-Ugarriza R, Mielgo-Ayuso J, Benito PJ, et al. Physical activity assessment in the general population; instrumental methods and new technologies. Nutr Hosp 2015;31 Suppl 3:219-26. [PubMed]
15. Vanhees L, Lefevre J, Philippaerts R, et al. How to assess physical activity? How to assess physical fitness? Eur J Cardiovasc Prev Rehabil 2005;12:102-14. [Crossref] [PubMed]
16. Sirard JR, Forsyth A, Oakes JM, et al. Accelerometer test-retest reliability by data processing algorithms: results from the Twin Cities Walking Study. J Phys Act Health 2011;8:668-74. [PubMed]
17. Helmerhorst HJ, Brage S, Warren J, et al. A systematic review of reliability and objective criterion-related validity of physical activity questionnaires. Int J Behav Nutr Phys Act 2012;9:103. [Crossref] [PubMed]
18. Plasqui G, Bonomi AG, Westerterp KR. Daily physical activity assessment with accelerometers: new insights and validation studies. Obes Rev 2013;14:451-62. [Crossref] [PubMed]
19. Van Remoortel H, Giavedoni S, Raste Y, et al. Validity of activity monitors in health and chronic disease: a systematic review. Int J Behav Nutr Phys Act 2012;9:84. [Crossref] [PubMed]
20. Pedišić Ž, Bauman A. Accelerometer-based measures in physical activity surveillance: current practices and issues. Br J Sports Med 2015;49:219-23. [Crossref] [PubMed]
21. Thompson WR. Worldwide survey of fitness trends for 2016: 10th Anniversary Edition. ACSM's Health & Fitness Journal 2015;19:9-18.
22. Laukkanen RM, Virtanen PK. Heart rate monitors: state of the art. J Sports Sci 1998;16 Suppl:S3-7. [Crossref] [PubMed]
23. Blair SN. Physical inactivity: the biggest public health problem of the 21st century. Br J Sports Med 2009;43:1-2. [PubMed]
24. Lee IM, Shiroma EJ, Lobelo F, et al. Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. Lancet 2012;380:219-29. [Crossref] [PubMed]
25. World Medical Association. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA 2013;310:2191-4. [Crossref] [PubMed]
26. Trost SG, McIver KL, Pate RR. Conducting accelerometer-based activity assessments in field-based research. Med Sci Sports Exerc 2005;37:S531-43. [Crossref] [PubMed]
27. Ojiambo R, Konstabel K, Veidebaum T, et al. Validity of hip-mounted uniaxial accelerometry with heart-rate monitoring vs. triaxial accelerometry in the assessment of free-living energy expenditure in young children: the IDEFICS Validation Study. J Appl Physiol (1985) 2012;113:1530-6. [PubMed]
28. Choi L, Liu Z, Matthews CE, et al. Validation of accelerometer wear and nonwear time classification algorithm. Med Sci Sports Exerc 2011;43:357-64. [Crossref] [PubMed]
29. Cain KL, Geremia CM. Accelerometer data collection and scoring manual. Accessed Dec. 2011.
30. Mâsse LC, Fuemmeler BF, Anderson CB, et al. Accelerometer data reduction: a comparison of four reduction algorithms on select outcome variables. Med Sci Sports Exerc 2005;37:S544-54. [Crossref] [PubMed]
31. Esliger DW, Copeland JL, Barnes JD, et al. Standardizing and optimizing the use of accelerometer data for free-living physical activity monitoring. J Phys Act Health 2005;3:366-83.
32. Matthews CE, Chen KY, Freedson PS, et al. Amount of time spent in sedentary behaviors in the United States, 2003-2004. Am J Epidemiol 2008;167:875-81. [Crossref] [PubMed]
33. Freedson PS, Melanson E, Sirard J. Calibration of the Computer Science and Applications, Inc. accelerometer. Med Sci Sports Exerc 1998;30:777-81. [Crossref] [PubMed]
34. Kerr J, Sallis JF, Owen N, et al. Advancing science and policy through a coordinated international study of physical activity and built environments: IPEN adult methods. J Phys Act Health 2013;10:581-601. [PubMed]
35. Engineering/Marketing A. ActiLife Users Manual. Pensacola, FL: ActiGraph, 2012.
36. Polar Electro [Internet]. June 2015. Polar [cited 2015 August 2]. Available online: http://www.polar.com/e_manuals/V800/Polar_V800_user_manual_English/manual.pdf
37. Mukaka MM. Statistics corner: A guide to appropriate use of correlation coefficient in medical research. Malawi Medical Journal 2012;24:69-71. [PubMed]
38. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-10. [Crossref] [PubMed]
39. Tudor-Locke C, Craig CL, Brown WJ, et al. How many steps/day are enough? For adults. Int J Behav Nutr Phys Act 2011;8:79. [Crossref] [PubMed]
40. Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem 1993;39:561-77. [PubMed]
41. Freedson PS, John D. Comment on "estimating activity and sedentary behavior from an accelerometer on the hip and wrist". Med Sci Sports Exerc 2013;45:962-3. [Crossref] [PubMed]
42. Bai Y, Welk GJ, Nam YH, et al. Comparison of Consumer and Research Monitors under Semistructured Settings. Med Sci Sports Exerc 2016;48:151-8. [Crossref] [PubMed]
43. Dempsey PC, Owen N, Biddle SJ, et al. Managing sedentary behavior to reduce the risk of diabetes and cardiovascular disease. Curr Diab Rep 2014;14:522. [Crossref] [PubMed]
44. Shuger SL, Barry VW, Sui X, et al. Electronic feedback in a diet- and physical activity-based lifestyle intervention for weight loss: a randomized controlled trial. Int J Behav Nutr Phys Act 2011;8:41. [Crossref] [PubMed]
45. Kozey SL, Lyden K, Howe CA, et al. Accelerometer output and MET values of common physical activities. Med Sci Sports Exerc 2010;42:1776-84. [Crossref] [PubMed]
46. Ainsworth BE, Haskell WL, Herrmann SD, et al. 2011 Compendium of Physical Activities: a second update of codes and MET values. Med Sci Sports Exerc 2011;43:1575-81. [Crossref] [PubMed]
47. Ferguson T, Rowlands AV, Olds T, et al. The validity of consumer-level, activity monitors in healthy adults worn in free-living conditions: a cross-sectional study. Int J Behav Nutr Phys Act 2015;12:42. [Crossref] [PubMed]
48. Neuls F. Validity and reliability of" step count" function of the ActiTrainer activity monitor under controlled conditions. Acta Gymnica 2009;38:55-64.
49. Franklin NC, Lavie CJ, Arena RA. Personal health technology: A new era in cardiovascular disease prevention. Postgrad Med 2015;127:150-8. [Crossref] [PubMed]
50. Cadmus-Bertram LA, Marcus BH, Patterson RE, et al. Randomized Trial of a Fitbit-Based Physical Activity Intervention for Women. Am J Prev Med 2015;49:414-8. [Crossref] [PubMed]
51. Lee J, Finkelstein J. Activity trackers: a critical review. Stud Health Technol Inform 2014;205:558-62. [PubMed]
52. Lee KY, Macfarlane DJ, Cerin E. Comparison of three models of actigraph accelerometers during free living and controlled laboratory conditions. Eur J Sport Sci 2013;13:332-9. [Crossref] [PubMed]