Extensive research has established that increased physical activity (PA) is beneficial for health and reduces the risk of chronic illnesses, including cardiovascular disease (CVD), diabetes, and various cancers (12,14,34). Self-reported PA levels have been shown to be related to physiological biomarkers of health, including lipid, inflammation, hyperinsulinemia, and blood pressure indices (18). Much of the evidence for such associations and for the current U.S. national guidelines recommending that adults engage in 150 min·wk−1 of moderate to vigorous PA (MVPA), preferably spread throughout the week, has been based on self-reported PA (34).
However, prior research has indicated that self-reported measures of PA are only weakly associated with objectively measured PA (11,27,31,32). In addition, self-reported measures of PA provide substantially higher estimates of those adhering to PA recommendations compared with objectively measured activity using accelerometers (33). Research using convenience samples suggests that self-reported (i.e., International Physical Activity Questionnaire) and objectively measured (i.e., pedometer) PA independently correlate with health-related biomarkers and that objectively measured PA is more consistently related to biomarkers (27). However, to our knowledge, no population-based studies have examined self-reported PA versus objectively measured PA in relation to health-related biomarkers. Objective measurement of PA using accelerometers was implemented in the National Health and Nutrition Examination Survey (NHANES) in 2003-2004 and 2005-2006 (with support from the National Cancer Institute of the National Institutes of Health). The present study examined the independent associations of self-reported (i.e., NHANES questionnaire) and objectively measured (i.e., accelerometers) MVPA with physiological and anthropometric biomarkers of health in a representative sample of U.S. adults. Biomarkers within NHANES were chosen on the basis of their common identification as risk factors across chronic diseases and their documented association with PA and included blood pressure measures (CVD risk factor), anthropometric measures (CVD, diabetes, and cancer risk factor), cholesterol measures (CVD risk factor), glycemic control measures (linked to CVD, diabetes, and cancer risk), inflammation measure (C-reactive protein; linked to CVD and cancer risk), and homocysteine (CVD risk factor).
The NHANES uses a complex sampling design to produce a representative sample of the U.S. civilian noninstitutionalized (both children and adults) population and oversamples low-income respondents, adolescents, persons 60 yr and older, African Americans, and Mexican Americans. The NHANES includes an in-person home interview and a visit to a mobile examination center (MEC) where laboratory and examination data are collected. The interview data include demographic, socioeconomic, and health-related questions. The examination component includes medical and physiological measurements as well as laboratory tests. The Centers for Disease Control and Prevention Ethics Review Board approved the survey protocols, and informed consent was obtained for all subjects. This analysis combined NHANES data across the 2003-2004 and 2005-2006 data sets. For these two cycles of NHANES, a total of 9515 respondents 20 yr and older were interviewed and examined. Accelerometer data were obtained from 8077 individuals. We excluded from analyses participants who had accelerometer data with less than 4 d of valid data (n = 1984). Women who were pregnant or lactating were also excluded from analyses (n = 296). Adults 20 yr or older (N = 5797) with data on self-reported PA and 4 d or more of accelerometer measured PA were analyzed in the present study.
A set of sociodemographic variables were used as covariates in the analyses. These included age (as a continuous variable), gender, race/ethnicity recoded as a dichotomous variable (white vs non-white), and education in three levels (less than high school, high school diploma/general equivalency diploma, and more than high school).
These variables included a dichotomous smoking status item (current smoker, not current smoker), body mass index (BMI = weight in kilograms/height in square meters), used as continuous variable), and general health condition (from 1= poor to 5 = excellent).
Variables representing different possible diagnoses were also included. These were all coded as dichotomous variables, indicating if respondents had ever been diagnosed with the following medical problems: 1) diabetes, 2) high blood pressure, 3) osteoporosis/brittle bones, 4) CHD, 5) angina/angina pectoris, and 6) heart attack.
Two measures of PA were used in this analysis. The self-report PA in the NHANES included questions assessing the mode, frequency, and duration of recreation (e.g., jogging, swimming, cycling, soccer, dancing, etc.), household (e.g., raking leaves, mowing the lawn, etc.), and transportation (e.g., walk or bicycle as part of getting to and from work, etc.) activity over the 30 d before the interview. Each type of PA was subsequently coded to reflect exercise intensity on the basis of estimated MET values for each activity. We conducted analyses with accelerometer wear time included as a covariate (data not shown); the pattern of results between the PA measures and the biomarkers did not change with wear time included as a covariate. For objective (accelerometry) PA, all ambulatory examined participants were asked to wear an ActiGraph (ActiGraph, LLC, Ft. Walton Beach, FL) model 7164 accelerometer over the right hip on an elasticized belt for the 7 d after their MEC examination. Participants were asked to wear the device while they were awake and to take it off for swimming or bathing. Monitors were returned by express mail to the NHANES contractor, where data were downloaded and the device was checked to determine whether it was still within manufacturer's calibration specifications using an ActiGraph calibrator. The details of the accelerometer protocol are available at the CDC Web site (5), and the SAS code for aggregating ActiGraph 7164 data is available at the National Cancer Institute Web site (19). The uniaxial ActiGraph measures and records vertical acceleration as "counts," providing an indication of intensity of PA associated with locomotion (35). Data were recorded in 1-min epochs for up to 1 wk. A detailed description of the validity of the accelerometer data is described elsewhere (33). Briefly, we used a modified 10-min bout requirement to define valid data; 10-min activity bouts were defined as 10 consecutive minutes or more above the relevant threshold, with allowance for interruptions of 1 or 2 min below threshold. A bout was terminated by 3 min below threshold. For the analyses presented here, a valid day was defined as having 10 h or more of monitor wear. Wear time was determined by subtracting nonwear time from 24 h. Nonwear was defined by an interval of at least 60 consecutive minutes of zero activity intensity counts, with allowance for 1-2 min of counts between 0 and 100. Prior research suggests that self-reported and objectively measures of PA, as measured by NHANES, discriminate between patients with hypercholesterolemia who reported increasing PA on the basis of physician advice (compliant) versus those who did not increase PA (noncompliant) (9). Consistent with prior research (33), this study focused on average minutes per day of MVPA. Both self-reported and objectively measured PA values were divided by 10 so that each unit represents 10 min·d−1 of MVPA.
Clinically measured or laboratory-based biomarkers from NHANES were used as outcome measures. Clinically measured biomarkers included systolic blood pressure (SBP), diastolic blood pressure (DBP), BMI, waist circumference, triceps skinfold, and subscapular skinfold. Laboratory-based biomarkers included total cholesterol, HDL, LDL, triglyceride, glycohemoglobin, plasma glucose, C-peptide, insulin, C-reactive protein, and homocysteine. Further details of these measurements can be found on the NHANES Web site (4).
All analyses were conducted accounting for the complex sample design used by NHANES. The National Center for Health Statistics supplied values of stratum and primary sampling units used to account for clustering of the sample. Adjusted sampling weights were used to account for oversampling of specific populations. Weights were adjusted for nonresponse by poststratifying the sample of adults with 4 d or more of valid accelerometer data to sex, age, and race/ethnicity control totals for the full MEC examined NHANES sample. All analyses were conducted using SAS (SAS Institute, Inc., Cary, NC) and SUDAAN (Research Triangle Park, NC).
Analyses were completed in three steps. The first step was to conduct a series of simple linear regression models, regressing each outcome on each PA index separately, using age as a covariate. The second step involved conducting multivariate linear regression models, regressing each outcome on both PA indices while adjusting statistically for sociodemographic, health behavior/status, and relevant clinical diagnosis variables. A third set of regressions was conducted in which BMI was added as a covariate for models where anthropometric indices were not outcomes. Listwise deletion was used for all analyses, and statistical significance was assessed with two-tailed tests with alpha = 0.05. We also conducted analyses on only those respondents without a medical condition, and the pattern of results (not shown) was identical with the results with the full sample.
The average daily minutes of self-reported MVPA was 54.8 (SE = 1.9), and the average daily minutes of objectively measured MVPA was 6.7 (SE = 0.3). The accelerometer cut point of 2020 counts per minute was used to define moderate-intensity PA. The average age of respondents was 46.9 yr (SE = 0.5), and the average BMI was 28.2 (SE = 0.2), with 66% of respondents being overweight or obese. Approximately half (49%) of respondents were men and 73% were non-Hispanic white. One-fourth of the sample had a high school diploma/general equivalency diploma, and 59% had attended college. The demographic characteristics of the analytic sample were virtually identical with those of the full sample (data not shown). Thirty-six percent of respondents reported ever having a medical condition (e.g., high blood pressure, osteoporosis, CHD, diabetes, etc.). Weighted mean values of the biomarkers were as follows: SBP = 123.0 (SE = 0.35), DBP = 71.3 (SE = 0.26), waist circumference = 97.0 cm (SE = 0.35 cm), triceps skinfold = 19.4 mm (SE = 0.18 mm), subscapular skinfold = 20.0 mm (SE = 0.18 mm), total cholesterol = 199.6 mg·dL−1 (SE = 0.60 mg·dL−1), HDL = 54.4 mg·dL−1 (SE = 0.30 mg·dL−1), LDL = 115.6 mg·dL−1 (SE = 0.90 mg·dL−1), triglyceride = 144.4 mg·dL−1 (SE = 3.23 mg·dL−1), glycohemoglobin = 5.5% (SE = 0.02%), glucose = 102.2 mg·dL−1 (SE = 0.75 mg·dL−1), c-peptide = 0.8 nmol·L−1 (SE = 0.01 nmol·L−1), insulin = 10.9 μU·mL−1 (SE = 0.25 μU·mL−1), C-reactive protein = 0.4 mg·dL−1 (SE = 0.01 mg·dL−1), and homocysteine = 9.1 μmol·L−1 (SE = 0.13 μmol·L−1).
The age-adjusted simple regressions indicated that self-reported PA was significantly related to anthropometric measures, HDL, glycohemoglobin, insulin, and C-reactive protein (Table 1). Objectively measured PA was significantly associated with most of the physiological biomarkers (except for DBP, total cholesterol, and LDL). Overall, the simple regressions suggest that objectively measured PA shows stronger associations with the various biomarkers than did self-reported PA.
Results of the multivariate analyses including both self-reported and objectively measured PA are presented in Table 2. Model 1 results indicate that both self-reported and objectively measured PA are independently associated with skinfold measures, HDL, and C-reactive protein, even after adjusting statistically for age, gender, race, education, smoking, general perceived health, diabetes, high blood pressure, osteoporosis, CHD, angina, and heart disease. Objectively measured PA was also independently associated with SBP, BMI, waist circumference, triglyceride, glycohemoglobin, plasma glucose, C-peptide, insulin, and homocysteine. In contrast, the significant relationships of self-reported PA with BMI, waist circumferences, glycohemoglobin, and insulin noted in the simple regressions became nonsignificant in the multivariate analyses. Both self-reported and objectively measured PA remained associated with HDL and C-reactive protein, even after adjustment for BMI. However, the relationships between objectively measured PA with SBP, triglyceride, and C-peptide became nonsignificant after adjusting for BMI.
We conducted exploratory analyses to test whether varying the criteria for activity intensity and bout duration in accelerometry could affect the significance of the association of self-reported PA with skinfold measures, HDL, and C-reactive protein in the presence of the objectively measured PA. In the first set of analyses, the accelerometer cut point was lowered to 760 to include lifestyle/light-intensity PA in addition to MVPA. In the second set of analyses, we maintained the original accelerometer cut point for MVPA but removed the 10-min bout requirement. Results (data not shown) indicated that both self-reported and objectively measured PA remained significantly associated with skinfold measures, HDL, and C-reactive protein when activity intensity and bout duration in accelerometry were modified.
Objective and subjective measures of PA in relation to biomarkers.
In this nationally representative sample of U.S. adults with measures of self-reported and accelerometer-based PA, objectively measured MVPA displayed much stronger associations with multiple biomarkers and anthropometric indices than did self-reported MVPA even after adjusting for numerous potential confounders. For several biomarkers, when self-reported and objectively measured MVPA were examined simultaneously, only objectively measured MVPA showed significant associations (BMI, waist circumferences, glycohemoglobin, insulin, and homocysteine). However, for several other biomarkers (skinfold indices, HDL, and C-reactive protein), associations were observed with both objectively measured and self-reported MVPA. Taken together, these findings suggest that self-reported and objectively measured PA may capture distinct aspects of PA that are associated with biomarkers of health, although objectively measured PA demonstrates significantly stronger associations.
Our findings are consistent with prior research suggesting objectively measured PA to be more strongly related with self-reported indicators of health (11) as well as clinically measured cardiometabolic factors (27) assessed in convenience samples. However, the present study used a nationally representative sample, examined a broader range of clinically measured or laboratory-based biomarkers, and adjusted statistically for various demographic, socioeconomic, and health/disease factors. Moreover, we adjusted for BMI to reduce the possibility that significant results were due to differences in body weight. Further adjusting for BMI had little effect on the independent relationships of self-reported and objectively measured PA with HDL and C-reactive protein. These findings support prior evidence that the health benefits of PA may be distinct from their effects on body weight and the relationship of body weight on health (14,30).
Self-reported PA appears to capture aspects of PA that are not assessed by accelerometer measured PA. The two measurement approaches are inherently different in that the questionnaire assesses behaviors as perceived by the respondent in terms of intensity, frequency, and duration, while the accelerometer measures movement and applies absolute criteria to determine activity intensity. The two methods also capture various activities to different extents. A possible explanation is that very low or intermittent types of PA may be included by individuals reporting PA but are not included in the accelerometer measured PA for this study because of the cut points we used. However, analyses in which activity intensity and bout duration in objectively measured PA were modified to include more low or intermittent types of PA argue against this possible explanation. Specifically, both self-reported and objectively measured PA remained independently associated with HDL and C-reactive protein when activity intensity and bout duration in accelerometry were modified.
An alternative explanation for the independent association of self-reported PA may be that higher levels of this measure serve as a marker for greater muscular strength (not assessed in the NHANES). Muscular strength has been noted to be associated with health and mortality independent of aerobic fitness (22,25), and prior research has found that resistance training has to be related to body composition (8,28,29), lipid profiles (15), and inflammatory markers (1,13,17,23). However, because it has been estimated that 20% of U.S. adults engage in strength/resistance training two times per week or more (6) and 5% engage in yoga for health (2), strength training may not completely account for the independent associations between self-reported PA and the biomarkers. Another explanation is that reporting higher levels of PA may be associated with an unexamined factor (e.g., optimistic outlooks on life (7,24) or dietary patterns (16,21,26)), which may contribute to better physiological health independent of objectively measured PA. Future research is warranted to more fully disentangle the independent associations of self-reported and objectively measured PA with various biomarkers. It may be useful to examine different aspects of activity (i.e., aerobic exercise, strength training, flexibility activities, etc.), as measured by self-report and objective measures, to assess whether various biomarkers are differentially related to specific types of activities.
This study has several limitations. The NHANES surveys are cross-sectional, limiting causal inference. Furthermore, the use of a single waist-mounted uniaxial accelerometer likely missed activities that primarily involved upper body movement and cannot capture the effects of weight bearing or load carrying on energy expenditure or account for grade change in walking, which may influence self-report. Nonetheless, objectively measured activity displayed stronger associations with multiple biomarkers. The self-reported PA was based on recall of the past 30 d of activities that were of at least 10 min in duration. Measurement error resulting from social desirability, memory distortions, and/or other cognitive biases may have attenuated the magnitude of associations between the self-reported PA and the various biomarkers (10). However, even after adjusting statistically for numerous covariates, objectively measured PA, and BMI, the associations of self-reported PA with HDL and C-reactive protein remained significant. It is possible that the relationships between the biomarkers and activity indicators are specific to the self-report and objective measures used in the NHANES or PA characteristics examined in this study. Additional research is needed to determine whether these results occur with other self-report and objective activity indicators as well as with other characteristics of activity (e.g., longest bout length, average intensity, frequency of activity, etc.).
In this study, the self-report measure provided significantly higher estimates of MVPA compared with accelerometry, a finding consistent with prior research (33). Because there is not currently a gold standard for assessing total MVPA in a large sample of free-living adults, it seems likely that the true level of MVPA is somewhere between these two estimates (33). Importantly, results from the present study suggest that researchers cannot assume that accelerometers are capturing the same aspects of PA as self-reported measures. Although modest overlap does exist between self-reported and objectively measured PA (r = 0.20-0.40), simply replacing self-report measures with accelerometers may not be warranted, given the independent associations noted between different measures and various outcomes. Instead, results support prior recommendations to include both self-reported and objective PA measures when examining the effects of PA behavior on health outcomes (10,11) and that research question and design considerations (e.g., cost, sampling, etc.) are all factors in deciding the measurement approach for any particular study.
Given the consistently stronger relationships between objectively measured PA and biomarkers noted in this study, further research is needed to determine optimal PA levels as measured by accelerometry; these levels may well differ from what has been estimated largely on the basis of self-reported measures of PA. The stronger relationships found between objectively measured PA and the various biomarkers may be due to the possibility that accelerometers are capturing occupational activity that could not be incorporated in the self-reported PA index, given the general nature of the occupation activity question in the NHANES. Prior research suggests that half of respondents in a national sample classified as sedentary using leisure-time measures report that they engage in 1 h of occupational activity (3). Additional research focused on determining the influence of occupational activity on biomarkers may help to disentangle this issue. The development of integrated mobile tools that can assess objective and self-reported PA concurrently (e.g., NIH Genes, Environment and Health Initiative (20)) may assist in further disentangling relationships of self-reported and objectively measured PA with health outcomes. In addition, new triaxial accelerometers and programming may allow researchers to determine specific types of PA from accelerometers rather than just intensity and duration. The inclusion of such measures to future NHANES surveys holds promise in contributing to understanding the associations of PA to intermediate biomarkers and health outcomes at the population level.
This study was not supported by external funding. The U.S. National Cancer Institute provided funds for the data collection.
The authors acknowledge Drs. Susan M. Krebs-Smith and Heather Bowles, National Cancer Institute Division of Cancer Control and Population Sciences, for critical comments on the manuscript.
No author reports any competing financial interest.
Conflict-of-interest notification: none to report.
The National Cancer Institute reviewed and approved this article before submission. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Department of Health and Human Services, the National Institutes of Health, or the National Cancer Institute.
The results of the present study do not constitute endorsement by the American College of Sports Medicine.
1. Balducci S, Zanuso S, Nicolucci A, et al. Anti-inflammatory effect of exercise training in subjects with type 2 diabetes and the metabolic syndrome is dependent on exercise modalities and independent of weight loss. Nutr Metab Cardiovasc Dis
2. Barnes PM, Powell-Griner E, McFann K, Nahin RL. Complementary and alternative medicine use among adults: United States, 2002. Adv Data
3. Centers for Disease Control and Prevention. Prevalence of leisure-time and occupational physical activity
among employed adults-United States, 1990. MMWR Morb Mortal Wkly Rep
4. Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey. Atlanta (GA): Centers for Disease Control and Prevention; [cited March 1, 2010]. Available from: http://www.cdc.gov/nchs/nhanes/nhanes2003-2004/nhanes03_04.htm
5. Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey (NHANES), Laboratory Procedures Manual. Atlanta (GA): Centers for Disease Control and Prevention; [cited March 1, 2010]. Available from: http://www.cdc.gov/nchs/data/nhanes/nhanes_03_04/lab_pm.pdf
6. Centers for Disease Control and Prevention. Trends in strength training-United States, 1998-2004. MMWR Morb Mortal Wkly Rep
7. Chida Y, Steptoe A. Positive psychological well-being and mortality: a quantitative review of prospective observational studies. Psychosom Med
8. Chromiak JA, Smedley B, Carpenter W, et al. Effect of a 10-week strength training program and recovery drink on body composition, muscular strength and endurance, and anaerobic power and capacity. Nutrition
9. Fan AZ, Ham SA, Muppidi SR, Mokdad AH. Validation of reported physical activity
for cholesterol control using two different physical activity
instruments. Vasc Health Risk Manag
10. Ferrari P, Friedenreich C, Matthews CE. The role of measurement
error in estimating levels of physical activity
. Am J Epidemiol
11. Harris TJ, Owen CG, Victor CR, Adams R, Ekelund U, Cook DG. A comparison of questionnaire, accelerometer, and pedometer: measures in older people. Med Sci Sports Exerc
12. 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
13. Heffernan KS, Jae SY, Vieira VJ, et al. C-reactive protein and cardiac vagal activity following resistance exercise training in young African-American and white men. Am J Physiol Regul Integr Comp Physiol
14. IARC Working Group. IARC Working Group on the Evaluation of Cancer-Preventive Strategies. IARC Handbook of Cancer Prevention
. Lyon (France): IARC Press; 2002. p. 315.
15. Kelley GA, Kelley KS. Impact of progressive resistance training on lipids and lipoproteins in adults: another look at a meta-analysis using prediction intervals. Prev Med
16. McNaughton SA, Dunstan DW, Ball K, Shaw J, Crawford D. Dietary quality is associated with diabetes and cardio-metabolic risk factors. J Nutr
17. Moinuddin I, Leehey DJ. A comparison of aerobic exercise and resistance training in patients with and without chronic kidney disease. Adv Chronic Kidney Dis
18. Mora S, Cook N, Buring JE, Ridker PM, Lee IM. Physical activity
and reduced risk of cardiovascular events: potential mediating mechanisms. Circulation
19. National Cancer Institute. SAS Programs for Analyzing NHANES 2003-2004 Accelerometer Data. Bethesda (MD): National Cancer Institute; [cited March 1, 2010] Available from: http://riskfactor.cancer.gov/tools/nhanes_pam/
20. National Institutes of Health. Genes, Environment and Health Initiative (GEI)-Exposure Biology Program-Improved Measures of Diet and Physical Activity
. Bethesda (MD): National Institutes of Health; [cited March 1, 2010]. Available from: http://www.gei.nih.gov/exposurebiology/program/improved.asp
21. Nettleton JA, Schulze MB, Jiang R, Jenny NS, Burke GL, Jacobs DR Jr. A priori-defined dietary patterns and markers of cardiovascular disease risk in the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Clin Nutr
22. Newman AB, Kupelian V, Visser M, et al. Strength, but not muscle mass, is associated with mortality in the health, aging and body composition study cohort. J Gerontol A Biol Sci Med Sci
23. Olson TP, Dengel DR, Leon AS, Schmitz KH. Changes in inflammatory biomarkers following one-year of moderate resistance training in overweight women. Int J Obes (Lond)
24. Rasmussen HN, Scheier MF, Greenhouse JB. Optimism and physical health: a meta-analytic review. Ann Behav Med
25. Ruiz JR, Sui X, Lobelo F, et al. Muscular strength and adiposity as predictors of adulthood cancer mortality in men. Cancer Epidemiol Biomarkers Prev
26. Samaha F. Effects of a Mediterranean-style diet on cardiovascular risk factors. Ann Intern Med
27. Schmidt MD, Cleland VJ, Thomson RJ, Dwyer T, Venn AJ. A comparison of subjective and objective measures of physical activity
and fitness in identifying associations with cardiometabolic risk factors. Ann Epidemiol
28. Shaw I, Shaw BS. Consequence of resistance training on body composition and coronary artery disease risk. Cardiovasc J S Afr
29. Sillanpaa E, Laaksonen DE, Hakkinen A, et al. Body composition, fitness, and metabolic health during strength and endurance training and their combination in middle-aged and older women. Eur J Appl Physiol
30. Sui X, LaMonte MJ, Laditka JN, et al. Cardiorespiratory fitness and adiposity as mortality predictors in older adults. JAMA
31. Treuth MS, Sherwood NE, Baranowski T, et al. Physical activity
self-report and accelerometry
measures from the Girls health Enrichment Multi-site Studies. Prev Med
. 2004;(38 suppl):S43-9.
32. Troiano R, Dodd K. Differences between objective and self-report measures of physical activity
: what do they mean? Korean J Meas Eval Phys Ed Sports Sci
33. Troiano RP, Berrigan D, Dodd KW, Masse LC, Tilert T, McDowell M. Physical activity
in the United States measured by accelerometer. Med Sci Sports Exerc
34. U.S. Department of Health and Human Services. Physical Activity
Guidelines Advisory Committee. Physical Activity
Guidelines Advisory Committee Report, 2008. In: Book Physical Activity Guidelines Advisory Committee. Physical Activity Guidelines Advisory Committee Report, 2008
. Washington (DC): Department of Health and Human Services; 2008. p. 683.
35. Welk GJ. Use of accelerometry
-based activity monitors to assess physical activity
. In: Welk GJ, editor. Physical Activity Assessments for Health-Related Research
. Champaign (IL): Human Kinetics; 2002. p. 125-39.