Share this article on:

00005768-199711000-0002200005768_1997_29_1535_coleman_relationships_11article< 112_0_9_3 >Medicine & Science in Sports & Exercise©1997The American College of Sports MedicineVolume 29(11)November 1997pp 1535-1542Relationships between TriTrac-R3D vectors, heart rate, and self-report in obese children[Special Communications: Methods]COLEMAN, KAREN J.; SAELENS, BRIAN E.; WIEDRICH-SMITH, MARGO D.; FINN, JEREMY D.; EPSTEIN, LEONARD H.Department of Psychology, University of Texas at El Paso, El Paso, TX; and Departments of Psychology, and Counseling and Educational Psychology, University at Buffalo, Buffalo, NYSubmitted for publication January 1996.Accepted for publication April 1997.Request for reprints: Leonard H. Epstein, Ph.D., Behavioral Medicine Laboratory, Psychology Department, Park Hall, University at Buffalo, Buffalo, NY 14260-4110.Address for correspondence: Karen J. Coleman, Ph.D., Psychology Department, University of Texas at El Paso, El Paso, TX, 79968-0553.ABSTRACTThe TriTrac (Professional Products, Inc., Madison, WI) triaxial accelerometer and diary self report were compared with adjusted heart rates to evaluate 3 d of leisure-time activity in 35 8- to 12-yr-old obese children. Adjusted heart rates were calculated by subtracting preexercise resting heart rates from heart rates measured in the field. TriTrac and self-reported data were converted to multiples of resting metabolic rate (METs). Correlations between accelerometer METs and adjusted heart rates (r = 0.71) were significantly higher (P < 0.001) than correlations between adjusted heart rates and self-reported METs (r = 0.36) or accelerometer and self-reported METs (r = 0.38). Self-reported METs had higher mean standard errors in estimating heart rates (13.93 ± 6.15 beats·min-1) than did accelerometer METs (10.94 ± 5.62 beats·min-1; P < 0.001), were significantly greater than accelerometer METs (2.50 ± 1.48 vs 1.80 ± 1.48; P< 0.05) and systematically overestimated accelerometer METs. The anteroposterior vector accounted for 36%, and the vector magnitude score accounted for 34% of the variance in unadjusted heart rates. The mediolateral vector and vector magnitude score accounted for 69% of the variance in self-reported METs. The vertical vector did not account for variance in either unadjusted heart rates or self-reported METs. It was concluded that the TriTrac yielded a better estimate of activity in obese children than self report. In addition, the vector magnitude composite score of the TriTrac accounted for significantly more variance in both self-reported activity and unadjusted heart rates as compared with the vertical directional vector of the TriTrac.Researchers have used questionnaires, direct observation, heart rate monitoring, doubly labeled water, and motion sensors to assess children's physical activity. Self-report questionnaires or recall interviews of physical activity represent the simplest procedures but have moderate reliability and validity in children (3). Direct observation provides reliable and valid estimates of activity (11), but it is very time consuming and expensive, particularly if activity patterns are to be assessed during child leisure time. Heart rate monitoring provides reliable measures of activity-induced energy expenditure(29,42), although it is affected by many variables other than changes in activity(15,16,26). Doubly labeled water procedures provide the most accurate estimates of energy expenditure; however, they are costly and do not provide information on specific patterns of activity (23,39).A practical and feasible option for monitoring activity is motion sensors. Motion sensors can either be used to assess activity(9,19,21,28,31,37), or to estimate energy expenditure(6-8,24). Much of the research using motion sensor assessments of activity and energy expenditure has used the Caltrac (Professional Products, A Division of Reining International, Ltd., Madison, WI) single-plane accelerometer. Investigators have reported significant correlations between the Caltrac and several methods of activity measurement, such as self-reported activity using questionnaires(27) and diaries (43), direct behavioral observation of activity in a controlled, experimental setting(9,29), and with heart rate monitoring in field settings (37,38). However, large errors in estimation have been reported with Caltrac measures of energy expenditure(24), including significant over-(7) or underestimation (8) of actual energy expenditure.These errors in estimation of energy expenditure may be a result of the Caltrac's single plane measurement of activity (28). Although there is disagreement as to whether triaxial accelerometers yield more information than vertical motion sensors, such as the Caltrac(18,35), a triaxial accelerometer may provide more accurate estimates of low levels of energy expenditure that are not well represented by movement in the vertical plane(7,25). Bouten et al. (6) found significant correlations during treadmill walking in adults (r = 0.96) between energy expenditure derived from the anteroposterior directional vector of a triaxial accelerometer and total energy expenditure measured with indirect calorimetry, while the composite vector was significantly associated with sedentary activity (r = 0.81). The vertical directional vector did not predict energy expenditure in any of the activities studied.A triaxial accelerometer, the TriTrac (TriTrac-R3D Research Ergometer, Professional Products, A Division of Reining International, Ltd., Madison, WI), has been made available for use in research. It collects movement information in three planes and stores minute by minute activity and estimated energy expenditure. It also has marked time intervals, an improvement on the triaxial accelerometer described by Bouten et al. (6). Matthews and Freedson (25) reported correlations as high as r = 0.82 between free-living self-reported and TriTrac activity measures in adults. In addition, Welk and Corbin (44) have demonstrated high correlations between TriTrac and Caltrac activity counts (r= 0.88), and moderate correlations between TriTrac activity counts and activity heart rates (r = 0.58) in 9-11 yr old boys. We completed a pilot study (36) using the TriTrac accelerometer to measure estimated rates of energy expenditure (METs) in 18 obese 8- to 12-yr-old children. Children were fitted with accelerometers and allowed access to sedentary (<2 METs) and active (≥3 METs) alternatives for 30 min in a controlled, experimental setting. Individual subject correlations between accelerometer and behavioral observation METs for the experimental session ranged from 0.40 to 0.85, with a mean of 0.66. The magnitude of these correlations is similar to that found between direct observation and the Caltrac (9,28).Although data exist on the relationships between self-report, heart rate, and the TriTrac (25,44), no investigation to date has analyzed whether the TriTrac can be used to index low levels of activity in populations such as the obese or whether the vector magnitude composite activity counts provide more information about movement than activity counts from a single directional vector. To address these issues, we assessed the validity of the TriTrac in monitoring leisure-time activity in 8- to 12-yr-old obese male and female children, using heart rate as the criterion measure. Heart rate was chosen as the criterion because it has been shown to be linearly related to oxygen consumption (15,26) and thus energy expenditure and is a good physiologic indicator of children's activity in the field (12,38). We also assessed the validity of diary self-report using heart rate as the criterion measure and compared these analyses to those of the TriTrac. Diary self-report is one of the most common ways to assess activity in the field(38) and has been shown to be the most accurate form of self-report in children (3).METHODSubjectsHeart rate, self-reported activity and accelerometer data were collected from 35 obese children between the ages of 8-12 who had been recruited through community advertisement and physician referral and accepted for a weight control program. Informed consent was obtained from the parents or legal guardians of all children before their participation. Approval for the study was granted by the social sciences institutional review board of the University at Buffalo. All activity measures were collected before treatment began. Average sample characteristics ± the SD for those subjects were as follows: age = 10.33 ± 1.16 yr; height = 1.46 ± 0.09 m; weight = 58.6 ± 11.4 kg; BMI = 27.24 ± 3.12 kg·m-2; percent overweight = 60.72 ± 17.61%; percent body fat = 33.89 ± 4.11%. The sample was 57% female and 43% male.MeasurementHeight, weight, and body fat. Height was measured to the nearest 0.32 cm using a stadiometer. Weight was measured to the nearest 0.25 lb on a standard balance scale, and BMI was calculated by dividing weight (kg) by height2 (m2). Percent overweight was calculated by referencing the observed BMI to the 50th percentile BMI for the specific child age and sex(30). Body fat was assessed using bioelectrical impedance (Model BIA-101; RJL Systems; Clinton TWP, MI). This model uses an equation developed by Houtkooper et al. (17) for white children 10-19 yr of age. Specifically this equation is FFB = 0.628RI + 0.246WT + 0.061 × C - 2.773 (FEB is fat free body mass; RI is[height2 (cm2)/resistance value]; WT is the weight in kilograms; and XC is the reactance value).Preexercise resting heart rates. Resting heart rate was assessed before each child began a submaximal fitness test on a cycle ergometer(preexercise). Heart rate was monitored using the Polar Vantage XL heart rate monitor (Polar CIC Inc., Port Washington, NY). Each subject was seated quietly on the bike and heart rate was monitored for 2 to 5 min. When heart rate was stable (heart rate within 8 beats·min-1) heart rate measures were taken every 15 s for 1 min. Preexercise resting heart rate was calculated as an average of these stable heart rate measures.Monitoring devices. Field measures of movement and energy expenditure were collected using the TriTrac accelerometer, and heart rates in the field were monitored using the Polar Vantage XL heart rate monitor. The accelerometer and heart rate monitoring devices were programmed before field testing to collect minute by minute data. No equipment failures occurred during field testing. Information was downloaded from all accelerometers and heart rate monitors to a personal computer using software provided by the manufacturer.All subjects were instructed on how to wear the heart rate monitor and accelerometer and how to operate the heart rate monitor wrist watch. The accelerometer was worn at the waist above the right hip with an elastic belt onto which the accelerometer was fastened in place with velcro and elastic straps. Positioning activity monitors around the waist over the hip has been shown to produce reliable measures of torso movement (2). Conducting gel was applied to the heart rate transmitter band, and it was fastened around the chest underneath the breastbone, while the receiver watch was worn on either wrist. Activity was assessed on Thursday and Friday, from the time the children got home from school until they went to bed, and Saturday, from the time they got up in the morning until they went to bed. They were told to take the monitors off while swimming or bathing. All measures were used to assess leisure time activity, defined as activity engaged in by choice, rather than that required during school hours (i.e., physical education).Self-report. Subjects were instructed to keep detailed self-reports of the types of behaviors they engaged in and the times they began and ended these behaviors during the times the accelerometer and heart rate monitor were worn. The self-report sheets for Thursday and Friday were divided into before school, during school, after school, and after dinner columns. Children only recorded those activities they did after school and after dinner. Saturday self-report sheets were divided into morning, afternoon, and evening. Children recorded all activities that they did on Saturday from the time they got out of bed in the morning until they got into bed at night. Each column included several general active and sedentary categories (watching television or videotapes, walking, running, etc.), as well as blank spaces to write in specific behaviors. Subjects were asked to record the time they began each activity in one space and the time they ended the activity in the following space. There were also blanks provided for subjects to describe the activity (i.e., what programs were watched on television). This “segmented” approach to designing the self-report sheet has been shown to produce more accurate indications of activity patterns in children than data gathered using blank sheets without any categorization (3). Self-report sheets were reviewed with a trained technician to ensure a detailed account of all daily activities. Self-reported data collected when the child did not wear the accelerometer and heart rate monitors (before and during school) were not used for any analyses.Data ReductionAccelerometer. The accelerometer collected minute by minute data in three planes of movement (mediolateral (X), anteroposterior (Y), and vertical (Z)), and provided one composite movement score for all three directions called the “vector magnitude” ([x2 + y2 + z2]1/2). The accelerometer computed estimated energy expenditure(kcal) for each minute of data collection using the vector magnitude movement count in a formula developed for the Caltrac activity monitor. In addition, the accelerometer provided an estimate of resting metabolic rate (RMR) using formulas that were developed at the hemokinetics facilities with healthy adults. These formulas overestimate resting metabolic rate by 7% ± 12.3% in children (7). Research by Bray et al.(7,8) has demonstrated that the Caltrac calculations for both RMR and activity kilocalories can be used to assess energy expenditure in children with an acceptable level of error.Each minute of accelerometer data was converted to METs (multiples of resting metabolic rate) by dividing the activity kilocalories computed by the accelerometer for that minute by each child's calculated RMR. MET values were used as an indication of a general level of intensity across various activity durations and different body weights (1). Average accelerometer METs were calculated for each nonschool 10-min time block on Thursday, Friday, and Saturday.Self report. Self-reported activity for each nonschool 10-min time block on Thursday, Friday, and Saturday was used for data analyses. Activities engaged in for a duration of 5 min or more were entered in 10-min blocks as MET values using the Compendium of Physical Activities(1). Most activities such as running, biking, swimming, and other common physical activities, as well as chores around the house, were assigned MET values from the general categories listed in the Compendium. Sports such as basketball, volleyball, and football were classified as nonteam sports, unless the children indicated they played as part of an organized team. Hopscotch, jumping rope and playground activities (swingset, dodgeball, kickball, general playing) were all considered under the general Compendium heading of “children's games.” Sedentary activities such as sitting, standing, talking on the phone, watching television, riding in a car, reading, lying quietly without sleeping, etc. were assigned MET values from the compendium.Heart rates. Adjusted heart rates were used for all correlational analyses. Adjusted heart rates were calculated by subtracting preexercise resting heart rates from heart rates measured in the field. Adjusted heart rates were calculated because of evidence provided by Welk and Corbin(44) that the TriTrac accounted for more variance in children's heart rates corrected for resting heart rate as compared to uncorrected heart rate data. In this study, Welk and Corbin(44) used an average of the five lowest heart rates throughout the day as an estimate of resting heart rate, following the methods presented by Sallis et al. (37). However, Durant et al.(12) felt that this calculation underestimated resting heart rate and recommended using actual resting heart rate whenever possible to obtain a measure of activity heart rate.We had obtained actual measures of resting heart rate before each participant's fitness test; however, we were concerned that these heart rates might have been artificially elevated owing to anticipation of the fitness test. Therefore, we conducted a repeated measures multivariate ANOVA comparing the resting heart rate of each child before fitness testing with both an average of the five lowest heart rates collected in the field and the single lowest heart rate collected in the field. The Hotelling's T2 statistic was used, with a Bonferroni adjustment for three comparisons (P = 0.05/3 = 0.02), for post hoc comparisons. We found that heart rates obtained before fitness testing (83 ± 10 beats·min-1) were significantly lower (F(1,34) = 7.85; P < 0.001) than an average of the five lowest heart rates collected in the field (88± 11 beats·min-1), and were not statistically different(F(1,34) = 0.46; P = 0.50) from the single lowest heart rate collected in the field (84 ± 9 beats·min-1). Based on these results and the suggestions of Durant et al.(12), adjusted heart rates were obtain for analyses by subtracting the resting heart rate recorded for each subject during their fitness test from each minute of their heart rate data collected in the field. An average adjusted heart rate was then calculated for each subject using each nonschool 10-min time block on Thursday, Friday, and Saturday.Statistical analyses. Anthropometric and correlational data were first analyzed for effects of gender. Gender was used as a between-subjects factor for all one-way ANOVA on body composition measures and all repeated measures MANOVA for differences among activity correlations. No effects of gender were found in any of the analyses; consequently, results were combined for male and female children.Correlations. Ten-minute interval data for each subject were used to generate Pearson Product Moment correlations between activity measures. This resulted in 35 individual correlations between adjusted heart rates and accelerometer METs (rhra), adjusted heart rates and self-reported METs(rhrsr), and accelerometer and self-reported METs (rasr). Only 10-min intervals that contained data for all three activity measures were used to generate correlations. Each subject had a different number of intervals used to compute correlations between activity measures. The distribution of 10-min intervals across all subjects for correlational analyses ranged from 8 to 132 intervals across the 3 d, with 26% of subjects having less than 30 intervals, 23% of subjects having between 30 and 50 intervals, and 51% of subjects having more than 50 intervals. All correlations were transformed to a Fisher's Z statistic to approximate the normal distribution(32) for analyses. Means and SDs for all correlations were computed using Z-transformed correlations and then converted back to Pearson's values for presentation.Correlations between activity measures were compared to determine how strongly accelerometer and self-reported data were related to adjusted heart rates. Before analyzing statistical differences among correlations, a test for reducibility (sphericity) was performed to determine if an ANOVA for repeated measures could be used (41). The conditions for reducibility were not met so that statistical differences among correlations were assessed through multivariate ANOVA for repeated measures using the Multivariance program (14). This analysis treated correlations between the activity measures as multiple dependent measures. All main effects for correlational analyses are expressed with the Wilks' LambdaF-statistic (41). The Hotelling's T2 statistic was used, with a Bonferroni adjustment for three comparisons(P = 0.05/3 = 0.02), to determine which correlations were contributing to any differences found. Post hoc analyses are expressed with the Wilks' Lambda F-statistic(41).Individual SEE were also obtained to determine the measurement error in predicting adjusted heart rates from self-reported and accelerometer METs. SEEs were compared using a paired-samples t-test.Comparisons of self-reported and accelerometer METs. Self-reported METs were compared with accelerometer METs using a technique developed by Bland and Altman (4,5) to assess the agreement between two methods of measurement. This method was used by Bray et al. (8) to evaluate Caltrac energy measurements in comparison with measures of oxygen consumption. For each subject, the difference between accelerometer and self-reported METs is plotted against the mean of these two measures. A zero line drawn on the graph (seeFig. 1) represents no error in measurement between the methods. A high correlation between the difference in METs and the mean of self-reported and accelerometer METs suggests systematic bias (an increase in bias as the magnitude of both measures increases) between methods under comparison. This analysis, in combination with a paired-samplest-test, was used to assess the differences between accelerometer and self-reported METs.Figure 1-The difference between self-reported and accelerometer METs plotted against the mean of self-reported and accelerometer METs for each subject. A line of identity is drawn (r = 0.93; P < 0.001), as well as a line at which there are no differences between self-reported and TriTrac METs. This represents zero error in measurement between the two methods. SR = self-report; TRI = TriTrac.Vector analyses. Stepwise regression analyses were performed to determine the amount of variance accounted for in unadjusted heart rates and self-reported METs by the mediolateral, anteroposterior, vertical, and vector magnitude composite activity counts of the accelerometer. Variables were included in the models at P = 0.05. To control for variance in field measures of heart rate resulting from resting heart rate, resting heart rates were entered first into all vector models. Unadjusted heart rates and self-reported METs were regressed on vector magnitude activity counts separately from the unidirectional vectors. Based upon the findings of Bouten et al. (6), directional vectors were entered into both unadjusted heart rates and self-reported METs regressions in the following order: anteroposterior counts, vertical counts, and then mediolateral counts.The z statistic to test differences in correlations(41) was used to compare regression models generated with vector magnitude activity counts to models resulting from the analysis of the unidirectional activity counts. This comparison provided a statistical basis for determining whether one or more vector(s) of the accelerometer accounted for more variance in unadjusted heart rates or self-reported METs than the accelerometer's composite vector magnitude score.All data reduction and statistical analyses were conducted using SYSTAT for Windows (SYSTAT, Inc.: Evanston, IL) and the Multivariance program(14).RESULTSMeans and SDs are presented for all activity data inTable 1. Mean and SD for all correlations between activity data, and the mean and SD for the SEE are shown inTable 2. A main effect for differences among activity measure correlations was found (F (2,33) = 19.69; P < 0.001). Hotelling's T2 comparisons revealed that correlations between accelerometer METs and adjusted heart rates (rhra = 0.71) were higher than both correlations between accelerometer and self-reported METs(rasr = 0.38) (F (1,34) = 30.77; P < 0.001), and adjusted heart rates and self-reported METs (rhrsr = 0.36) (F(1,34) = 40.21; P < 0.001). There were no statistical differences among adjusted heart rates and self-reported METs correlations and accelerometer and self-reported METs correlations (F (1,34) = 0.12;P = 0.73). Mean self-reported standard errors of estimating adjusted heart rates (13.93 ± 6.15 beats·min-1) were higher than mean accelerometer standard errors of estimating adjusted heart rates (10.94± 5.62) (t (34) = 5.19; P < 0.001).TABLE 1. Mean and SD of accelerometer METs (TRI), adjusted heart rates(AHR), self-reported METs (SR), vector magnitude activity counts (VEC), antero-posterior activity counts (Y), vertical activity counts (Z), and medio-lateral activity counts (X) (N = 35).TABLE 2. Means and SDs of correlations, and means and SDs of SEE for the accelerometer (TRI), adjusted heart rate (AHR), and self-report (SR)(N = 35).The difference between self-reported and accelerometer METs is plotted against the mean of self-reported and accelerometer METs for each subject inFigure 1. Most of the points are above the zero line. There is a high correlation between the difference and mean of self-reported and accelerometer METs (r = 0.93; P < 0.001), indicating systematic bias between these measurement methods. It appears that self-reported activity was higher than accelerometer activity as children became more active. In addition to these findings, overall self-reported METs were significantly higher than accelerometer METs (t (34) = 5.76;P < 0.001).For all vector regression analyses, preexercise resting heart rates did not account for a significant amount of variance in unadjusted heart rates. After the addition of preexercise resting heart rates, vector magnitude activity counts accounted for 34% (P = 0.001) of the variance in unadjusted heart rates (R2adj = 0.34). In a separate regression, the anteroposterior activity counts, added after preexercise resting heart rates, accounted for 36% (P < 0.001) of the variance in unadjusted heart rates (R2adj = 0.36), while activity counts from the mediolateral and vertical directional vectors did not account for any additional variance explained in unadjusted heart rates.Activity counts from the vector magnitude and the anteroposterior vector were highly correlated (r = 0.98; P < 0.0001). Different models predicting variance in unadjusted heart rate were compared using the z statistic. This analysis revealed that vector magnitude and anteroposterior vector activity counts accounted for the same amount of variance (34% and 36%, respectively) in unadjusted heart rates (z (32) = -0.75; P > 0.10).Vector magnitude activity counts accounted for 69% of the variance in self-reported METs (R2adj = 0.69;P < 0.001). In a separate regression, activity counts from the mediolateral vector also accounted for 69% of the variance in self-reported METs (R2adj = 0.69; P < 0.001), while anteroposterior and vertical activity counts did not account for any additional variance in self-reported METs. Vector magnitude and mediolateral activity counts were also highly correlated(r = 0.93; P < 0.001).DISCUSSIONThe mean correlation between adjusted heart rates and accelerometer METs(rhra = 0.71) was higher than correlations reported in children between adjusted heart rates and the Caltrac (37) and the TriTrac (44). Despite the high correlations between adjusted heart rates and accelerometer METs, there was a higher SEE using the TriTrac (10.94 ± 5.62 beats·min-1) as compared with that reported by Sallis et al. (37) for the Caltrac (7.5 beats·min-1). This may reflect variability in heart rates resulting from factors other than physical activity rather than error in the TriTrac's measurement of activity. It has been well established that heart rate is subject to change in response to factors other than increases in activity (12,15,16,26), particularly during low intensity activities(26,33,34,45). Racette et al.(34) found that there was a weak relationship between oxygen consumption and heart rates for activities below a 3 MET intensity level in obese adults. In the present study, the mean accelerometer METs was only 1.80 ± 1.48, indicating that children primarily engaged in low intensity activities throughout the sample period and were below the activity intensity reported by Racette et al. (34) for a linear relationship between heart rates and oxygen consumption. At this level of activity, heart rates may be elevated by cognitive factors such as stress, independent of any changes in body movement (33). Meijer et al. (26) also reported a great deal of heart rate variability for adults during low energy tasks in the laboratory, as well as much more variability for heart rate in the field, as compared to a triaxial accelerometer they developed.Individual correlations between adjusted heart rates and accelerometer METs were higher than both correlations between adjusted heart rates and self-reported METs (rhrsr = 0.36), and accelerometer and self-reported METs (rasr = 0.38). Research has demonstrated a poor relationship between self-reported activity and that detected using motion sensors in adults (10) and children (40). Low levels of agreement between self-reported and motion sensor activity are assumed to result from the fact that children and adults over-report both the duration and intensity of their physical activity(22,38,40). There is also evidence in adults that the frequency and duration of sedentary activities are under-reported (20). The present study provides data in 8- to 12-yr-old obese children to support these issues, demonstrating that there was a systematic overestimation of self-reported METs as compared with accelerometer METs and that this overestimation increased as the intensity of the activity increased. In addition, overall self-reported METs were significantly higher than accelerometer METs, self-reported activity was poorly correlated with both accelerometer activity and adjusted heart rates, and there were higher SEEs (13.93 ± 6.15 beats·min-1) when self-reported METs were used to predict variance in adjusted heart rates as compared with accelerometer SEEs (10.94 ± 5.62 beats·min-1). Sedentary, obese adults may over-report their activity because of the pressure to become more physically active(34), and young children may have an exaggerated perception of the time and effort level involved in physical activity(3).There has been some controversy concerning the additional activity information a triaxial monitor may provide above that recorded with uniaxial activity monitors (18,35). Redmond and Hegge(35) designed and tested their own triaxial activity monitor and contended that the vertical vector provided the same activity information as the anteroposterior and mediolateral vectors. However, this comparison was done with a wrist-worn activity monitor to evaluate adult patterns of activity. Bouten et al. (6) evaluated the use of a triaxial accelerometer, worn around the waist, that was based on the same piezoresistive technology as the TriTrac. This accelerometer provided information about the anteroposterior, mediolateral, and vertical planes of motion, as well as a summation vector with the same calculation as the vector magnitude for the TriTrac. Using oxygen consumption as the standard, they found that the summation vector was the best indicator of sedentary energy expenditure (r = 0.81), while the anteroposterior vector was the best indicator of physically active (primarily walking) energy expenditure (r = 0.96). They also found that energy expended during sitting with arm movement was best explained with the mediolateral vector. For all activities combined, both the summation (r = 0.95) and anteroposterior (r = 0.97) vectors were significantly related to energy expenditure.We supported the findings of Bouten et al. (6) in that only the anteroposterior and the vector magnitude vectors of the TriTrac significantly accounted for variance in unadjusted heart rates (36% and 34%, respectively). However, we also found that both mediolateral and vector magnitude activity counts accounted for 69% of the variance in children's self-reported METs. Activity counts from the anteroposterior vector did not significantly account for variance in self-reported energy expenditure.The differences in the relationships between activity vectors and unadjusted heart rates, and self-reported METs and activity vectors may be a reflection of the type of activity that heart rate and self report represent. Bouten et al. (6) found that the anteroposterior vector was related to walking, while the mediolateral vector was related to sitting with arm movement. It is possible that heart rate is more influenced by movement in the anteroposterior direction such as walking, and that children are more likely to self report activities that occur in the mediolateral plane of motion such as sitting while playing video or computer games. Both of these activities would be represented in the vector magnitude composite movement counts. Neither Bouten et al. (6) nor this investigation found that the vertical vector was the best characterization of physically active or sedentary energy expenditure, suggesting that the vertical directional vector of these triaxial accelerometers may not accurately reflect energy expenditure in the laboratory with adults or in the field with obese children.The results of this investigation suggest the TriTrac was related to changing activity in obese children and was a better estimate of energy expenditure in the field than self report. In addition, the vector magnitude composite activity counts of the TriTrac may provide a more comprehensive estimate of sedentary and physical activity in obese children than the vertical directional vector of the TriTrac. Further investigation is needed using more established field measures of energy expenditure, such as doubly-labeled water, to determine the accuracy of the TriTrac in estimating energy expenditure for a wide range of activities in various groups of children and adults.The correlational results of this investigation can only be applied to activity measured in time series throughout an extended period such as a day or several days. The validity of using the TriTrac to assess energy expenditure for single time points has yet to be established. Because of the collection of minute by minute activity data in three planes of motion for up to 30 d with specific interval markers, research with the TriTrac will be able to generate analyses of patterns of activity in a wide range of sedentary and active populations.Author's Note: Since the publication of this study, Professional Products has developed a new version of the TriTrac now available for research. This version is smaller than the TriTrac used in our study, and the anteroposterior directional vector is now referred to as the “Z” vector, while the vertical directional vector is now referred to as the“Y” vector. The mediolateral vector is still called the“X” vector.Appreciation is expressed to Katarina Anderson for assistance in data management, and Dr. Juliet Shaffer, Department of Statistics, University of California at Berkeley, for her consultation regarding correlational analyses.This study was funded in part by NIH grant RO1 HD 25997 awarded to Leonard H. Epstein, Ph.D.REFERENCES1. Ainsworth, B. E, W. L. Haskell, A. S. Leon, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med. Sci. Sports Exerc. 25:71-80, 1993. [CrossRef] [Full Text] [Medline Link] [Context Link]2. Balogun, J. A., L. O. Amusa, and I. U. Onyewadume. Factors affecting Caltrac® and Calcount® accelerometer output.Physiol. Ther. 68:1500-1504, 1988. [Context Link]3. Baranowski, T. Validity and reliability of self report measures of physical activity: an information processing perspective.Res. Q. Exerc. Sport 59:314-327, 1988. [CrossRef] [Context Link]4. Bland, J. M. and D. G. Altman. Statistical methods for assessing agreement between two methods of clinical measurement.Lancet 1(8476):307-310, 1986. [Medline Link] [Context Link]5. Bland, J. M. and D. G. Altman. Comparing methods of measurement: why plotting difference against standard method is misleading.Lancet 346(8982):1085-1087, 1995. [CrossRef] [Full Text] [Medline Link] [Context Link]6. 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. [CrossRef] [Full Text] [Medline Link] [Context Link]7. Bray, M. S., J. R. Morrow, Jr., J. M. Pivarnik, and J. T. Bricker. Caltrac® validity for estimating caloric expenditure with children. Pediatr. Exerc. Sci. 4:166-179, 1992. [Context Link]8. Bray, M. S., W. W. Wong, J. R. Morrow, Jr., N. F. Butte, and J. M. Pivarnik. Caltrac® versus calorimeter determination of 24-h energy expenditure in female children and adolescents. Med. Sci. Sports Exerc. 26:1524-1530, 1994. [CrossRef] [Full Text] [Medline Link] [Context Link]9. Danner, F., M. Noland, M. McFadden, K. Dewalt, and J. M. Kotchen. Description of the physical activity of young children using movement sensor and observation methods. Pediatr. Exerc. Sci. 3:11-20, 1991. [Context Link]10. Dishman, R. K., C. R. Darracott, and L. T. Lambert. Failure to generalize determinants of self-reported activity to a motion sensor. Med. Sci. Sports Exerc. 24:904-910, 1992. [CrossRef] [Full Text] [Medline Link] [Context Link]11. Durant, R. H., T. Baranowski, J. Puhl, et al. Evaluation of the children's activity rating scale (CARS) in young children.Med. Sci. Sports Exerc. 25:1415-1421, 1993. [CrossRef] [Full Text] [Medline Link] [Context Link]12. Durant, R. H., T. Baranowski, H. Davis, et al. Reliability and variability of heart rate monitoring in 3-, 4-, or 5-yr-old children. Med. Sci. Sports Exerc. 24:265-271, 1992. [CrossRef] [Full Text] [Medline Link] [Context Link]13. Ferguson, G. A. and Y. Takane. Statistical Analysis in Psychology and Education, 6th Ed. New York: McGraw-Hill Book, 1989, pp. 198-225.14. Finn, J. D. and R. D. Bock. Multivariance VII User's Guide. Chicago: Scientific Software, Inc., 1985. [Context Link]15. Freedson, P. S. Electronic motion sensors and heart rate as measures of physical activity in children. J. Sch. Health 61:220-223, 1991. [CrossRef] [Medline Link] [Context Link]16. Haskell, W. L., A. S. Leon, C. J. Caspersen, et al. Cardiovascular benefits and assessment of physical activity and physical fitness in adults. Med. Sci. Sports Exerc. 24:S201-S220, 1992. [Medline Link] [Context Link]17. Houtkooper, L. B., S. B. Going, T. G. Lohman, A. F. Roche, and M. Van Loan. Bioelectrical impedance estimation of fat-free body mass in children and youth: a cross-validation study. J. Appl. Physiol. 72:366-373, 1992. [Medline Link] [Context Link]18. Janz, K. F. Validation of the CSA accelerometer for assessing children's physical activity. Med. Sci. Sports Exerc. 26:369-375, 1994. [CrossRef] [Full Text] [Medline Link] [Context Link]19. Klesges, L. M., and R. C. Klesges. The assessment of children's physical activity: a comparison of methods. Med. Sci. Sports Exerc. 19:511-517, 1987. [CrossRef] [Full Text] [Medline Link] [Context Link]20. Klesges, R. C., L. H. Eck, M. W. Mellon, W. Fullington, G. W. Somes, and C. L. Hanson. The accuracy of self-reports of activity.Med. Sci. Sports Exerc. 22:690-697, 1990. [CrossRef] [Full Text] [Medline Link] [Context Link]21. Klesges, R. C., L. M. Klesges, A. M. Swenson, and A. M. Pheley. A validation of two motion sensors in the prediction of child and adult physical activity levels. Am. J. Epidemiol. 122:400-410, 1985. [Medline Link] [Context Link]22. Lichtman, S. W., K. Pisarska, E. R. Berman, et al. Discrepancy between self-reported and actual caloric intake and exercise in obese subjects. N. Engl. J. Med. 327:1893-1898, 1992. [Context Link]23. Livingstone, M. B. E., W. A. Coward, A. M. Prentice, et al. Daily energy expenditure in free-living children: comparison of heartrate monitoring with the doubly labeled water (2H218O) method.Am. J. Clin. Nutr. 56:343-352, 1992. [Context Link]24. Maliszewski, A. F., P. S. Freedson, C. J. Ebbeling, J. A. Crussemeyer, and K. B. Kastango. Validity of the Caltrac® accelerometer in estimating energy expenditure and activity in children and adults.Pediatr. Exerc. Sci. 3:141-151, 1991. [Context Link]25. Matthews, C. E. and P. S. Freedson. Field trial of a three-dimensional activity monitor: comparison with self-report. Med. Sci. Sports Exerc. 27:1071-1078, 1995. [CrossRef] [Full Text] [Medline Link] [Context Link]26. Meuer, G. A., K. A. Westerterp, J. Koper, and F. T. Hoor. Assessment of energy expenditure by recording heart rate and body acceleration. Med. Sci. Sports Exerc. 21:343-347, 1989. [Context Link]27. Miller, D. J., P. S. Freedson, and G. M. Kline. Comparison of activity levels using the Caltrac® accelerometer and five questionnaires. Med. Sci. Sports Exerc. 26:376-382, 1994. [CrossRef] [Full Text] [Medline Link] [Context Link]28. Mukeshi, M., B. Gutin, W. Anderson, P. Zybert, and C. Basch. Validation of the Caltrac® movement sensor using direct observation in young children. Pediatr. Exerc. Sci. 2:249-254, 1990. [Context Link]29. Murgatroyd, P. R., P. S. Shetty, and A. M. Prentice. Techniques for the measurement of human energy expenditure: a practical guide.Int. J. Obes. 17:549-568, 1993. [Context Link]30. Must, A., G. E. Dallal, and W. H. Dietz. Reference data for obesity: 85th and 95th percentiles of body mass index (wt/ht2) and triceps skinfold thickness. Am. J. Clin. Nutr. 53:839-846, 1991. [Medline Link] [Context Link]31. Noland, M., F. Danner, K. Dewalt, M. McFadden, and J. M. Kotchen. The measurement of physical activity in young children.Res. Q. Exerc. Sport 61:146-153, 1990. [CrossRef] [Medline Link] [Context Link]32. Olkin, I., and J. D. Finn. Correlations redux.Psychol. Bull. 118:155-164, 1995. [Context Link]33. Patterson, S. M., D. S. Krantz, L. C. Montgomery, P. A. Deuster, S. M. Hedges, and L. E. Nebel. Automated physical activity monitoring: validation and comparison with physiological and self-report measures. Psychophysiology 30:296-305, 1993. [CrossRef] [Medline Link] [Context Link]34. Racette, S. B., D. A. Schoeller, and R. F. Kushner. Comparisons of heart rate and physical activity recall with doubly labeled water in obese women. Med. Sci. Sports Exerc. 27:126-133, 1995. [Context Link]35. Redmond, D. P. and F. W. Hegge. Observations on the design and specification of a wrist-worn human activity monitoring system.Behavior Research Methods, Instruments, & Computers 17:659-669, 1985. [Context Link]36. Saelens, B. E., L. H. Epstein, and J. G. O'Brien. The Tritrac-R3D® activity monitor and behavioral observation: a validity study. Ann. Behav. Med. 17:S113 (Abstract), 1995. [Context Link]37. Sallis, J. F., M. J. Buono, J. J. Roby, D. Carlson, and J. A. Nelson. The Caltrac® accelerometer as a physical activity monitor for school-age children. Med. Sci. Sports Exerc. 22:698-703, 1990. [CrossRef] [Full Text] [Medline Link] [Context Link]38. Saris, W. H. M. Habitual physical activity in children: methodology and findings in health and disease. Med. Sci. Sports Exerc. 18:253-263, 1986. [CrossRef] [Full Text] [Medline Link] [Context Link]39. Schoeller, D. A., L. G. Bandini, and W. H. Dietz. Inaccuracies in self-reported intake identified by comparison with the doubly labeled water method. Can. J. Physiol. Pharmacol. 68:941-949, 1990. [CrossRef] [Medline Link] [Context Link]40. Simons-Morton, B. G., W. C. Taylor, and I. W. Huang. Validity of the physical activity interview and caltrac with preadolescent children. Res. Q. Exerc. Sport 65:84-88, 1994. [Medline Link] [Context Link]41. Tabachnick, B. G. and L. S. Fidell. Using Multivariate Statistics, 2nd Ed. New York: Harper Collins, 1989, pp. 158-159. [Context Link]42. Treiber, F. A., L. Musante, S. Hartdagan, H. Davis, J. Levy, and W. B. Strong. Validation of a heart rate monitor with children in laboratory and field settings. Med. Sci. Sports Exerc. 21:338-342, 1989. [CrossRef] [Full Text] [Medline Link] [Context Link]43. Washburn, R. A., C. A. Janney, and J. R. Fenster. The validity of objective physical activity monitoring in older individuals.Res. Q. Exerc. Sport 61:114-117, 1990. [CrossRef] [Medline Link] [Context Link]44. Welk, G. J. and C. B. Corbin. The validity of the Tritrac-R3D® activity monitor for the assessment of physical activity in children. Res. Q. Exerc. Sport 66:1-8, 1995. [Context Link]45. Williams, E., R. C. Klesges, C. L. Hanson, and L. H. Eck. A prospective study of the reliability and convergent validity of three physical activity measures in a field research trial. J. Clin. Epidemiol. 42:1161-1170, 1989. [CrossRef] [Medline Link] [Context Link]CORRELATION; ACCELEROMETER; DIRECTIONAL VECTORS;|00005768-199711000-00022#xpointer(id(R1-22))|11065213||ovftdb|00005768-199301000-00011SL000057681993257111065213P62[CrossRef]|00005768-199711000-00022#xpointer(id(R1-22))|11065404||ovftdb|00005768-199301000-00011SL000057681993257111065404P62[Full Text]|00005768-199711000-00022#xpointer(id(R1-22))|11065405||ovftdb|00005768-199301000-00011SL000057681993257111065405P62[Medline Link]|00005768-199711000-00022#xpointer(id(R3-22))|11065213||ovftdb|SL0000696619885931411065213P64[CrossRef]|00005768-199711000-00022#xpointer(id(R4-22))|11065405||ovftdb|SL000055311986130711065405P65[Medline Link]|00005768-199711000-00022#xpointer(id(R5-22))|11065213||ovftdb|00005531-199510210-00017SL000055311995346108511065213P66[CrossRef]|00005768-199711000-00022#xpointer(id(R5-22))|11065404||ovftdb|00005531-199510210-00017SL000055311995346108511065404P66[Full Text]|00005768-199711000-00022#xpointer(id(R5-22))|11065405||ovftdb|00005531-199510210-00017SL000055311995346108511065405P66[Medline Link]|00005768-199711000-00022#xpointer(id(R6-22))|11065213||ovftdb|00005768-199412000-00016SL00005768199426151611065213P67[CrossRef]|00005768-199711000-00022#xpointer(id(R6-22))|11065404||ovftdb|00005768-199412000-00016SL00005768199426151611065404P67[Full Text]|00005768-199711000-00022#xpointer(id(R6-22))|11065405||ovftdb|00005768-199412000-00016SL00005768199426151611065405P67[Medline Link]|00005768-199711000-00022#xpointer(id(R8-22))|11065213||ovftdb|00005768-199412000-00017SL00005768199426152411065213P69[CrossRef]|00005768-199711000-00022#xpointer(id(R8-22))|11065404||ovftdb|00005768-199412000-00017SL00005768199426152411065404P69[Full Text]|00005768-199711000-00022#xpointer(id(R8-22))|11065405||ovftdb|00005768-199412000-00017SL00005768199426152411065405P69[Medline Link]|00005768-199711000-00022#xpointer(id(R10-22))|11065213||ovftdb|00005768-199208000-00012SL0000576819922490411065213P71[CrossRef]|00005768-199711000-00022#xpointer(id(R10-22))|11065404||ovftdb|00005768-199208000-00012SL0000576819922490411065404P71[Full Text]|00005768-199711000-00022#xpointer(id(R10-22))|11065405||ovftdb|00005768-199208000-00012SL0000576819922490411065405P71[Medline Link]|00005768-199711000-00022#xpointer(id(R11-22))|11065213||ovftdb|00005768-199312000-00016SL00005768199325141511065213P72[CrossRef]|00005768-199711000-00022#xpointer(id(R11-22))|11065404||ovftdb|00005768-199312000-00016SL00005768199325141511065404P72[Full Text]|00005768-199711000-00022#xpointer(id(R11-22))|11065405||ovftdb|00005768-199312000-00016SL00005768199325141511065405P72[Medline Link]|00005768-199711000-00022#xpointer(id(R12-22))|11065213||ovftdb|00005768-199202000-00018SL0000576819922426511065213P73[CrossRef]|00005768-199711000-00022#xpointer(id(R12-22))|11065404||ovftdb|00005768-199202000-00018SL0000576819922426511065404P73[Full Text]|00005768-199711000-00022#xpointer(id(R12-22))|11065405||ovftdb|00005768-199202000-00018SL0000576819922426511065405P73[Medline Link]|00005768-199711000-00022#xpointer(id(R15-22))|11065213||ovftdb|SL0000531819916122011065213P76[CrossRef]|00005768-199711000-00022#xpointer(id(R15-22))|11065405||ovftdb|SL0000531819916122011065405P76[Medline Link]|00005768-199711000-00022#xpointer(id(R16-22))|11065405||ovftdb|SL00005768199224s20111065405P77[Medline Link]|00005768-199711000-00022#xpointer(id(R17-22))|11065405||ovftdb|SL0000456019927236611065405P78[Medline Link]|00005768-199711000-00022#xpointer(id(R18-22))|11065213||ovftdb|00005768-199403000-00015SL0000576819942636911065213P79[CrossRef]|00005768-199711000-00022#xpointer(id(R18-22))|11065404||ovftdb|00005768-199403000-00015SL0000576819942636911065404P79[Full Text]|00005768-199711000-00022#xpointer(id(R18-22))|11065405||ovftdb|00005768-199403000-00015SL0000576819942636911065405P79[Medline Link]|00005768-199711000-00022#xpointer(id(R19-22))|11065213||ovftdb|00005768-198710000-00015SL0000576819871951111065213P80[CrossRef]|00005768-199711000-00022#xpointer(id(R19-22))|11065404||ovftdb|00005768-198710000-00015SL0000576819871951111065404P80[Full Text]|00005768-199711000-00022#xpointer(id(R19-22))|11065405||ovftdb|00005768-198710000-00015SL0000576819871951111065405P80[Medline Link]|00005768-199711000-00022#xpointer(id(R20-22))|11065213||ovftdb|00005768-199010000-00022SL0000576819902269011065213P81[CrossRef]|00005768-199711000-00022#xpointer(id(R20-22))|11065404||ovftdb|00005768-199010000-00022SL0000576819902269011065404P81[Full Text]|00005768-199711000-00022#xpointer(id(R20-22))|11065405||ovftdb|00005768-199010000-00022SL0000576819902269011065405P81[Medline Link]|00005768-199711000-00022#xpointer(id(R21-22))|11065405||ovftdb|SL00000429198512240011065405P82[Medline Link]|00005768-199711000-00022#xpointer(id(R25-22))|11065213||ovftdb|00005768-199507000-00017SL00005768199527107111065213P86[CrossRef]|00005768-199711000-00022#xpointer(id(R25-22))|11065404||ovftdb|00005768-199507000-00017SL00005768199527107111065404P86[Full Text]|00005768-199711000-00022#xpointer(id(R25-22))|11065405||ovftdb|00005768-199507000-00017SL00005768199527107111065405P86[Medline Link]|00005768-199711000-00022#xpointer(id(R27-22))|11065213||ovftdb|00005768-199403000-00016SL0000576819942637611065213P88[CrossRef]|00005768-199711000-00022#xpointer(id(R27-22))|11065404||ovftdb|00005768-199403000-00016SL0000576819942637611065404P88[Full Text]|00005768-199711000-00022#xpointer(id(R27-22))|11065405||ovftdb|00005768-199403000-00016SL0000576819942637611065405P88[Medline Link]|00005768-199711000-00022#xpointer(id(R30-22))|11065405||ovftdb|SL0000042019915383911065405P91[Medline Link]|00005768-199711000-00022#xpointer(id(R31-22))|11065213||ovftdb|SL0000696619906114611065213P92[CrossRef]|00005768-199711000-00022#xpointer(id(R31-22))|11065405||ovftdb|SL0000696619906114611065405P92[Medline Link]|00005768-199711000-00022#xpointer(id(R33-22))|11065213||ovftdb|SL0000684119933029611065213P94[CrossRef]|00005768-199711000-00022#xpointer(id(R33-22))|11065405||ovftdb|SL0000684119933029611065405P94[Medline Link]|00005768-199711000-00022#xpointer(id(R37-22))|11065213||ovftdb|00005768-199010000-00023SL0000576819902269811065213P98[CrossRef]|00005768-199711000-00022#xpointer(id(R37-22))|11065404||ovftdb|00005768-199010000-00023SL0000576819902269811065404P98[Full Text]|00005768-199711000-00022#xpointer(id(R37-22))|11065405||ovftdb|00005768-199010000-00023SL0000576819902269811065405P98[Medline Link]|00005768-199711000-00022#xpointer(id(R38-22))|11065213||ovftdb|00005768-198606000-00001SL0000576819861825311065213P99[CrossRef]|00005768-199711000-00022#xpointer(id(R38-22))|11065404||ovftdb|00005768-198606000-00001SL0000576819861825311065404P99[Full Text]|00005768-199711000-00022#xpointer(id(R38-22))|11065405||ovftdb|00005768-198606000-00001SL0000576819861825311065405P99[Medline Link]|00005768-199711000-00022#xpointer(id(R39-22))|11065213||ovftdb|SL0000277919906894111065213P100[CrossRef]|00005768-199711000-00022#xpointer(id(R39-22))|11065405||ovftdb|SL0000277919906894111065405P100[Medline Link]|00005768-199711000-00022#xpointer(id(R40-22))|11065405||ovftdb|SL000069661994658411065405P101[Medline Link]|00005768-199711000-00022#xpointer(id(R42-22))|11065213||ovftdb|00005768-198906000-00019SL0000576819892133811065213P103[CrossRef]|00005768-199711000-00022#xpointer(id(R42-22))|11065404||ovftdb|00005768-198906000-00019SL0000576819892133811065404P103[Full Text]|00005768-199711000-00022#xpointer(id(R42-22))|11065405||ovftdb|00005768-198906000-00019SL0000576819892133811065405P103[Medline Link]|00005768-199711000-00022#xpointer(id(R43-22))|11065213||ovftdb|SL0000696619906111411065213P104[CrossRef]|00005768-199711000-00022#xpointer(id(R43-22))|11065405||ovftdb|SL0000696619906111411065405P104[Medline Link]|00005768-199711000-00022#xpointer(id(R45-22))|11065213||ovftdb|SL00005077198942116111065213P106[CrossRef]|00005768-199711000-00022#xpointer(id(R45-22))|11065405||ovftdb|SL00005077198942116111065405P106[Medline Link]2585007Relationships between TriTrac-R3D vectors, heart rate, and self-report in obese childrenCOLEMAN, KAREN J.; SAELENS, BRIAN E.; WIEDRICH-SMITH, MARGO D.; FINN, JEREMY D.; EPSTEIN, LEONARD H.Special Communications: Methods1129