The accuracy of accelerometry to estimate energy expenditure has been evaluated during daily life by indirect calorimetry (5,7) and during walking and jogging by the oxygen consumption rate (V̇O2) (4,7,10,12,14). Although these previous results support the suitability of the method of calorimetry in the field, there have been few studies to apply this method to estimate walking efficiency.
The vertical displacement during walking as measured by triaxial accelerometry is likely associated with decreased walking efficiency. Eston et al. (7) measured three-dimensional accelerations in 9-yr-old children with a triaxial accelerometer placed on their back waist during graded walking and running on a treadmill at 4–10 km·h−1, suggesting that the ratio of vertical to total acceleration at a given walking velocity was twofold higher in children than that reported in adults elsewhere (4), because the higher energy expenditure at a given walking velocity of 4–5 km·h−1 reported in children 9 yr of age compared with adults may be partially explained by the energy waste in the vertical direction (14). Bouten et al. (4) measured three-dimensional accelerations in adults during walking at a graded velocity of 3–7 km·h−1 on a treadmill and reported that the sum of accelerations in all directions increased with walking velocity, which was highly correlated with V̇O2. They also reported that the increasing rate of acceleration with walking velocity was highest in the vertical direction, but they did not suggest the involvement in walking efficiency. Because the mechanical energy in the vertical direction does not drive the body forth, the increase in acceleration rate in the vertical direction suggests a decrease in walking efficiency.
It has been suggested that thigh-muscle strength is closely associated with maximal walking velocity in older subjects (8,16,17). Moreover, the vertical displacement of energy expenditure was reportedly improved by preventing decline in thigh-muscle strength for older subjects (13). So far, these biomechanical parameters have been investigated using video and/or force plate systems in laboratories (6,11,15), which are expensive and not readily available in the field. The purposes of the present study was to therefore assess the ability of triaxial accelerometry to accurately represent baseline walking efficiency and to detect improvements of walking efficiency after an exercise training program.
The subjects were recruited from 130 participants in the health promotion program by Matsumoto City for 9 months before the present study in which they were encouraged to perform the targeted walk of 10,000 steps·d−1 every day. The averaged walking steps and days that the participants accomplished in the program was ∼6000 steps·d−1 and 13 d·month−1. Before participating in the present study, subjects were confirmed to be apparently healthy and not to have any prominent cardiovascular or orthostatic diseases determined by a medical examination including a blood chemical test. They did not participate in any other regular physical training program except for the present study. This study was approved by the Review Board on Human Experiments at the Shinshu University School of Medicine. After the experimental protocols were fully explained, all subjects gave their written informed consent before participating in this study.
We verified the accuracy to estimate energy expenditure, V̇O2, during walking from the accelerations in 13 male and 27 female subjects. After measuring body weight and height, isometric extension torque (Fmax) was determined in each subject. The physical characteristics of the subjects were; 67 ± 2 kg in body weight, 166 ± 2 cm in height, and 228 ± 18 N·m in Fmax for male subjects, and 55 ± 2 kg in body weight, 154 ± 1 cm in height, and 140 ± 6 N·m in Fmax for female subjects, not significantly different from those in the second experiment in Table 2.
We constructed an instrument to measure triaxial accelerations from three uniaxial semiconductor acceleration sensors (ADXL202EB, Analog Devices, Wilmington, MA). The characteristics of the device were 9 × 6 × 4 mm in size, ± 3 V in power supply, 0–500 Hz in frequency response, and ±2 g in detective range. The acceleration sensors, a band-pass filter (0.3–100 Hz), and a battery were mounted in an electrical circuit and were placed in a plastic box of 7.0 × 5.0 × 1.5 cm in size with 40-kilobyte memory. The box was equipped with an RS232C port to transport the accelerations in the x, y, and z directions to an external computer (Active Tracer 301, GMS, Tokyo). The total weight of the device was 65 g. The appearance and a picture of a subject fitted with the device are illustrated in Figure 1.
After baseline measurements at rest, the subjects with the device on their back-waist performed a 3-min walk on a flat floor at four graded subjective velocities; slow, moderate, fast, and fastest, with intermittent 30-s rest each, while the three-dimensional accelerations and V̇O2 (a portable gas analyzer, Meta Max 3B, Cortex, Germany) were simultaneously measured at 10-ms intervals and recorded with memories at 1 min as averaged values. At first, subjects were asked to walk at moderate velocity at which they performed a walk every day and then asked to walk at their maximal velocity. The slow velocity was instructed as 50% of the moderate velocity, and the fast velocity was intermediate between the moderate and the fastest velocities. The subjects were supervised by trainers to maintain the given velocity during the measurements. After the measurements, all the data were transferred to a host computer for analyses.
The total impulse (Itotal, N·min−1) was calculated from the accelerations as reported previously (4). Briefly, the absolute value of total acceleration was calculated as the square root of the summed vectors of each direction and then Itotal was calculated by integrating the absolute acceleration at a given period.
The Itotal, V̇O2 for the final 1 min in each graded walk for 3 min, was adopted for regression analyses (Fig. 2). The V̇O2 was measured with a portable gas analyzer (Meta MAX 3B, Cortex) at 10-ms intervals and recorded with a computer by telemetry. The analyzer was calibrated before starting the measurements in each subject. To confirm the precision to determine V̇O2 with the analyzer, we simultaneously measured V̇O2 in three subjects with another gas analyzer (AE230, Minato, Tokyo) during graded exercise on a cycle ergometer in an artificial climate chamber maintained at 25°C of atmospheric temperature and 50% of relative humidity. The V̇O2 determined with both analyzers were highly correlated (r = 0.989) with a regression equation of y = 1.069 x − 0.059. The 95% of confidence limit to determine V̇O2 with the portable gas analyzer was ± 90 mL·min−1 over the range of 200–2500 mL·min−1.
We assessed the relations of thigh-muscle strength, maximal walking velocity, and three-dimensional accelerations before and after training regimen; endurance/resistance was 6–9 months in older subjects. The physical characteristics of the subjects in the second experiment are shown in Table 1. They remained unchanged after training. Fmax and maximal walking velocity (Vmax) were measured before and after the 6 and 9 months of training. Fmax was measured in each side of the knee with a dynamometer (Biodex, Biodex Medical System, NY). After regular warming up and familiarization protocols, the anatomical axis of the knee joint was aligned with the mechanical axis of the dynamometer arm to adjust the angle between the tibia and femur to 105°. Then, three 3-s maximum voluntary contractions, with 10-s intermittent recovery, were conducted. The peak torque averaged for two trials was adopted for the value for one knee side and represented as an averaged value of both knee sides. Vmax was determined by measuring the faster time of two trials to walk a distance of 50 m on a flat ground. During ambulation, the accelerations were measured with a triaxial accelerometer (Active Tracer 301, GMS).
The subjects trained once per week under our supervision for 9 months, between June 1, 2001, and February 28, 2002, according to the protocol recommended by ACSM (1). Before exercise training program, maximal work intensity (Wmax) was determined from the relation between exercise intensity and HR during graded exercise to submaximal intensity. Moreover, 1RM was determined as maximal isometric knee extension and flexion force as described above. The training program was shown in Table 2. Briefly, after 6-min self-paced walking and 5-min stretch exercise for warming up, subjects performed against gravity exercise at two sets of 10 repetitions every 5 s. Then, they performed two 15-min bicycle ergometer sets (Combi, Aerobike 800, Tokyo) at 40–60% Wmax. In addition, they performed leg extensions and leg curls at 40–60% of 1RM, two sets of 10 repetitions for resistance training. The exercise intensity for endurance training was increased with the training days: 40%, 50%, and 60% of the pretraining Wmax, in the first, second, and third month of training, respectively. Similarly, the exercise intensity for resistance training was also increased with the training days: 40%, 50%, and 60% of the pretraining 1RM in the first, second, and third month of training, respectively. After the fourth month, the exercise intensities for endurance and resistance training were performed at 60% of Wmax and 60% of 1RM, respectively, determined at the beginning of every month. The environmental condition for the training room was controlled at ∼20°C atmospheric temperature and ∼50% relative humidity.
The effects of training on physical characteristics (Table 1), Fmax, Vmax, Ix/Itotal, Iy/Itotal, and Iz/Itotal (Table 3) were tested by a 2 (male, female) × 3 (before, 6 months, and 9 months) ANOVA for repeated measures. Subsequent post hoc tests to determine significant differences in the various pair wise comparisons were performed with Fisher’s least significant difference test. Regression analyses were performed to test the differences in regression coefficients between male and female subjects were examined using the t-test (19). Values are presented as means ± SE.
Figure 2 shows the relations between V̇O2 and total impulse (Itotal) during graded walking in the male and female subjects. V̇O2 was highly correlated with Itotal in male subjects (r = 0.949, P < 0.0001) with a regression equation of y = 0.053 x + 288 and in female subjects (r = 0.964, P < 0.0001) with y = 0.060 x + 218. Because there was no significant difference in the regression coefficients between the groups, a regression equation was determined on the data pooled from both groups as y = 0.056 x + 249 (r = 0.958, P < 0.0001). The accuracy to estimate V̇O2 by substituting Itotal in the equation was 20–83 mL·min−1 with a 95% confidence limit over the range of 250–2200 mL·min−1 of V̇O2.
As shown in Table 3, Fmax and Vmax increased significantly after 6 and 9 months of training in the male and female groups. In addition, Ix/Itotal and Iy/Itotal increased significantly after 6 and 9 months in the male and female groups except for Iy/Itotal after 6 months in the female group. On the other hand, Iz/Itotal decreased significantly after 6 and 9 months of training in both groups. The percent changes from the baselines for the male groups were +36 ± 7% in Fmax, +27 ± 4% in Vmax, +24 ± 7% in Ix/Itotal, +16 ± 5% in Iy/Itotal, and −19 ± 5% in Iz/Itotal, and for the female group, +52 ± 9% in Fmax, +19 ± 2% in Vmax, +17 ± 2% in Ix/Itotal, and +8 ± 2% in Iy/Itotal and −13 ± 2% in Iz/Itotal after 9 months of training. Because there were no significant differences in the percent change in these variables between the male and female groups, they were pooled as a total group and were +48 ± 7% in Fmax, +21 ± 2% in Vmax, +18 ± 2% in Ix/Itotal, +10 ± 2% in Iy/Itotal, and −14 ± 2% in Iz/Itotal.
Figure 3 shows the relations between Vmax versus Ix/Itotal, Iy/Itotal, or Iz/Itotal at Vmax in the male (A) and female (B) groups. The regression equations in each figure were determined from the pooled values from before and after the 6 months and 9 months of training. As shown in the figures, the Ix/Itotal was positively correlated with Vmax in the male group (r = 0.677, P < 0.0001) with a regression equation of y = 0.0016 x + 0.07 and in the female group (r = 0.505, P < 0.0001) with y = 0.0025 x + 0.04. The Iy/Itotal was slightly but significantly correlated with Vmax in the male group (r = 0.483, P < 0.01) with a regression equation of y = 0.0012 x + 0.05 and also in the female group (r = 0.234, P < 0.05) with y = 0.0021 x + 0.03. The Iz/Itotal was negatively correlated with Vmax in the male group (r = −0.676, P < 0.0001) with a regression equation of y = −0.0025 x + 0.82 and in the female group (r = −0.509, P < 0.0001) with y = −0.0034 x + 0.91.
Figure 4 shows the relations between Vmax versus Ix/Itotal, Iy/Itotal, or Iz/Itotal at Vmax on the data pooled from the male and female groups since there were no significant differences in the regression coefficients between the groups in Figure 3. In the total group, Ix/Itotal was positively correlated with Vmax (r = 0.561, P < 0.0001) with a regression equation of y = 0.0021 x + 0.014. Iy/Itotal was positively correlated with Vmax (r = 0.212, P < 0.01) with a regression equation of y = 0.0018 x + 0.002. The Iz/Itotal was negatively correlated with Vmax (r = - 0.522, P < 0.0001) with a regression equation of y = −0.0030 x + 0.85.
The small panels attached in each of Figures 3 and 4 were means and SE before and after 6 months and 9 months of exercise training. The changes in Ix/Itotal, Iy/Itotal, and Iz/Itotal at a given increase in Vmax were almost identical to the slope of the regression equations determined in the pooled values.
Figure 5 shows the relations between Fmax and the product of Vmax and body weight in the male (A), female (B), and total (C) groups. As shown in the figures, the product of Vmax and body weight increased with Fmax in the male group (r = 0.662, P < 0.0001) with a regression equation of y = 36.2 x + 2635 and in the female group (r = 0.419, P < 0.0001) with y = 32.3 x + 3106. Because there were no significant differences between regression coefficients between the male and female groups, the data were pooled in the total group, where the product of Vmax and body weight was significantly correlated with Fmax (r = 0.633, P < 0.0001) with a regression equation of y = 34.3 x + 2883. The small panels attached in each figure were means and SE before and after 6 months and 9 months of exercise training. The increase in Vmax × body weight at a given increase in Fmax was almost identical to the slope of the regression equations determined in the pooled values.
In the present study, we verified the suitability of triaxial accelerometry to estimate V̇O2 during walking in older male and female subjects. In addition, we found by this method that the energy loss in the vertical direction was reduced, whereas the energy utilization in the anterior-posterior/mediolateral directions was increased in male and female older subjects as their maximal walking velocity and thigh muscle strength increased after 6 or 9 months endurance/resistance training.
In the present study, we calculated Itotal according to the formula of Bouten et al. (4). They reported that energy expenditure during walking on a treadmill at 3–7 km·h−1, equivalent to V̇O2 of 5–20 mL·kg−1·min−1, could be estimated with an accuracy of about 15% in 11 younger male subjects. More recently, Nichols et al. (14) reported a similar accuracy in this method using 30 young male and 30 young female subjects during treadmill walking and running at 3–10 km·h−1. In the present study, the accuracy was about 10% during walking on flat ground at 2–8 km·h−1 in 53 older male and female subjects. These results suggest that the energy expenditure estimated from Itotal in the present study was reliable enough to discuss walking efficiency in older subjects.
As shown in Table 3 and Figures 3 and 4, the ratios of Ix/Itotal and Iy/Itotal in the male and female groups increased whereas Iz/Itotal decreased, which were significantly correlated with the increase in Vmax. Because Itotal was closely correlated with V̇O2 or total energy expenditure over the range of walking velocity (Fig. 2), the decreased Iz/Itotal and the increased Ix/Itotal and Iy/Itotal suggest an improved walking efficiency in subjects with higher Vmax.
Eston et al. (7) measured, by triaxial accelerometry, the impulses in three directions during walking at 4 and 6 km·h−1 and running at 8 and 10 km·h−1 in children 9 yr old. These authors suggested that the ratio of Iz/Itotal was 0.8, 60% higher than that reported in younger adults (4). Because energy expenditure at a given walking velocity of 8–13 km·h−1 was reported about 10% higher in children 9 yr old than those above 17 yr old (2), the low walking efficiency in children may be partially explained by the energy waste in the vertical direction. Although there has been no study to compare the walking efficiency between younger and older subjects by triaxial accelerometry, Martin et al. (13) reported that aerobic demand during walking at 2–7 km·h−1 was 8% higher in older subjects than in younger subjects, suggesting the higher energy waste in the vertical direction for older subjects. The results in the present study suggest that the walking efficiency was improved in male and female older subjects with higher Vmax by decreasing energy waste in the vertical direction whereas increasing energy utilization in the anterior-posterior/mediolateral directions.
The reason for the improved walking efficiency in subjects with higher Vmax is not clear. However, as shown in Figure 5, because the product of Vmax and body weight was closely correlated with Fmax, the increased Fmax may be involved in the mechanisms. Gleim et al. (9) studied the influence of the body flexibility on the efficiency of walking and jogging, and suggested that the flexibility in trunk rotation and lower limb turnout decreased with increased leg muscle strength, which increased the efficiency of walking and jogging over the range of 0.3–12 km·h−1. As for the mechanism, the authors hypothesized that a thigh muscle, once stretched, should be able to generate more tension from the elastic and less from the contractile component, resulting in less joint movement in the vertical direction. They also suggested that a rigid musculoskeletal system is more resistant to the rotatory forces imparted by increased speed of movement, suggesting decreased motion in the vertical direction whereas increased motion in anterior-posterior/mediolateral directions. The results of the present study suggest the reduced Iz/Itotal was the result of increased knee extensor strength.
As a limitation of the present study, we measured only knee extensor and flexion forces before and after resistance/endurance training. Resistance training for 5–6 months increased not only thigh-muscle strength but also hip- and lower leg-muscle strength (3,18). Thus, we did not exclude the involvement of these muscles in the improvement of walking efficiency after training. Another limitation of the present study was that the subjects trained only once a week, less frequently than the protocol recommended by ACSM, due to limited staff resources and institutional capacity. They were not on a home training program except for walking at the targeted steps of 10,000 per day. Despite the milder training program, thigh-muscle strength increased with improved walking efficiency, which was detected by triaxial accelerometry.
Regarding clinical implication of these findings, the device will be used for trainers to check the effects of exercise training program on patients/trainees at home that they design. If the data are transferred to a host computer in a health administration center via the computer network, trainers will know the up-to-date improvement of walking efficiency in patients/trainees. The data on individuals will be used for trainers to design the optimal exercise training programs for patients/trainees.
In summary, as detected by triaxial accelerometry, Iz/Itotal decreased, whereas Ix/Itotal and Iy/Itotal increased at a higher Vmax in older subjects postendurance/resistance training.
We thank the volunteers from Matsumoto Health Promotion Program for the Elderly, for participating in this study.
This study was supported in part by grants from the Ministry of Health, Labor, and Welfare and the Ministry of Education, Culture, Sports, Science, and Technology of Japan.
1. American College of Sports Medicine Position Stand. Exercise and physical activity for older adults. Med. Sci. Sports Exerc. 30: 992–1008, 1998.
2. Bar-Or, O. Pediatric Sports Medicine for the Pediatrician: From Physiologic Principles to Clinical Applications. New York: Springer-Verlag, 1983, pp. 6–9.
3. Bemben, D. A., N. L. Fetters, M. G. Bemben, N. Nabavi, and E. T. Koh. Musculoskeletal responses to high- and low-intensity resistance training
in early postmenopausal women. Med. Sci. Sports Exerc. 32: 1949–1957, 2000.
4. Bouten, C. V., K. R. Westerterp, M. Verduin, and J. D. Janssen. Assessment of energy expenditure for physical activity using a triaxial accelerometer. Med. Sci. Sports Exerc. 26: 1516–1523, 1994.
5. Chen, K. Y., and M. Sun. Improving energy expenditure estimation by using a triaxial accelerometer. J. Appl. Physiol. 83: 2112–2122, 1997.
6. Elble, R. J., S. S. Thomas, C. Higgins, and J. Colliver. Stride-dependent changes in gait of older people. J. Neurol. 238: 1–5, 1991.
7. 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.
8. Fiatarone, M. A., E. F. O’Neill, N. D. Ryan, et al. Exercise training and nutritional supplementation for physical frailty in very elderly people. N. Engl. J. Med. 330: 1769–1775, 1994.
9. Gleim, G. W., N. S. Stachenfeld, and J. A. Nicholas. The influence of flexibility on the economy of walking and jogging. J. Orthop. Res. 8: 814–823, 1990.
10. Jakicic, J. M., C. Winters, K. Lagally, J. Ho, R. J. Robertson, and R. R. Wing. The accuracy of the TriTrac-R3D accelerometer to estimate energy expenditure. Med. Sci. Sports Exerc. 31: 747–754, 1999.
11. K errigan, D. C., M. K. T odd, U. Della Croce, L. A. Lipsitz, and J. J. Collins. Biomechanical gait alterations independent of speed in the healthy elderly: evidence for specific limiting impairments. Arch. Phys. Med. Rehabil.
12. Levine, J. A., P. A. Baukol, and K. R. Westerterp. Validation of the Tracmor triaxial accelerometer system for walking. Med. Sci. Sports Exerc. 33: 1593–1597, 2001.
13. Martin, P. E., D. E. Rothstein, and D. D. Larish. Effects of age and physical activity status on the speed-aerobic demand relationship of walking. J. Appl. Physiol. 73: 200–206, 1992.
14. Nichols, J. F., C. G. Morgan, J. A. Sarkin, J. F. Sallis, and K. J. Calfas. Validity, reliability, and calibration of the Tritrac accelerometer as a measure of physical activity. Med. Sci. Sports Exerc. 31: 908–912, 1999.
15. O strosky, K. M., J. M. V answearingen, R. G. B urdett, and Z. G ee. A comparison of gait characteristics in young and old subjects. Phys. Ther.
74:637–644; discussion 644–636, 1994.
16. Schlicht, J., D. N. Camaione, and S. V. Owen. Effect of intense strength training on standing balance, walking speed, and sit-to-stand performance in older adults. J. Gerontol. A Biol. Sci. Med. Sci. 56: M281–M286, 2001.
17. Sipila, S., J. Multanen, M. Kallinen, P. Era, and H. Suominen. Effects of strength and endurance training
on isometric muscle strength and walking speed in elderly women. Acta Physiol. Scand. 156: 457–464, 1996.
18. Sipila, S., and H. Suominen. Effects of strength and endurance training
on thigh and leg muscle mass and composition in elderly women. J. Appl. Physiol. 78: 334–340, 1995.
19. Sokal, R. R., and F. J. Rohlf. Biometry, 3rd Ed. New York: Freeman, 1995, pp. 469–477.