Share this article on:

Patterns of Accelerometer-Derived Estimates of Inactivity in Middle-Age Women


Medicine & Science in Sports & Exercise: January 2012 - Volume 44 - Issue 1 - p 104–110
doi: 10.1249/MSS.0b013e318229056e

Purpose The study’s purpose was to characterize accelerometer-derived estimates of physical inactivity collected during five consecutive weeks in middle-age women.

Methods Data were obtained from 63 participants (95.5%) enrolled in the Evaluation of Physical Activity Measures in Middle-Age Women Study. Inactive time (min·d−1) was estimated as the sum of activity counts <100, and inactive-to-active transitions were defined as an interruption in which a period of inactivity was immediately followed by a minute or more above 100 counts. A repeated-measures ANOVA using PROC MIXED (SAS/STAT software, v. 9.2) was used to describe hourly, daily, and weekly variation in estimates of physical inactivity.

Results Participants were 52.7 ± 5.5 yr, 85.7% non-Hispanic white, and 63.5% postmenopausal, with a body mass index of 26.7 ± 5.1 kg·m−2. Inactive time gradually increased as the day continued, particularly on weekend days. When compared with weekdays, average inactive time was lower on Saturday and Sunday (all P < 0.01 except for Saturday vs Monday, P < 0.10); Saturdays were not significantly different from Sundays. Breaks in inactive time were significantly lower on Sunday when compared with weekdays and Saturday (all P < 0.05), and fewer breaks were noted on Saturday when compared with Wednesday and Friday (both P < 0.01). After adjustment for total wear time or inactive time, most day-to-day differences were attenuated. Week-by-week differences in physical inactivity estimates were also not statistically significant.

Conclusions The results of this study suggest that inactive time increases as the day continues and that daily physical inactivity estimates are more stable after 1) adjustment for wear time or 2) when averaged over the week. Researchers should carefully consider the intended application of physical inactivity estimates before data collection and processing, analysis, and final data reporting.

1School of Public Health–Austin Regional Campus, The University of Texas Health Science at Houston, Houston, TX; 2Risk Factor Monitoring and Methods Branch, Applied Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, MD; 3Department of Biostatistics, University of Nebraska Medical Center, Omaha, NE; and 4Healthy Lifestyles Research Center, Program in Exercise and Wellness, Arizona State University, Mesa, AZ

Address for correspondence: Kelley Pettee Gabriel, Ph.D., Division of Epidemiology,Human Genetics, and Environmental Sciences, School of Public Health–Austin Regional Campus, The University of Texas HealthScience Center at Houston, The University of Texas AdministrationBuilding,1616 Guadalupe Street, Suite 6.300, Austin, TX 78701; E-mail:

Submitted for publication May 2011.

Accepted for publication June 2011.

Recent research has sought to clarify the independent role of prolonged inactive or sedentary time on obesity (4,9) and cardiometabolic risk factors (2,7), as well as chronic disease outcomes including type 2 diabetes mellitus (8,9) and cardiovascular disease mortality (24). As a result, interest in the measurement of physical inactivity and subsequent relationships with health outcomes is quickly gaining momentum as an important focus in public health research (17). However, further work in this area is needed to achieve the knowledge base that physical activity–related research has acquired during the past 50-plus years.

Given the Latin root of the word sedentary (i.e., sedēre, to sit), sedentary behaviors have been defined as seated activities that occur across several domains including work, nonactive commuting, and certain leisure time activities (e.g., television watching or computer/Internet use). In turn, this has become a generally accepted definition and is widely used in population-based research (17). Unfortunately, waist-mounted activity monitors are not able to effectively predict the wearer’s posture. Despite this, activity monitors have been used in several previous investigations (12,17) with a cut point threshold (<100 counts per minute) used to characterize time spent in sedentary pursuits. Given this inability of waist-worn devices to detect posture, other nonambulatory activities, such as time spent lying, reclining, and standing, may also be included in the summary estimate. With this in mind, we refer to derived estimates as physical inactivity or inactive time, rather than sedentary or sitting time, to reduce the potential for misclassification bias and aid in the interpretation of results.

In an adult population, much of what we know about physical inactivity originates from national surveillance data systems, where questions primarily focus on active parameters to describe physical activity levels within nationally representative samples. In 2003–2006, the National Health and Nutrition Examination Survey used accelerometers to provide a direct measure of human movement in US children and adults age ≥6 yr. This, in turn, has fostered novel research that explores the associations between movement or lack thereof with health risk factors and related chronic disease outcomes (14). Methods for quantifying physical activity volume (frequency, intensity, and duration) using accelerometer-based motion sensors have evolved during the past 20 yr (13,20,23). In particular, the variability of accelerometer-derived physical activity estimates during repeated sampling days or periods has been an important methodological focus (11,21,25). Common sources of variance that are known to influence the quantification of physical activity levels include intraindividual differences, day-of-week effects, and seasonal influences (11,26). However, little is known about the variability of these measures acutely (i.e., within a day) or during extended periods of time. This is important because it will provide a better understanding of the methodological requirements for measurement when examining relationships with health outcomes, which, in turn, will provide an enhanced appreciation for how to best target and reduce inactive time to realize health benefit. Further, examining patterns of physical inactivity in specific population subgroups may better inform health promotion research. Therefore, the purpose of this study was to characterize hourly, daily, and weekly patterns of accelerometer-derived estimates of physical inactivity in middle-age women representing pre- to post-menopause during five consecutive weeks.

Back to Top | Article Outline


Study Design and Participants

This descriptive study of accelerometer-derived estimates of physical inactivity was conducted within the larger Evaluation of Physical Activity Measures in Middle-Age Women (PAW) Study. The PAW Study was designed to evaluate the psychometric properties of six physical activity measures (i.e., five questionnaires and a walking-based performance measure) used in epidemiological studies of physical activity and women’s health. The protocol used in the PAW Study has been previously reported (18,19). Briefly, participants completed six consecutive weekly visits from August 2007 to May 2008, each lasting 30–60 min. Seventy-seven women were screened, and 66 (85.7%) were enrolled into the study. Among those who were not enrolled (n = 11), reasons included lack of time (n = 9), family obligations (n = 1), and a preexisting health condition (n = 1) that precluded participation in the study. All participants provided written informed consent, and the study protocol was approved by the institutional review board at Arizona State University.

Back to Top | Article Outline

ActiGraph GT1M Accelerometer

Objective data on physical activity and inactivity were collected using the ActiGraph GT1M accelerometer (Pensacola, FL). The ActiGraph is a small (3.8 × 3.7 × 1.8 cm) uniaxial piezoelectric accelerometer that is typically worn at the waist, which measures acceleration in the vertical plane. Data outputs from the ActiGraph accelerometer are activity counts, which quantify the amplitude and frequency of detected accelerations. Activity counts are summed over a researcher-specified time interval (i.e., epoch). In the current study, a 1-min epoch is reported. The sum of the activity counts in a given epoch is related to activity intensity and can be categorized on the basis of validated activity count cut points (10). Technical specifications, as well as the reliability and validity of the ActiGraph, have been described previously (10,16).

Participants wore the ActiGraph (dominant hip) everyday, during all waking hours, throughout the 6-wk study. Participants were asked to record the time at which they put on the monitors in the morning and the time they took off the monitors at night in a physical activity diary provided by study staff. At the end of each week, the participant returned the physical activity diary to study staff and was given another diary to complete during the following week. Each week, data from the accelerometer were downloaded and screened for wear time using methods reported by Troiano et al. (20). Briefly, device nonwear time was defined as 60 consecutive minutes of 0 counts, with an allowance for 1–2 min of detected counts between 0 and 100. Wear time was determined by subtracting derived nonwear time from 24 h (15,20). A minimum of 10 h of wear time was required for data to be considered for further use in calculating hourly and daily estimates of physical inactivity. For weekly estimates, a minimum of 10 h of wear time per day for at least 4 of 7 d was necessary for each week of data included in analyses.

Back to Top | Article Outline

Accelerometer-derived estimates of physical inactivity

Inactive time (min·d−1) was estimated as the amount of time accumulated below 100 counts per minute during detected periods of monitor wear (12). Inactive-to-active transitions (i.e., breaks in inactive time) were defined as an interruption in which a period of physical inactivity was immediately followed by a minute or more above 100 counts; breaks were summed over each day (7). For hourly estimates, inactive time and breaks in inactive time were averaged over the entire hour (e.g., the hour specified as 9:00 a.m. represents 9:00–9:59 a.m.). For daily and weekly estimates, inactive time (min·d−1) and the number of breaks in inactive time (number of breaks per day) were adjusted for total wear time and the amount of inactive time accumulated per day, respectively. Weekly summary estimates of accelerometer-determined physical inactivity were compiled for all participants with at least four valid days of ≥10 h of wear time.

Back to Top | Article Outline

Participant Characteristics

Age (yr) and demographic characteristics including race/ethnicity, educational attainment, and health behavior information (i.e., smoking status) were collected using standardized questionnaires. Anthropometric measures, including body mass index (kg·m−2), were calculated from height (m) and weight (kg) measured with a stadiometer and calibrated balance beam scale, respectively.

Back to Top | Article Outline

Statistical Methods

Univariate analyses were conducted on measured parameters, and all variables were assessed for normality. Normally distributed variables were reported as mean and SD, and proportions were noted for categorical variables. Day-to-day and week-by-week physical inactivity estimates were reported as mean (SE). A repeated-measures ANOVA was used to describe hourly, daily, and weekly estimates of physical inactivity. Simulated confidence intervals on differences were used to examine day-to-day and week-by-week comparisons of the average values for all physical inactivity estimates (22). This is the recommended method for adjustments due to multiplicity for the unbalanced repeated-measures model (3).

Back to Top | Article Outline


Sixty-three PAW Study participants (95.5%) had five valid (each week, ≥10 h for ≥4 of 7 d) weeks of accelerometer data. The average daily accelerometer wear time during 5 wk was 868.7 ± 59.6 min or approximately 14.5 h·d−1. Further, average accelerometer wear time ranged from 837.3 (Sunday) to 910.1 (Wednesday) min·d−1. Wear time was significantly lower on Sunday when compared with all other days (all P < 0.01), and Saturday was significantly lower than Wednesday and Friday (both P < 0.01). The mean age and body mass index of study participants were 52.7 ± 5.5 yr and 26.7 ± 5.1 kg·m−2, respectively. Other participant characteristics include 85.7% non-Hispanic white, 60.3% married, 50.8% (4 yr) college graduates, 21.8% with family income <$50,000, 63.5% postmenopausal, 7.9% current smokers, and 33.3% reported having no chronic disease or condition.

Back to Top | Article Outline

Hourly Variation in Accelerometer-Derived Physical Inactivity Estimates

Inactive time (min·d−1) and inactive-to-active transitions (number of breaks per day)

The hour-by-hour mean (95% confidence intervals) inactive time (min·h−1) and inactive-to-active transitions (number of breaks per hour) are presented averaged across all weekdays and weekend days during the hours of 9:00 a.m. to 9:59 p.m. only, given that the full analytical sample was represented during this period (Figs. 1A, B). On weekdays, the mean (95% confidence interval) inactive time per hour (min·h−1) gradually increased from 9:00 a.m. (31.8 (30.2–33.3) min·h−1) to 1:00 p.m. (36.2 (34.8–37.5) min·h−1), was stable throughout the afternoon, dipped slightly from 5:00 to 7:00 p.m. (36.3 (35.0–38.2) and 34.9 (33.5–36.3) min·h−1, respectively), and then gradually increased until 9:00 p.m. (36.3 (34.7–37.8) min·h−1). On weekend days, average inactive time steadily increased from 9:00 a.m. to 9:00 p.m. (26.4 (24.8–27.9) to 37.5 (36.0–39.0) min·h−1) (Fig. 1A). Given that the ability to break a period of inactive time is contingent upon being inactive, the hour-by-hour patterns of inactive-to-active transitions (number of breaks per hour) for weekdays and weekend days are relatively consistent with those described above for hour-by-hour inactive time (min·h−1) (Fig. 1B).



Back to Top | Article Outline

Day-to-Day Variation in Accelerometer-Derived Physical Inactivity Estimates

Inactive time unadjusted (min·d−1) and adjusted for wear time (min·d−1)

When averaged over 5 wk of observation, the highest accumulated time spent being inactive occurred on Tuesday (mean ± SE = 569.8 ± 10.8 min·d−1), followed by Wednesday (568.1 ± 10.7 min·d−1), Friday (566.1 ± 10.1 min·d−1), and Thursday (564.3 ± 10.7 min·d−1). On average, participants had the lowest amount of physical inactivity on Monday (552.1 ± 10.5 min·d−1) and weekend days (525.8 ± 10.1 and 511.1 ± 10.2 min·d−1 on Saturday and Sunday, respectively). The day-to-day differences in inactive time adjusted for wear time showed a similar pattern, with Tuesday having the highest (64.0% ± 1.0%) and Saturday and Sunday having the lowest (both 61.0% ± 1.0%) proportion of the day spent in inactive pursuits (Table 1).



The difference in average daily accumulated inactive time (min·d−1) between days of the week shows that the time spent being inactive (min·d−1) on Sunday was significantly lower than on all weekdays (all P < 0.001); a similar pattern was shown when Saturday was compared with Tuesday (P < 0.001), Wednesday (P < 0.001), Thursday (P < 0.01), and Friday (P < 0.001) (Table 2). When comparing day-to-day mean inactive time (min·d−1) adjusted for total wear time (min·d−1), average values on both Saturday and Sunday were significantly lower than those on Tuesday (P < 0.05 and P < 0.01, respectively). No other comparisons in mean inactive time (min·d−1) adjusted for total wear time (min·d−1) were statistically significant (Table 2).



Back to Top | Article Outline

Inactive-to-active transitions (number of breaks per day) unadjusted and adjusted for inactive time (h·d−1)

During 5 wk of observation, the highest average number of inactive-to-active transitions occurred on Wednesday (mean ± SE = 99.4 ± 2.1 breaks per day), followed by Friday (98.2 ± 2.0 breaks per day), Tuesday and Thursday (both 97.2 ± 2.1 breaks per day), and Monday (96.1 ± 2.1 breaks per day). When compared with weekdays, fewer breaks in inactive time occurred on weekend days (92.7 ± 2.0 and 87.9 ± 2.0 breaks per day on Saturday and Sunday, respectively). After adjustment for inactive time (h·d−1), the number of inactive-to-active transitions was relatively consistent between days of the week (i.e., ranged from 10.7 ± 0.3 breaks per day per inactive hour on Tuesday to 11.0 ± 0.3 breaks per day per inactive hour on Saturday) (Table 1).

The difference in the daily average inactive-to-active transitions (number of breaks per day) between days of the week illustrates that the frequency of breaks was significantly lower on Sunday when compared with all weekdays (all P < 0.001) and Saturday (P < 0.05) (Table 2). Further, fewer breaks in inactive time were accumulated on Saturday when compared with Wednesday and Friday (both P < 0.01). No other day-to-day pairwise comparisons were statistically significant (Table 2). When inactive-to-active transitions (number of breaks per day) were adjusted for inactive time (h·d−1), there were no significant differences noted between days (Table 2).

Back to Top | Article Outline

Week-by-week variation in accelerometer-derived physical inactivity estimates

The amount of time (min·d−1) spent being inactive averaged over the number of valid days per week (i.e., ≥10 h·d−1 for ≥4 of 7 d) for each of the 5 wk the accelerometer was worn is shown unadjusted and after adjustment for wear time (min·d−1). Regardless of whether the estimate was unadjusted or adjusted, mean values in inactive time were not statistically different between any of the 5 wk. Similar to the estimates of daily inactive time, accumulated inactive-to-active transitions (number of breaks per day), averaged over the number of valid days per week, were not significantly different between weeks for either the unadjusted or adjusted values (Table 1).

Back to Top | Article Outline


In the current investigation, hourly, daily, and weekly patterns of commonly reported accelerometer-based estimates of physical inactivity were examined in middle-age women across five consecutive weeks. The results of the current study revealed several key findings. First, participants became increasingly more inactive as the day continued, particularly on weekend days. Although we did not collect contextual information (i.e., via a physical activity log or record) to complement accelerometer data, this description of hourly patterns has important public health implications for health promotion researchers interested in targeting reductions in inactive time for health benefit. More specifically, the finding that inactivity was prevalent across all hours of the day suggests that intervention strategies designed to increase opportunities to be more physically active across all domains (i.e., work and home) might reduce overall time spent being inactive. Second, average inactive time was significantly higher on weekdays when compared with weekend days, which is consistent with findings from a 2002 study by Matthews et al. (11). This finding further supports the implementation of work site interventions that are designed to increase physical activity levels throughout the workday. Third, when average daily inactive time and frequency of inactive-to-active transitions were adjusted for wear and inactive time, respectively, differences in estimates between days were attenuated. Finally, there were no significant differences for any estimate of physical inactivity when weekly averages from 1 wk were compared with other weeks (e.g., week 1 vs week 2, 3, 4, or 5).

In this study, weekly data suggest that participants spent approximately 62%–63% (i.e., ∼9.25 h of 14.5 total hours of wear) of the day being nonambulatory or inactive, which may include time spent reclining, sitting, and standing as assessed using the <100 cut point threshold. This proportion of daily time spent being inactive is consistent with results obtained from postmenopausal women enrolled in the Woman on the Move through Activity and Nutrition Study (5). Although 62%–63% represents a substantial portion of the day, it is important to note that derived estimates are contingent upon total daily activity monitor wear time. To illustrate this point, we will consider two examples in which participants A and B slept for 8 h each night. In our first example, participant A wore the activity monitor for 14.5 h, took it off, and then spent the remaining 1.5 h (i.e., 8-h sleep + 14.5-h monitoring period + 1.5 h = 24 h) doing light chores around the house to get ready for the next day. In this example, if participant A wore the monitor during the last 1.5 h of the day, the proportion of her day spent inactive would likely decrease. Conversely, participant B wore the monitor for 14.5 h, took it off, and then spent the remaining 1.5 h of the day sitting on the couch watching television. Here, had participant B worn the monitor, the proportion of the day spent inactive would increase. Therefore, it is important to consider the monitoring period when interpreting results from device-based studies, particularly when data collection was not conducted during a complete 24-h day.

The findings from this report have several other important implications for epidemiologic studies and health promotion programs that incorporate accelerometer-based methods to quantify physical inactivity. First, when describing inactive time, it is important for researchers or practitioners to consider whether it is most appropriate to report hourly, daily, or weekly averages that are either unadjusted or adjusted for total wear or sedentary time. This will likely differ depending upon the type of study and purpose(s) for assessment. For example, intervention studies or health promotion programs targeting specific reductions in time spent being inactive may focus on daily estimates that reflect total unadjusted inactive time or number of inactive-to-active transitions per day. From these measures, participants can be classified as meeting targeted daily or weekly goals related to physical inactivity. However, in research studies evaluating the role of inactive time on health-related outcomes, estimates that maximize stability over time (i.e., less day-to-day variability) might be more advantageous in detecting significant associations. In the current study, results showed that accelerometer-derived estimates of physical inactivity became more stable when 1) expressed as proportions or rates (i.e., adjusted for total wear time or inactive time) rather than total duration or accumulation of inactive-to-active transitions or 2) summed across multiple days of observation (i.e., 4–7 d). Thus, the adjusted values may provide a better period estimate of physical inactivity, which, in turn, may better relate to health outcomes that develop over time (e.g., obesity).

There are important limitations to consider when interpreting the results of the current investigation. First, the study population consisted of a convenience sample of healthy middle-age women who consented to participate in multiple physical activity measures, including five consecutive weeks of accelerometer wear, which may limit the generalizability of the findings to more diverse populations. Also, because behaviors can vary drastically from one population to another, it may be more appropriate to characterize patterns of physical inactivity in specific subgroups so that any relevant consideration can be applied when processing derived estimates or analyzing data in individuals or across populations. Second, summary estimates for sedentary behavior were classified and computed on the basis of accumulated time at <100 counts. Given that waist-worn activity monitors cannot accurately differentiate sedentary (i.e., sitting) time from other nonambulatory activities (e.g., lying down or standing), it may be less appropriate to characterize sedentary time (vs inactive time) using a cut point classification of <100 counts. Previous works (1,6) suggest that sitting may be physiologically different from standing (i.e., requires isometric contraction of antigravity or postural muscles). Given that the underlying biological plausibility with health outcomes may differ depending on whether the individual spent inactive time sitting or standing, investigators should consider incorporating postural recognition devices (e.g., activPAL PAL Technologies Ltd., Glasgow, UK) into their assessment strategy. However, because of issues of associated cost and the need for additional participant instruction about proper wear and monitor placement, incorporating these devices into large epidemiological studies may be less feasible at this time.

This study describes hourly, daily, and weekly patterns of accelerometer-derived physical inactivity estimates in healthy middle-age women. Overall, the total volume of inactive time and frequency of inactive-to-active transitions were higher during weekdays when compared with weekend days. Further, the estimates became more stable after adjustment for daily wear or inactive time and when averaged over the week. The results from the current study suggest that researchers and practitioners should carefully consider the intended application of accelerometer-based estimates of physical inactivity or sedentary behavior when developing data collection protocols and analytic procedures to final data reporting.

This research was funded by the American College of Sports Medicine Paffenbarger-Blair Endowment for Epidemiological Research on Physical Activity that was awarded to Dr. Kelley Pettee Gabriel while working as a postdoctoral research associate at Arizona State University.

The authors declare no conflicts of interest.

The authors thank the 66 dedicated PAW Study participants and study staff.

The results of the present study do not constitute endorsement by the American College of Sports Medicine.

Back to Top | Article Outline


1. Bey L, Hamilton MT. Suppression of skeletal muscle lipoprotein lipase activity during physical inactivity: a molecular reason to maintain daily low-intensity activity. J Physiol. 2003; 551 (Pt 2): 673–82.
2. Dunstan DW, Salmon J, Owen N, et al.. Associations of TV viewing and physical activity with the metabolic syndrome in Australian adults. Diabetologia. 2005; 48 (11): 2254–61.
3. Edwards D, Berry JJ. The efficiency of simulation-based multiple comparisons. Biometrics. 1987; 43 (4): 913–28.
4. Foster JA, Gore SA, West DS. Altering TV viewing habits: an unexplored strategy for adult obesity intervention? Am J Health Behav. 2006; 30 (1): 3–14.
5. Gabriel KP, McClain JJ, Schmid KK, et al.. Issues in accelerometer methodology: the role of epoch length on estimates of physical activity and relationships with health outcomes in overweight, post-menopausal women. Int J Behav Nutr Phys Act. 2010; 7: 53.
6. Hamilton MT, Hamilton DG, Zderic TW. Role of low energy expenditure and sitting in obesity, metabolic syndrome, type 2 diabetes, and cardiovascular disease. Diabetes. 2007; 56 (11): 2655–67.
7. Healy GN, Dunstan DW, Salmon J, et al.. Breaks in sedentary time: beneficial associations with metabolic risk. Diabetes Care. 2008; 31 (4): 661–6.
8. Hu FB, Leitzmann MF, Stampfer MJ, Colditz GA, Willett WC, Rimm EB. Physical activity and television watching in relation to risk for type 2 diabetes mellitus in men. Arch Intern Med. 2001; 161 (12): 1542–8.
9. Hu FB, Li TY, Colditz GA, Willett WC, Manson JE. Television watching and other sedentary behaviors in relation to risk of obesity and type 2 diabetes mellitus in women. JAMA. 2003; 289 (14): 1785–91.
10. Matthews CE. Calibration of accelerometer output for adults. Med Sci Sports Exerc. 2005; 37 (11 suppl): S512–22.
11. Matthews CE, Ainsworth BE, Thompson RW, Bassett DR Jr. Sources of variance in daily physical activity levels as measured by an accelerometer. Med Sci Sports Exerc. 2002; 34 (8): 1376–81.
12. 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 (7): 875–81.
13. Meijer GA, Westerterp KR, Verhoeven FM, Koper HB, ten Hoor F. Methods to assess physical activity with special reference to motion sensors and accelerometers. IEEE Trans Biomed Eng. 1991; 38 (3): 221–9.
14. Metzger JS, Catellier DJ, Evenson KR, Treuth MS, Rosamond WD, Siega-Riz AM. Patterns of objectively measured physical activity in the United States. Med Sci Sports Exerc. 2008; 40 (4): 630–8.
15. National Institutes of Health. National Cancer Institute: Risk Factor Monitoring and Methods. [cited 2011 Jun 13]. Available from:
16. Nichols JF, Morgan CG, Chabot LE, Sallis JF, Calfas KJ. Assessment of physical activity with the Computer Science and Applications, Inc., accelerometer: laboratory versus field validation. Res Q Exerc Sport. 2000; 71 (1): 36–43.
17. Owen N, Healy GN, Matthews CE, Dunstan DW. Too much sitting: the population health science of sedentary behavior. Exerc Sport Sci Rev. 2010; 38 (3): 105–13.
18. Pettee Gabriel KK, McClain JJ, Lee CD, et al.. The evaluation of physical activity measures used in middle-aged women. Med Sci Sports Exerc. 2009; 41 (7): 1403–12.
19. Pettee Gabriel KK, Rankin RL, Lee CD, Charlton ME, Swan PD, Ainsworth BE. Test–retest reliability and validity of the 400-meter walk test in healthy, middle-aged women. J Phys Act Health. 2010; 7 (5): 649–57.
20. Troiano RP, Berrigan D, Dodd KW, Mâsse LC, Tilert T, McDowell M. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc. 2008; 40 (1): 181–8.
21. Trost SG, Pate RR, Freedson PS, Sallis JF, Taylor WC. Using objective physical activity measures with youth: how many days of monitoring are needed? Med Sci Sports Exerc. 2000; 32 (2): 426–31.
22. Tukey JW. The philosophy of multiple comparisons. Stat Sci. 1991; 6 (1): 100–16.
23. Ward DS, Evenson KR, Vaughn A, Rodgers AB, Troiano RP. Accelerometer use in physical activity: best practices and research recommendations. Med Sci Sports Exerc. 2005; 37 (11 suppl): S582–8.
24. Warren TY, Barry V, Hooker SP, Sui X, Church TS, Blair SN. Sedentary behaviors increase risk of cardiovascular disease mortality in men. Med Sci Sports Exerc. 2010; 42 (5): 879–85.
25. Welk GJ. Principles of design and analyses for the calibration of accelerometry-based activity monitors. Med Sci Sports Exerc. 2005; 37 (11 suppl): S501–11.
26. Wickel EE, Welk GJ. Applying generalizability theory to estimate habitual activity levels. Med Sci Sports Exerc. 2010; 42 (8): 1528–34.


©2012The American College of Sports Medicine