Predicting Walking METs and Energy Expenditure from Speed... : Medicine & Science in Sports & Exercise (original) (raw)
Walking is both the most commonly reported physical activity in population surveys (31) and the primary exercise recommended by public health authorities. It has been described as the “nearest activity to perfect exercise” because walking requires no equipment, is low impact, and is usually performed at a sufficient intensity to confer health benefits. The accuracy of exercise intensity (V̇O2 mL·kg−1·min−1 or METs [1 MET = V̇O2 of 3.5 mL·kg−1·min−1]) and energy cost (kcal·min−1) predictions for walking are therefore of great consequence.
Historically, walking intensity and energy expenditure were predicted via combinations of speed and body mass (26). However, pedometers, motion sensors, and accelerometers introduced over the last decade have superseded speed for predicting walking intensity. The limitations of the aforementioned devices include their inability to: detect changes in incline or load during walking (15,24), identify the initiation or cessation of walking, and accurately predict the energy requirements of activities other than level terrain walking (3,20,28,33).
The MTI actigraph (Manufacturing Technologies Inc., Ft. Walton Beach, FL; formerly the CSA monitor) is a common uniaxial accelerometer that has been the focus of many walking validation studies (6,12,14,17,20,22,24,25,28,30,35). However, only Brage et al. (6) noted that “walking/running speed may be an equally, or even a better estimate of the total metabolic turnover” than the CSA monitor. The literature suggests that level-terrain walking intensity (METs) and energy expenditure (kcal·min−1) can be predicted with comparable accuracy via speed (2,5,8,23,27,32) or CSA accelerometry (6,12,14,17,20,22,24,25,28,30,35), but nobody has compared the accuracy of these two predictors on the same sample.
Our hypotheses are: 1) speed will display a stronger association with level terrain walking METs and energy expenditure than CSAhip (Computer Science Applications accelerometer positioned on the hip); 2) cross-validation of speed- and CSAhip-based prediction equations will confirm that the former is a superior predictor of walking METs and energy expenditure; 3) energy cost per kilogram of body mass and kilometer walked (i.e., kcal·kg−1·km−1) should be constant in our sample of middle-aged men and women.
METHODS
Subjects.
Walking data were collected on 72 middle aged volunteers (Table 1) during a larger study on the energy expenditure of household/garden activities. Household/garden activity data for the men and women are presented in Gunn et al. (19) and Brooks et al. (7), respectively.
Descriptive statistics for the 35- to 45-yr-old men (N = 36) and women (N = 36).
All volunteers were recruited from the local community via advertisements and word of mouth. Current smokers and persons suffering from diseases or taking medication known to affect energy metabolism were excluded. We attempted to recruit a sample that had a QI (Quetelet’s Index; kg·m−2) distribution similar to that of the Australian population (1). Our sample size allowed us to estimate population self-paced walking means within a 95% confidence interval of ± 0.15 METs (11). This project was approved by the Clinical Research Ethics Committee of the Flinders Medical Centre. All experimental procedures, possible risks, and benefits were explained to the subjects before their written informed consent was obtained.
Experimental design.
Volunteers were instructed to walk around a level, paved, and sheltered quadrangle (perimeter = 141 m) at what they perceived to be a “moderate pace” for 15 min. Walking was performed between 10 a.m. and 12 noon. To minimize the confounding influences of learning, prior exercise, and food/caffeine intake, volunteers performed self-paced walking in a habituated, rested, and postabsorptive state. The effect of the menstrual cycle on energy metabolism and self-pacing was also reduced by testing females between days 7 and 20 (day 1 = start) of their self-reported menstrual cycle.
Mass and height were determined using an electronic balance (model FW-150K; A&D Mercury Pty Ltd, South Australia) and wall stadiometer, respectively. Fat free mass (FFM), fat mass (FM), and percent body fat (% BF) were measured via dual energy x-ray absorptiometry. Lunar DPX-L (Lunar Corp, Madison, WI) and Prodigy (General Electric, Madison, WI) scanners measured the aforementioned body composition variables in 61 and 11 volunteers, respectively.
The “classical” Douglas bag method was used to calculate oxygen consumption during walking. METs were determined by dividing a resting V̇O2 constant of 3.5 mL·kg−1·min−1 into walking oxygen consumption (mL·kg−1·min−1). Energy expenditure (kcal·min−1) was calculated using: kcal·min−1 = 15.913 + [(5.207 × respiratory exchange ratio) × V̇O2 (L·min−1)]/4.186 (13).
A continuous 15-min walk consisted of a 5-min warm-up (minutes 1–5), and two consecutive 5-min (minutes 5–10 and 10–15) expirate collections in aluminized, Mylar®-lined Douglas bags. Fractional concentrations of O2 and CO2 in mixed expirate were determined via calibrated Applied Electrochemistry S-3A oxygen and CD-3A carbon dioxide analyzers (Pittsburgh, PA), respectively. Expired volume was measured using a calibrated 350-L Tissot spirometer. It is salient to note that the Douglas bags and associated timing equipment were supported by the experimenter so that the total mass supported by the volunteer was ∼450 g. A detailed explanation of our methodology for measuring V̇O2 is contained in Gunn et al. (18).
Mean walking speed was calculated for minutes 5–15 of walking. Structural beams evenly spaced around the quadrangle allowed us to estimate distance walked to the nearest 2.5 m. The CSA monitor (model 7164) was attached to a waist belt and positioned on the volunteer’s right hip along their midaxillary line. CSA monitor output was verified using a calibration rig designed by the manufacturer (model CAL71). Heart rate (HR) was measured using a Polar X-Trainer Plus (Polar Electro OY, Kempele, Finland) that had been validated via a pulse generator. The HR and CSA accelerometer were temporally synchronized to V̇O2 measurement. Perceived exertion was estimated using a 15-point Borg rating scale (Borg RPE) at the end of the walking task.
Reliability and precision of measurement.
The reliability and precision of our V̇O2 measurement technique were assessed using a convenience sample of 12 volunteers (three men and nine women: mean ± SD = 43 ± 14 yr, 78 ± 16 kg, 172 ± 13 cm) who walked on a treadmill at 5.0 km·h−1 for 15 min. Intraclass correlations (ICC) and percent technical error of measurements (% TEM) were calculated for two consecutive (minutes 5–10 and 10–15) V̇O2 measurements after a 5-min warm-up.
The ability of an individual to reproduce their self-paced, moderate walking speed and V̇O2 on a separate day was assessed using a repeated walking trial for 36 volunteers from our experimental sample. The duplicate trials were separated by a median of 3 d (range: 1–8 d).
Statistics.
The 0.05 level was used for all tests of statistical significance. Independent _t_-tests were used to test for gender differences. Reliability and precision for V̇O2 measurement at a standardized (N = 12) and self-paced workload (N = 36) were examined using ICC and % TEM, respectively.
Pearson correlation coefficients (r) were calculated between METs or kilocalories per minute and CSAhip, speed, mass, FFM, FM, % BF, height, and gender. A bivariate linear prediction equation was developed to predict METs and energy expenditure from speed or CSAhip. Mass, gender, and % BF were added via a forward stepwise selection procedure. FFM, FM, and height were excluded as potential predictors because of their near collinearity with mass. A quadratic curve was also fitted to the speed–MET data because of a nonlinear relationship between METs and walking speeds > 6.0 km·h−1 (8,23,26).
Other prediction equations were cross-validated and assessed for goodness of fit (r2), standard deviation of differences (SDdiff) between measured and predicted values, mean difference (X̄diff), and range of individual differences. The best prediction equations were chosen by assessing the relative balance between explained variances, SDdiff, and X̄diff. CSAhip-based equations that required other predictors (6,24) and equations developed on children (14,30) or lifestyle activities (28) were excluded from cross-validation. A speed-based equation that predicted energy expenditure without using mass as a copredictor (2) was also excluded from cross-validation. The best speed- and CSAhip-based predictions from the literature are presented via Bland and Altman plots (4). Predictive accuracy was assessed by 95% prediction limits, calculated as mean ± 2SD of the differences between predicted and measured values.
RESULTS
Descriptive statistics.
Women were significantly (P < 0.001) shorter, lighter, and had a higher % BF than their male counterparts (Table 1). One percent, 58%, and 40% of our sample were in the underweight (QI < 18.5), normal (18.5 ≤ QI < 25), and overweight QI categories (QI ≥ 30) compared with 2, 47, and 50% in the 35- to 44-yr-old Australian population (1).
MET and energy expenditure prediction equations.
Walking intensity (METs) significantly correlated with speed (r = 0.78; Fig. 1), CSAhip (r = 0.72; Fig. 1), mass (r = −0.39), height (r = −0.30), FM (r = −0.28), and FFM (r = −0.28). Energy expenditure (kcal·min−1) significantly correlated with mass (r = 0.72), FFM (r = 0.60), speed (r = 0.50), height (r = 0.44), FM (r = 0.42), and CSAhip (r = 0.41). % BF did not correlate significantly with either self-paced walking METs (r = −0.13; P = 0.28) or energy expenditure (r = 0.09; P = 0.45) for the combined data, but the association between METs and % BF was significant for the female data (r = −0.43; P = 0.01).
FIGURE 1—Walking speed vs METs (:
left ) and CSAhip vs METs ( right ) for female ( broken line ), male ( dotted line ), and combined data ( solid line ).
Speed accounted for 61% of MET variance, with this value increasing to 72% when mass was added to the prediction equation (Table 2; Equations 1 and 2). Gender was also selected via forward stepwise selection (P = 0.046), but contributed only 1% to the explained MET variance. A quadratic speed-based equation explained 3% more MET variance than a linear equation (Table 2; Equations 1 and Quadratic). CSAhip accounted for 51% of MET variance, with this value increasing to 61% when mass was added (Table 2; Equations 4 and 5).
MET and energy expenditure equations (N = 72).
Speed and CSAhip explained 25% and 17% of the energy expenditure variance, respectively (Table 2; Equations 6 and 9). The majority of energy expenditure variance (52%) was explained by body mass. Table 2 shows that gender was selected via forward stepwise regression in the speed-based energy expenditure equation, but was nonsignificant in the CSAhip-based equation.
Cross-validation of other equations.
Table 3 presents other prediction equations and their cross-validation results for our measured METs and energy expenditure. Table 3 and Figures 2 and 3 highlight the superiority of speed-based over CSAhip-based predictions for both METs and energy expenditure (kcal·min−1).
Cross-validation of other MET and energy expenditure (kcal·min−1) prediction equations.
4 ) plots of the best speed- and CSAhip-based equations for predicting METs during walking; broken lines , 95% prediction limits; solid lines , mean error.
4 ) plots of the best speed- and CSAhip-based equations for predicting energy expenditure (kcal·min−1) during walking; broken lines , 95% prediction limits; solid lines , mean error.
Walking METs for our sample were predicted with the greatest accuracy from a speed-based quadratic equation published by Bubb et al. (8; Table 3). The equation underestimated our mean by 0.2 METs with 95% prediction limits of ± 0.7 METs (Fig. 2).
Walking energy expenditure (kcal·min−1) was predicted with greatest precision using equations that utilized both speed and mass (23; Table 3). The male and female specific equations resulted in a mean overestimation of 0.2 kcal·min−1 and 95% prediction limits of ± 1.04 kcal·min−1 (Fig. 3).
Reliability and precision of measurement.
The ICC and % TEM for consecutive V̇O2 measurements at 5.0 km·h−1 on a treadmill (N = 12) were 0.98 and 1.2%, respectively. Self-paced trials on separate days (N = 36) yielded an ICC and % TEM of 0.90 and 4.7%, respectively.
Self-paced walking intensity and energy expenditure.
Women walked at a faster mean speed than the men (P = 0.02), which resulted in their higher mean METs (P = 0.02), CSAhip (P = 0.02), HR (P < 0.001), and percentage of age predicted HRmax (P < 0.001; Table 4 (29)). There was no significant gender difference for Borg RPE (P = 0.52), and the females’ lower mean energy expenditure (kcal·min−1) was due to their smaller body mass. There was also no gender difference for the energy required to move 1 kg of body mass 1 km (kcal·kg−1·km−1) over level terrain with an overall coefficient of variation of ± 10.7% (Table 4).
METs, energy expenditure, and predictors for self-paced walking.
DISCUSSION
Prediction of level walking METs and energy expenditure.
We are the first group to compare speed- and CSAhip-based predictions of level terrain walking METs and energy expenditure. Contemporary studies have utilized pedometers, motion sensors and accelerometers, possibly because of a perceived increase in accuracy using new equipment. Our data suggest otherwise, with speed explaining 8–10% more variance in our walking METs and energy expenditure compared to CSAhip (Table 2). Cross-validation of previously published equations reinforced our conclusion (Table 3).
Our data.
Speed (r = 0.78) and CSAhip (r = 0.72) both correlated significantly with METs. The highly significant inclusion of mass in our MET prediction equations (Table 2; Equations 2 and 5 compared with Equations 1 and 4) was a surprise because walking METs should be independent of mass. In other words, a heavy and light person walking together should have identical METs (assuming the same mechanical efficiency) because the oxygen consumption required to move each kilogram of their respective body masses (i.e., mL O2·kg−1·min−1 or MET) would be the same. However, at the same walking speed, our equations predict a lower MET intensity for a heavier person compared with a lighter one. This leads us to conclude that dividing gross V̇O2 (L·min−1) by body mass does not fully standardize the V̇O2 of our volunteers, who varied significantly in body composition, stature, and mass. Gender was also selected for the speed-based MET prediction model but its inclusion impacted minimally on the explained variance and predictive precision (Table 2: Equation 3 compared with Equation 2). Gender’s greater negative weighting for women infers that, at a constant walking speed and mass, men have slightly higher METs than women. Even though McDonald (23) agrees that “the metabolic cost of walking is on average 12% greater for a man than for a woman of the same weight walking at the same speed,” the remaining literature (5,8) suggests no gender effect on METs when body mass and walking speed are held constant. Gender is therefore unlikely to enhance speed-based MET prediction for walking. Only 9 of our 72 volunteers walked at a pace ≥ 6.0 km·h−1, so a quadratic speed-based equation, which accounts for the nonlinear speed–MET relationship at higher walking speeds, did not improve predictive precision substantially compared with a linear equation (Table 2; Equation 1 and Quadratic). In contrast, previously published speed-based prediction equations are predominately curvilinear (2,8,27,32) because they were developed using a wider range of walking speeds (Table 3).
Our data support the association between body mass and walking energy expenditure (kcal·min−1; 23,32), with mass explaining a greater proportion of energy expenditure variance (52%) than either speed (25%) or CSAhip (17%). Mass therefore significantly impacted on the standard error of estimate (SEE) and explained variance when it was added to the bivariate speed- or CSAhip-based energy expenditure predictions (Table 2; Equations 7 and 10 compared with Equations 6 and 9). Similar to MET prediction, the final model for speed-based energy expenditure prediction included gender as a significant predictor but its effect on predictive precision was again negligible (Table 2; Equation 8 compared with Equation 7).
Our equations (Table 2) highlight the superiority of walking speed as a predictor of METs and energy expenditure. However, on a population basis they are unlikely to provide a more accurate estimate of walking METs and energy expenditure than other equations (Table 3) because ours were developed using only one point per subject and within a restricted range of walking speeds (4.0–6.5 km·h−1).
Cross-validation data.
Cross-validation of other speed- and CSAhip-based prediction equations revealed that speed-based MET and energy expenditure prediction equations were superior to those utilizing accelerometry (Table 3).
Figure 2 shows that the best speed-based MET prediction equation (8) predicted within tighter 95% prediction limits (± 0.7 METs) compared with the best CSAhip-based (20) equation (± 1.0 METs). Although the magnitude of mean differences was similar for both (speed: −0.2 METs; CSAhip: + 0.3 METs), speed-based prediction explained 13% more variance of our measured METs (Table 3).
Energy expenditure (kcal·min−1) was predicted with greatest precision (Fig. 3; 95% prediction limit = ± 1.0 kcal·min−1) and accuracy (Table 3; meanerror = 0.2 kcal·min−1) using gender specific speed-based equations that were developed from a review of historical walking studies (1912–58; 23). In contrast, two CSAhip-based energy expenditure prediction equations (12,17) overestimated mean energy expenditure by 1.2 and 1.7 kcal·min−1 (Table 3). Furthermore, the best CSA-based equation (12) predicted our kilocalories per minute within slightly wider 95% confidence limits (speed ± 1.0 kcal·min−1; CSAhip ± 1.2 kcal·min−1; Fig. 3) and explained less variance (speed, 82%; CSAhip, 66%) than the speed-based energy expenditure regression equation (23). It should be noted that the mean overestimation of energy expenditure by the Ekelund prediction equation (12; Table 3) is probably due to their different positioning of the CSA accelerometer. Ekelund et al. (12) developed their equation using a lower back placement of the CSA, which has been shown to elicit a lower CSA output compared with hip placement during walking (35). Even though it is therefore not a genuine CSAhip-based equation, it was included because it is far superior to the only other CSAhip-based energy expenditure prediction equation in the literature (17).
It is worthwhile noting that our cross-validation SDdiff are often smaller than the SEE reported for the original equations (Table 3) because cross-validation only occurred within a small range of speeds (4.0–6.5 km·h−1) for which the criterion–predictor relationship is strong. If we tested the predictive precision of the equations at higher walking speeds, where the criterion–predictor relationship weakens, cross-validation SDdiff would also increase.
Self-paced walking reliability and precision.
The reliability and precision statistics for self-paced walking METs in this study (ICC = 0.90; % TEM = 4.7%) were comparable to those previously reported by our team for 24 similarly aged volunteers (18; ICC = 0.93: % TEM = 6.1%). Although Bussman et al. (9) reported an ICC of 0.87 for self-paced walking speed, we could not locate any other report of the reliability and precision for self-paced walking METs. Our standard deviation of the day 1–2 differences for self-paced walking (0.26 METs) was comparable to that reported by Wergel-Kolmert and Wohlfart (0.28 METs; (34)) for a standardized treadmill speed of 5.0 km·h−1 in adolescent females. This suggests that our volunteers reproduced their self-paced walking intensity within an error margin similar to that obtained for a standardized treadmill workload in adolescent females.
The interday reliability for self-paced activities encompasses self-pacing variation, biological variation, and measurement error. The latter was estimated at ± 1.2% (% TEM) via repeated V̇O2 measurements at a standardized treadmill walking pace (5.0 km·h−1) on the same day. Our 95% confidence interval for the reproduction of self-paced walking METs on a separate day of ± 9.2% (i.e., 1.96 × % TEM of 4.7) can therefore be partitioned into ± 2.4% of measurement error, with the remaining ± 6.8% attributed to self-pacing differences and biological variation.
Self-paced walking METs and energy expenditure.
The ranges in Table 1 suggest that we recruited a cross-section of the 35- to 45-yr-old Australian population. Our measured moderate-paced walking speed and associated MET intensity could therefore be generalized to middle aged Australians and possibly other Westernized countries.
After being instructed to walk at “what they perceived to be a moderate pace,” women self-paced at a higher mean speed compared to men. Their Borg RPE demonstrated that they did not perceive themselves to be working any harder than the men even though their METs, CSAhip, HR, and percentage of age predicted HRmax were significantly higher (Table 4). Our male and female speed comparison contrasts with the findings of Blessey et al. (5), who reported a significantly lower (P < 0.01) self-paced walking speed in women (4.4 km·h−1) compared with men (5.3 km·h−1) on an oval concrete track. Finley et al. (16) also covertly observed a lower mean walking speed in women (4.5 km·h−1) compared with men (4.8 km·h−1) in shopping centers, small commercial areas, residential areas, and the city.
We expected men to self-select a faster walking pace than the women because, on average, men are aerobically fitter than women. Evidence of this in our sample is provided by gender specific MET–HR regression equations revealing that walking at 4.0 METs would elicit a substantially lower HR in men (96 beats·min−1) compared with women (110 beats·min−1). However, a relationship between aerobic fitness and self-selected exercise intensity may not be valid for low intensity workloads. Cunningham et al. (10) reported no significant difference between aerobically fit (mean V̇O2max = 57 mL·kg−1·min−1) and unfit (mean V̇O2max = 33 mL·kg−1·min−1) groups for self-paced slow walking speed, despite the expected differences (faster speeds for the fitter group) when walking normally, fast, and as fast as possible. Ygnve et al. (35) also found that there is no significant gender difference in V̇O2 (mL·kg−1·min−1) during self-paced normal and fast walking even though men self-paced at a higher jogging V̇O2. Higher self-paced walking METs in women could be an artifact of the Hawthorne effect, or the measured values might reflect a real gender difference in self-paced walking speed and intensity. If the latter is true, women will accrue greater health benefits from a given duration of walking compared with men because they work at a slightly higher MET intensity (4.1 vs 3.8 METs), which results in a substantial increase in cardiovascular stress (61 vs 52% of age predicted HRmax) because they are less aerobically fit.
Energy expenditure of walking one kilometer (kcal·kg−1·km−1).
The energy expended to move 1 kg of body mass 1 km (kcal·kg−1·km−1) over level terrain should remain constant. However, an 11% coefficient of variation for this variable was probably caused by variations in mechanical efficiency, with inefficient walking styles eliciting higher values. This natural variation in mechanical efficiency weakens the walking speed–MET relationship. No gender difference for kilocalories per kilogram per kilometer suggests that a presumed shorter stride length (women were ∼15 cm shorter than men) did not alter the mechanical efficiency of walking at self-selected speeds.
The mean of 0.75 kcal·kg−1·km−1 (Table 4) is useful for estimating the gross energy costs of walking. Hence, assuming constancy of energy intake, walking an extra 1.0 km·d−1 over level terrain would result in an annual caloric deficit equivalent to 2.3 kg of fat (1 kg fat = 9100 kcal) for a 75-kg individual. This minor lifestyle modification (1.0 km takes ∼11 min 5.3 km·h−1), which results in ∼60 kcal·d−1 being expended, has been suggested to offset the gradual weight gain in about 90% of the American population (21).
Summary.
Accelerometers, motion sensors, and pedometers can provide a person with motivation and feedback by roughly estimating overall daily activity. Used in conjunction with HR, they can also contribute to energy expenditure and exercise intensity predictions of other activities. However, when level walking METs and energy expenditure are to be estimated in a research setting, speed will provide more accurate and precise estimates of exercise intensity (METs) and energy cost (kcal·min−1) than hip accelerometry. We base this conclusion on the fact that: 1) speed displayed a stronger association with our measured METs and energy expenditure compared with CSAhip (Table 2); and 2) cross-validation of other equations showed that speed-based predictions were closer to, and within tighter prediction limits, of our measured values compared to CSAhip-based predictions (Table 3). In addition to a more precise estimate of METs and energy expenditure, speed can be measured covertly, which eliminates the Hawthorne effect. Furthermore, average speed is no more difficult to measure than time and distance. Our findings for CSAhip may extend to other accelerometers because of moderate to strong intercorrelations between these instruments (3,24,33), but further research is required to confirm this.
Self-selected moderate-paced walking for a cross-sectional sample of 35- to 45-yr-old Australian men (3.8 METs at 52% HRmax) and women (4.1 METs at 61% HRmax) suggests that women are likely to accrue greater health benefits from a given duration of walking. The mean 0.75 kcal·kg−1·km−1 is a simple conversion to estimate energy expenditure per level kilometer walked that is independent of speed. This assists in setting achievable goals for weight maintenance via regular walking sessions.
REFERENCES
1. Australian Bureau of Statistics. National Health Survey, Summary of Results. Canberra: Australian Bureau of Statistics, Publication 4364.0, 2001, pp. 82–83.
2. Balogun, J. A., D. A. Martin, and M. A. Clendenin. Human energy expenditure during level walking on a treadmill at speeds of 54–130 m·min−1. Int. Disabil. Stud. 11:71–74, 1989.
3. Bassett, D. R., Jr., B. E. Ainsworth, A. M. Swartz, S. J. Strath, W. L. O’Brien. and G. A. King. Validity of four motion sensors in measuring moderate intensity physical activity. Med. Sci. Sports Exerc. 32:S471–S480, 2000.
4. Bland, J. M., and D. Altman. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1:307–310, 1986.
5. Blessey, R. L., H. J. Hislop, R. L. Waters, and D. Antonelli. Metabolic energy cost of unrestrained walking. Phys. Ther. 56:1019–1025, 1976.
6. Brage, S., N. Wedderkopp, P. W. Franks, L. B. Andersen, and K. Froberg. Reexamination of validity and reliability of the CSA monitor in walking and running. Med. Sci. Sports Exerc. 35:1447–1454, 2003.
7. Brooks, A. G., R. T. Withers, C. J. Gore, A. J. Vogler, J. Plummer, and J. Cormack. Measurement and prediction of METs during household activities in 35–45 yr old females. Eur. J. Appl. Physiol. 91:638–648, 2004.
8. Bubb, W. J., D. A. Martin, and E. T. Howley. Predicting oxygen uptake during level walking at speeds of 80-130 m/min. J. Card. Rehabil. 5:462–465, 1985.
9. Bussmann, J. B., H. J. Van Den Berg-Emons, S. M. Angulo, T. Stijnen, and H. J. Stam. Sensitivity and reproducibility of accelerometry and heart rate in physical strain assessment during prosthetic gait. Eur. J. Appl. Physiol. 91:71–78, 2004.
10. Cunningham, D. A., P. A. Rechnitzer, M. E. Pearce, and A. P. Donner. Determinants of self-selected walking pace across ages 19 to 66. J. Gerontol. 37:560–564, 1982.
11. Dobson, A. J. Calculating sample size. Trans. Menzies Found. 7:75–79, 1984.
12. Ekelund, U., P. Tingstrom, K. Kamwendo, et al. The validity of the Computer Science and Applications activity monitor for use in coronary artery disease patients during level walking. Clin. Physiol. Funct. Imaging 22:248–253, 2002.
13. Elia, M., and G. Livesey G. Energy expenditure and fuel selection in biological systems: the theory and practice of calculations based on indirect calorimetry and tracer methods. In: Metabolic Control of Eating, Energy Expenditure and the Bioenergetics of Obesity, A. P. Simopoulos (Ed.). Basel: Karger, 1992, pp. 68–131.
14. Eston, R. G., A. V. Rowlands, and D. K. Ingledew. Validity of heart rate, pedometry, and accelerometry for predicting the energy cost of children’s activities. J. Appl. Physiol. 84:362–371, 1998.
15. Fehling, P. C., D. L. Smith, S. E. Warner, and G. P. Dalsky. Comparison of accelerometers with oxygen consumption in older adults during exercise. Med. Sci. Sports Exerc. 31:171–175, 1999.
16. Finley, F. R., and K. A. Codey. Locomotive characteristics of urban pedestrians. Arch. Phys. Med. Rehabil. 51:423–426, 1970.
17. Freedson, P. S., E. Melanson, and J. Sirard. Calibration of the Computer Science and Applications, Inc. accelerometer. Med. Sci. Sports Exerc. 30:777–781, 1998.
18. Gunn, S. M., A. G. Brooks, R. T. Withers, et al. Determining energy expenditure during some household and garden tasks. Med. Sci. Sports Exerc. 34:895–902, 2002.
19. Gunn, S. M., G. E. van der Ploeg, R. T. Withers, et al. Measurement and prediction of energy expenditure in males during household and garden tasks. Eur. J. Appl. Physiol. 91:61–70, 2004.
20. Hendelman, D., K. Miller, C. Baggett, E. Debold, and P. Freedson. Validity of accelerometry for the assessment of moderate intensity physical activity in the field. Med. Sci. Sports Exerc. 32:S442–S449, 2000.
21. Hill, J. O., H. R. Wyatt, G. W. Reed, and J. C. Peters. Obesity and the environment: where do we go from here? Science 299:853–855, 2003.
22. Leenders, N. Y., T. E. Nelson, and W. M. Sherman. Ability of different physical activity monitors to detect movement during treadmill walking. Int. J. Sports Med. 24:43–50, 2003.
23. McDonald, I. Statistical studies of recorded energy expenditure of man. Part 2: expenditure on walking related to weight, sex, age, height, speed and gradient. Nutr. Abs. Rev. 31:739–762, 1961.
24. Melanson, E. L, Jr., and P. S. Freedson. Validity of the Computer Science and Applications, Inc. (CSA) activity monitor. Med. Sci. Sports Exerc. 27:934–940, 1995.
25. Nichols, J. F., C. G. Morgan, L. E. Chabot, J. F. Sallis, and K. J. Calfas. Assessment of physical activity with the Computer Science and Applications, Inc., accelerometer: laboratory versus field validation. Res. Q. Exerc. Sport 71:36–43, 2000.
26. Passmore, R., and J. V. G. A. Durnin. Human energy expenditure. Physiol. Rev. 35:801–835, 1955.
27. Pearce, M. E., D. A. Cunningham, A. P. Donner, P. A. Rechnitzer, G. M. Fullerton, and J. H. Howard. Energy cost of treadmill and floor walking at self-selected paces. Eur. J. Appl. Physiol. Occup. Physiol. 52:115–119, 1983.
28. Swartz, A. M., S. J. Strath, D. R. Bassett, Jr., W. L. O’Brien, G. A. King, B. E. Ainsworth. Estimation of energy expenditure using CSA accelerometers at hip and wrist sites. Med. Sci. Sports Exerc. 32:S450–S456, 2000.
29. Tanaka, H., K. D. Monahan, and D. R. Seals. Age-predicted maximal heart rate revisited. J. Am. Coll. Cardiol. 37:153–156, 2001.
30. Trost, S. G., D. S. Ward, S. M. Moorehead, P. D. Watson, W. Riner, and J. R. Burke. Validity of the computer science and applications (CSA) activity monitor in children. Med. Sci. Sports Exerc. 30:629–633, 1998.
31. U.S. Department of Health and Human Services. Physical Activity and Health: A Report of the Surgeon General. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, 1996, pp. 185.
32. van der Walt, W. H., and C. H. Wyndham. An equation for prediction of energy expenditure of walking and running. J. Appl. Physiol. 34:559–563, 1973.
33. Welk, G. J., S. N. Blair, K. Wood, S. Jones, and R. W. Thompson. A comparative evaluation of three accelerometry-based physical activity monitors. Med. Sci. Sports Exerc. 32:S489–S497, 2000.
34. Wergel-Kolmert, U., and B. Wohlfart. Day-to-day variation in oxygen consumption and energy expenditure during submaximal treadmill walking in female adolescents. Clin. Physiol. 19:161–168, 1999.
35. Yngve, A., A. M. Sjostrom, and U. Ekelund. Effect of monitor placement and of activity setting on the MTI accelerometer output. Med. Sci. Sports Exerc. 35:320–326, 2003.
Keywords:
CSA; MTI; VELOCITY; SELF-PACED
©2005The American College of Sports Medicine