People who exercise are interested in knowing the number of calories they expend during exercise. However, physical activity logs or questionnaires have not always been able to provide sufficient accuracy for many needs ⁽¹⁰⁾. Accelerometers provide a convenient tool for assessing physical activity outside of the laboratory setting: they are relatively simple to use and allow for little subject influence. The Caltrac and Tritrac (Hemokinetics, Madison, WI) are two commercially available accelerometers that are sensitive to the body's acceleration/deceleration. The Caltrac senses vertical acceleration and converts this measurement into a caloric expenditure, whereas the Tritrac senses acceleration in three separate directions: vertical, lateral-medial, and anterior-posterior. It is important to know whether these commercially available instruments are giving people accurate information.

The accelerometers have been tested in various laboratory settings to determine their validity and reliability. Limited research has been conducted on a triaxial accelerometer (early version of the Tritrac accelerometer used in this study) and this research has been limited to young adults ^(5,19). Several studies have investigated the reliability of the Caltrac accelerometer ^(12,17,18). Investigators have tested the reliability of the Caltrac using a variety of exercises; walking (normal and faster speeds), running, and occupational physical activities (postal carriers). The results from the studies using walking and running activities have concluded that the Caltrac is highly reliable, with reliability coefficients ranging from r = 0.7 ⁽¹⁷⁾ to r = 0.98 ⁽¹²⁾. The results of the day-to-day reliability during occupational activity also demonstrated that the Caltrac was moderately reliable, r = 0.58 ⁽¹⁸⁾, although exhibiting a slightly lower strength of relationship. The validity of the Caltrac has been assessed in children ^(6,15,19) and young adults ^{(2-4,8,9,13,14,17,20)}. Several studies conducted on children during free living conditions compared the Caltrac measures to heart rate (HR) response ^(15,19). Welk and Corbin ⁽¹⁹⁾ report a moderate correlation with HR (r = 0.52), while Simons-Morton et al. ⁽¹⁵⁾ reported higher correlations than Welk and Corbin ⁽¹⁹⁾ between Caltrac measures and HR. Simons-Morton et al. ⁽¹⁵⁾ concluded that strength of the correlations were dependent upon age of the children and intensity of the activity (r = 0.55 to r = 0.80). Studies using young adult populations to validate the Caltrac using indirect calorimetry as the gold standard report that the Caltrac typically overestimates EE ^(3,8,13). Balogun et al. ⁽²⁾ reported an overestimation ranging from 13.3 to 52.9% during various walking speeds, while Pambianco et al. ⁽¹³⁾ reported an overestimation of 9-13% during walking activity. Haymes and Byrnes ⁽⁸⁾ tested subjects during different walking and running speeds. It was reported that the Caltrac overestimated EE at the higher walking speeds (∼3.6 kcal·min⁻¹) and at the lower running speeds (∼2.6 kcal·min⁻¹). Although numerous studies have examined the validity of accelerometers on children and young adults, there have been very few completed on older adults.

The purpose of this study, therefore, was to compare the Caltrac and Tritrac accelerometer with indirect calorimetry to assess caloric expenditure in older adults during treadmill walking and bench stepping exercise performed at various work rates.

METHODS

Subjects. The volunteers in this study were participants in the Sites Testing for Osteoporosis Prevention/Intervention and Treatment trial (STOP/IT) at the University of Connecticut Health Center, Osteoporosis Center Exercise Research Laboratory. Descriptive characteristics of the volunteers (N = 86) are presented in Table 1. All volunteers were healthy males and females who had been randomized into an exercise or control group in a 2-yr exercise study. Before testing all volunteers gave written informed consent indicating that they understood the risks and benefits of the study, as well as the testing procedures, and that their participation was voluntary. As part of the STOP/IT trial testing protocol all volunteers regardless of group assignment (exercise or control) underwent semiannual testing which included a submaximal treadmill test with metabolic measurements. Only those volunteers randomized to the exercise group participated in an additional stepping exercise bout with metabolic measurements. Therefore, the number of volunteers per testing group varied depending on the mode of exercise test. Additionally, the number of volunteers per work rate was determined by the ability of the subject to complete each testing level. Only those subjects who completed all three work rates were used in the final analysis.

Protocol. Before any submaximal testing, all volunteers performed a treadmill incremental maximal aerobic capacity (V˙O_2max) test. The protocol for the V˙O_2max test consisted of a 5-min warm-up on the treadmill. During this warm-up, the treadmill speed was increased to a speed that elicited 70% of the maximal HR achieved during a screening graded exercise, and this speed was maintained throughout the test. The speed used varied among volunteers, but the speed in each case constituted a brisk walking speed. HR and blood pressure (BP) were monitored during each stage. The test started at 0% grade and was increased by 2% every 2 min until the respiratory exchange ratio (RER) reached 1.0. At that point grade was increased 1% each minute until volitional fatigue was reached. Two of the following three criteria were used to determine whether a maximal effort was given: 1) an increase in absolute O₂ uptake of less than or equal to 100 mL·min⁻¹, 2) RER greater than or equal to 1.15, and 3) HR greater than or equal to maximal HR from the subject's initial graded exercise screening test.

On separate days volunteers performed three submaximal, discontinuous (5 min of exercise, 2 min of rest), progressive levels of treadmill walking and bench stepping (Table 2). The work rates were generally performed from lowest to highest for each exercise mode. During the first work rate of the treadmill exercise volunteers walked at a constant speed set to elicit a response of approximately 60% of maximal aerobic capacity (V˙O_2max). Treadmill grade was increased to elicit 70% and 80% of V˙O_2max for the second and third work rates, respectively. The stepping exercise bouts consisted of three 5-min discontinuous, progressively increasing work rates, during which the volunteers performed a stepping sequence at 24 steps·min⁻¹ on a 15.2-, 20.3-, and a 25.4-cm bench. A wooden aerobic step bench (Step-A-Head, New York, NY) with a slip resistant rubberized top was used for the stepping exercise. All volunteers were instructed to perform a four step-cycle: step up, step up, step down, and step down. Volunteers alternated lead legs between cycles and were asked not to hyperextend their knees. Caltrac data were recorded during a 2-min rest between exercise bouts.

Instrumentation. Two commercially available accelerometers, Caltrac and Tritrac (Hemokinetics), were worn on a belt attached to the waist of each subject. The Tritrac was worn on the left hip and the Caltrac was worn on the right hip. Each instrument was placed on the body in line with the anterior axillary line and just anterior to the top of the iliac crest. Each of the instruments was worn according to manufacturer's suggestions. The Caltrac was placed in the "normal" mode, and the subject's age, height, weight, and gender were programmed into each unit. Data were recorded and the Caltrac cleared after each exercise bout. The Tritrac was initialized, via a computer interface, before testing. Height, weight, age, and gender were entered into the computer program, and all subsequent data were downloaded at the completion of the three exercise bouts. All EE data are reported on a calorie (kcal) per minute basis. Metabolic gases were collected throughout the 5 min of activity, but only the last 3 min (an average of the last six 30 s measures) of treadmill walking and the last 2 min (an average of the last four 30 s measures) of stepping exercise were used for calculating EE. Metabolic expired gases were analyzed with Ametek analyzers (Pittsburgh, PA; R-1 Flow Control meters, carbon dioxide sensor P-61B, oxygen sensor N-22 M, carbon dioxide analyzer CD-3A, oxygen analyzer S-3A1) using Rayfield software (Waitsfield, VT; version REP-401). The nonprotein respiratory exchange ratio (RER) was used to determine the kilocalories per liter of oxygen (O₂) consumed. This figure was used in subsequent calculations to determine calorie expenditure using the following equation: oxygen (O₂) consumed (L) multiplied by kilocalories produced per liter of O₂ = metabolic kilocalorie expenditure ⁽¹⁾.

Statistics. A 3 × 3 ANOVA with repeated measures was used to test for differences in EE as measured by indirect calorimetry and estimated by the Caltrac and Tritrac accelerometers at the various exercise intensities. Only those subjects that completed all three work rates were used in the repeated measures analysis for each mode (treadmill exercise, N = 63; stepping exercise, N = 23). If a significant main effect was found an ANOVA was used to determine differences within instrument and within intensity. A Tukey-b post-hoc test was used determine the location of significant differences. All statistical analyses were performed using SPSS for Windows (version 7.5.1, 1996) with the level of probability set at P < 0.05.

RESULTS

Descriptive data for the entire subject pool are presented in Table 1. EE data for the treadmill walking and stepping exercise are presented in Figures 1 and 2, respectively.

Treadmill walking. A repeated measures ANOVA for the treadmill exercise (Fig. 1) revealed a significant main effect for instrument and intensity and a significant interaction (P < 0.05). The Caltrac estimate of EE was significantly greater than measured EE at each intensity, with a percent difference that ranged from 10% at work rate 3 to 52% difference at work rate 1. The Tritrac estimate of EE was significantly lower than the measured EE at each work rate, with percent differences that ranged from −12% at work rate 1 to −37% at work rate 3. When separate ANOVAs were performed within each instrument, there was a significant difference in EE as measured by indirect calorimetry with each subsequent work rate requiring greater EE than the preceding work rate. However, no significant differences in estimates of EE at the different work rates were found for either the Caltrac or Tritrac accelerometers.

Stepping exercise. A repeated measures ANOVA for the stepping exercise (Fig. 2) also revealed a significant main effect for instrument and intensity and a significant interaction (P < 0.05). The Caltrac significantly underestimated EE at all three work rates, with a percent difference that ranged from −19% at work rate 1 to a −28% difference at work rates 2 and 3. The Tritrac significantly underestimated EE at all three work rates with a percent difference of −58 to −60%. The ANOVA performed within each instrument revealed a significant difference in EE as measured by indirect calorimetry with each subsequent work rate requiring greater EE than the preceding work rate. For both accelerometers, estimates of EE at the second and third work rates were significantly greater than EE at the first work rate, but not different between the second and third work rates. However, the Tritrac counts (X counts (lateral-medial), Z counts (vertical), and vector magnitude) were significantly different at each subsequent work rate (Table 3).

DISCUSSION

The results of this study indicate that the Caltrac and Tritrac accelerometers do not provide accurate estimates of EE during treadmill walking and stepping exercise in older adults. The Caltrac accelerometer is not sensitive to change in oxygen consumption with increasing changes in treadmill grade but does have some ability to sense changes in oxygen consumption resulting from changes in stepping height. The Caltrac significantly overestimated EE for treadmill exercise and significantly underestimated EE for stepping exercise. The Tritrac accelerometer always underestimated EE for both modes of exercise (treadmill walking or stepping) but appeared to be sensitive to some changes in work rate during stepping exercise. Interestingly, two of the individual activity counts (X and Z) and the vector magnitude were sensitive to changes in work during the stepping exercise.

The inability of the Caltrac accelerometer to detect changes in treadmill walking grade is consistent with the findings of other researchers who investigated changes in walking speed ^(8,9,11). The present study is in agreement with a study by Montoye et al. ⁽¹¹⁾ that used a Caltrac prototype which provided only activity counts. The subjects in the Montoye et al. ⁽¹¹⁾ study exercised for 4 min while performing a variety of exercises. During the walking exercise, which was performed at three different grades (speed held constant), the accelerometer prototype failed to detect differences in activity counts with changes in treadmill grade.

Pambianco et al. ⁽¹³⁾ tested young adults of normal weight and who were overweight during 15 min of treadmill walking at several walking speeds and reported that the Caltrac was sensitive to changes in the readings caused by changes in walking speeds. The present study kept treadmill speeds constant and increased grade. However, the difference in measured EE and EE estimated from the Caltrac was the smallest (10%) at the highest work rate. This work rate was a vigorous relative work rate (82% of V˙O_2max). In fact, when the results are analyzed by gender (not presented here) the only nonsignificant difference between measured EE and Caltrac estimated EE occurred in the females during treadmill walking at 80% of V˙O_2max. Welk and Corbin ⁽¹⁹⁾ also report that in children the highest association between Caltrac measures and HR occurred during times of free play or more vigorous physical activity.

The Tritrac accelerometer has the ability to detect movement in three planes: sagittal, transverse, and frontal, but it does not appear that the multiplane configuration enhances the ability to accurately detect changes in EE during treadmill walking, at least not as the Tritrac is currently designed. The question remains: Why are these accelerometers not able to detect differences in work rates during graded treadmill walking?

Montoye et al. ⁽¹¹⁾ stated that the inability of the accelerometer to detect changes in EE during increases in grade walking may not be a serious problem. They stated that if these instruments are used to detect changes in "habitual physical activity" then not sensing changes in grade (walking on uneven terrain) would not be seen as a major limiting factor of the instrument. A more serious flaw would be an inability to sense changes in walking speeds.

Haymes and Brynes ⁽⁸⁾ reported on the ability of the Caltrac to sense changes in running and walking speeds. The subjects walked at speeds of 2-5 mph and ran at speeds of 4-8 mph (horizontal treadmill). The authors report that the Caltrac could not detect changes among running speeds but appeared to be more sensitive during the walking exercise. While the subjects were walking at 2 mph, there were no significant differences between the Caltrac estimation of EE and EE as measured by indirect calorimetry. Walking speeds greater than 2 mph the Caltrac significantly overestimated EE. The present study did not change speeds during either mode of exercise. A suggestion for future research would be to compare the EE from the accelerometers during different speeds of stepping exercise. It appears that during stepping exercise the accelerometers are sensitive to changes in work rates resulting from different step heights, and this may provide information necessary for the accurate estimation of EE during step aerobic classes.

The results of the stepping exercise revealed a slightly different response for both accelerometers. Both the Caltrac and Tritrac provided an increase in the estimate of caloric expenditure at the two higher step heights when compared with the lowest height. To our knowledge the only other study that has used stepping was the study published by Montoye et al. ⁽¹¹⁾. The subjects in this study stepped at two work rates, 20 and 35 steps·min⁻¹. It was reported that there was an increase in the accelerometers reading as the work rates increased. This Caltrac prototype did not provide a caloric estimate of energy expended. Therefore, a comparison of reported calories with calories expended based on indirect calorimetry cannot be made. It appears that both the Tritrac and Caltrac accelerometers have some ability to distinguish increases in work rates during higher levels of stepping exercise. But even with this sensitivity, both accelerometers significantly underestimate EE at each stepping work rate.

The Tritrac activity counts in the lateral-medial (X counts) and vertical (Z counts) directions and the vector magnitude were able to detect changes in movement among the three stepping work rates. As expected, since stepping rate was constant across work rates, the Y counts, which detect movement in the anterior-posterior direction, were not different across work rates. It appears that during this type of stepping movement several of the individual detectors (X and Z counts) can sense the change associated with the increase in stepping height, but the conversion of this information to an estimated EE from the manufacturer's equation is not accurate. It is possible that future software or programming alterations by the Tritrac manufacturer may yield more accurate estimations than the present unit.

It is estimated that by the year 2025 the number of individuals aged 65 and older will increase from 32 million to 70 million ⁽⁷⁾. There has been very little validation research conducted on the Caltrac, and none on the Tritrac, in an older population. Washburn et al. ⁽¹⁶⁾ used an older population, but related the Caltrac measurements with physical activity diaries. Nichols et al. ⁽¹²⁾ compared the EE from indirect calorimetry with Caltrac estimated EE in young subjects (26 yr) to older subjects (64 yr) during treadmill walking. Subjects walked on a horizontal treadmill at speeds of 2.4-6.4 km·h⁻¹ (1.5-4.0 mph). In the older group, when speed was statistically controlled, there was very little relationship between Caltrac activity counts and EE (r = −0.02), whereas in the younger group there was a much stronger association (r = 0.51).

The present study used two types of ambulatory, weight-bearing movement to compare the accuracy of the Caltrac and Tritrac energy estimates to indirect calorimetry. The mode of activity and the intensity at which it was performed affected the accuracy of both instruments. If accelerometers or any other instrument that estimates EE is based on movement, then age, the mode of exercise, and the use of data obtained need to be considered. We are hopeful that with the advent of portable metabolic unit, the validation of these accelerometers in the field, using different age groups and a wide variety of physical activities, can be completed in the near future.