Oxygen uptake dynamics during high-intensity exercise in children and adults.

We hypothesized that the O2 uptake (Vo2) response to high-intensity exercise would be different in children than in adults. To test this hypothesis, 22 children (6-12 yr old) and 7 adults (27-40 yr old) performed 6 min of constant-work-rate cycle-ergometer exercise. Sixteen children performed a single test above their anaerobic threshold (AT). In a separate protocol, six children and all adults exercised at low and high intensity. Low-intensity exercise corresponded to the work rate at 80% of each subject's AT. High-intensity exercise (above the AT) was determined first by calculating the difference in work rate between the AT and the maximal Vo2 (delta). Twenty-five, 50, and 75% of this difference were added to the work rate at the subject's AT, and these work rates were referred to as 25% delta, 50% delta, and 75% delta. For exercise at 50% delta and 75% delta, Vo2 increased throughout exercise (O2 drift, linear regression slope of Vo2 as a function of time from 3 to 6 min) in all the adults, and the magnitude of the drift was correlated with increasing work rates in the above-AT range (r = 0.91, P less than 0.0001). In contrast, no O2 drift was observed in over half of the children during above-AT exercise. The O2 drifts were much higher in adults (1.76 +/- 0.63 ml O2.kg-1.min-2 at 75% delta) than in children (0.20 +/- 0.42, P less than 0.01).(ABSTRACT TRUNCATED AT 250 WORDS)

ercise are regulated in healthy children to maintain cellular homeostasis in a period characterized by rapid change in body size and organ development (7). Although the maximal 0, uptake (Vo2 ,,) and anaerobic threshold (AT) change in almost direct proportion with body weight, other responses, such as the time course of the rise in 0, uptake (VO,) during moderate-intensity (below-AT) constant-work-rate exercise, are the same in children and young adults(g). The dynamics of ventilation (VE) and CO, output (VCO,) after the onset of moderate-intensity exercise are faster in children than in adults (lo), but the differences in VE and VCO, are not nearly as great as the discrepancy in age and body size.
The mechanisms of these growth-related differences are not fully understood but may result from factors such as differences in CO, storage capacity and cellular metabolic adjustment to sudden increases in energy demand (10,29). These responses are particularly important during high-intensity (above-AT) exercise when the relationship between 0, delivery and tissue metabolic demand reaches physiological limitations.
The goal of this study was to determine whether growth-related differences in the VO, response existed for work performed in the high-intensity range. Finding such differences could be helpful in determining the mechanism of maturation of cellular energy metabolism.
Previous studies of the 00, response to constantwork-rate low-intensity exercise have demonstrated that the steady-state VO, is achieved by 3 min (-6 time constants) (16). However, when adults perform constantwork-rate exercisein the high-intensity range, the difference between the VO, at 6 and 3 min is positive, indicating that the attainment of a steady state is delayed if it is ever achieved (3, 23). The high correlation between lactate concentration and the magnitude of the 0, drift in adults (as indicated by the slope in the vo2 response between 3 and the 0, drift (3, 23). 6 min was, of in exercise) pro lmpted speculation that fact, related to lactate metabolism Whether an 0, drift existed in children-was not known, but there were reasons to speculate that VO, responses to high-intensity exercise in children would be different from those in adults. Children reach lower lactate concentrations and are less able to sustain exercise in the high-intensity range than are adults (1). These observations suggest that the magnitude of the 0, drift would likely be lower in children than in adults. This hypothesis was tested by examining the dynamic VO, response to exercise in a group of adults and children. Studies were done ove r a range of wo Irk rates from below to abov 'e the subjects' AT. Althou & methodological constraints prohibit direct measurements of cellular energy metabohsm in children (such as might be obtained by muscle biopsy), there is growing evidence that gas exchange measured at the mouth indirectly reflects cellular function (5,8,27).

Study Population
Twenty-two healthy children [12 females and 10 males, age range 6-12 yr, mean 10.0 t 2.2 (SD); mean 0161-7567/91 $1.50 Copyright 0 1991 the American Physiological Society 841 body weight 33.0 + 11.1 kg] and seven healthy adult males (27-40 yr of age, mean 32 t 5; mean body weight 70 t 8 kg) participated in the study. Informed consent was obtained from each subject or, when appropriate, a parent.

Protocols
The first protocol consisted of an incremental cycleergometer exercise. Each subject performed a progressive exercise test to the limit of tolerance by use of a ramp pattern of increasing work rate as described previously (11,32). This was used to determine the peak VO, or vo 2 mar and. the AT. The AT corresponds to the metabolic rate (VO,) above which anaerobic metabolism supplements aerobic energy production. Lactic acidosis then occurs, and its noninvasive determination has been previously described (30,31). The difference between work rates corresponding to the AT and to the Vo2,, (A) was also determined from the ramp protocols.
The second protocol consisted of constant work rate exercise. Sixteen children performed a single ergometer exercise test. Our goal was to select a work rate that was above the AT corresponding to 50% of the difference between V02max and the AT (e.g., 5O%A means the sum of the work rate at the AT and 50% of the work rate difference between the AT and VO, ,,). However, we found that some children were unable to complete 6 min of exercise at this work rate, and in others the lack of an apparent slope of VO, led us to believe that the chosen work rate was not sufficiently high for the particular subject. (Six minutes of exercise is about the limit of tolerance for many subjects exercising in the high-intensity range but still provides sufficient data for linear regression analysis.) In these cases, the children repeated the studies at appropriately higher or lower work rates and only the repeated data were included in the study. All in all, work rates ranged from 27 to lOO%A in these 16 children.
Based on the analysis of the initial 16 children, we then studied individual adults and children who performed constant-work-rate ergometer exercise at a series of exercise intensities. All the adults and six additional children (age range 6-12; 5 males) performed constantwork-rate ergometry for 6 min corresponding to 80% AT and to 25%A, 5O%A, and 75%A. For example, if the subject's AT corresponded to 100 W and if the A was 100 W, then the 25%A work rate would be chosen as 125 W. Not all subjects performed all four protocols. A 3-min period of unloaded pedaling preceded the constant-work exercise.

Gas Exchange Measurement
The subjects breathed through a mouthpiece connected to a low-impedance turbine volume transducer and a breathing valve with a combined dead space of 90 ml. 0, and CO, tensions were determined by mass spectrometry from a sample drawn continuously from the mouthpiece at 1 ml/s. The inspired and expired volumes and gas tension signals underwent analog-to-digital con- 1. 90, response to high-intensity exercise (5O%A) in a representative adult and child. A 6-min constant-work-rate exercise started at time 0. Slope was calculated using linear regression techniques to correlate VO, with time between 3rd and 6th min of exercise.
previously described (2). The breath-by-breath data were then interpolated to l-s time intervals.

Data Analysis
Normalization. To compare VO, responses of different-sized subjects, we used two strategies.
WORK RATES. The effort of each subject was scaled to his/her metabolic capability by using the physiological landmarks AT and V02max rather than absolute work rate. We used the work intensity to connote the relative work rate as %AT or %A. In addition, the work rate divided by body weight provided a numerical means to compare the work performed by adults and children.. Thebreath-by-breathV0, data were divided by body weight for each subject. 0, drift. We used linear regression techniques to correlate VO, with time between the 3rd and 6th min of exercise. The slope of the best-fit line was considered to be positive if the 95% confidence interval was >O. The slopes were normalized to body weight (ml l minh2. kg-l), and the normalized slopes were defined as the 0, drift. Typical responses to high-intensity constant-work-rate exercise in an adult and child are shown in Fig. 1. As can be seen in these examples, the linear model was a good description of the Vo2 of the exercise study. 0 data2 subject at difference cost. To calculate the 0, cost, the breath-by-breath were scaled to work performed (in watts) for each response in the 3-to 6-min portion each work ra between VO, . te. This was done by finding the at time t during exercise and the preexercise VO, (baseline, unloaded pedaling). This value was then divided by the work rate (in watts). Careful calibration of our ergometers re vealed that unl .oaded cycling represented 12 W for the adults' ergometer and 7 W for the children's ergometer. Thus these respective values were subtracted from the constant work rate value to correctly describe the Avo2-A watt relationship. To exclude mechanical differences between the two ergometers that might contribute to differences in the responses between the two groups, one adult subject repeated the exercises on the children's ergometer at work rates below and above the AT. Analysis of his data showed virtually no difference between the two ergometers in the subject's VO, response.
Muthematical modeling of responses. In addition to the linear regression analysis of the VO, response from 3 to 6 min of constant exercise, we attempted to characterize the whole response (i.e., from the onset of exercise through the 6th min). When exercise is performed below a subject's AT, the VO, response can be modeled by a single-exponential (1) where net ho,(t) is the increase in VO, above the baseline (during unloaded pedaling) at any given time (t) after * where s represents the coefficient of the linear term. Nonlinear iterative curve-fitting techniques were used to calculate the parameters of this model (i.e., 7 and s) for each subject at each work rate (3). In addition to the analysis of each curve for each subject, we used the noise-reducing technique of superimposing and averaging values. In this manner we could graphically display the differences between adults and children for a given work intensity (i.e., 80%AT, 25%A, 5O%A, and 75%A).
The finding of a linear component for high-intensity exercise complicates the use-of 7 to compare the temporal response dynamics of TO, in children and adults. If an asymptote (steady state) of VO, is not achieved, selection of an end point for determining temporal responses is necessarily arbitrary. In the particular case of the VO, cost for the highest-intensity work rates where both adults and children reached the same value by the 6th min of exercise, we calculated the half time (t1/2) of the response (i.e., the time required to achieve one-half of the baseline to end-exercise value). The t,,, was then used to compare the overall speed of response between the adults and children in this case.

Statistical Analysis
For the 16 children who performed a single high-intensity exercise test, linear regression was used to determine whether there was a correlation between the work rate (as work rate per kilogram) and the magnitude of the 0, drift (as the slope of the best-fit line). In the remaining children and adults who completed the several work rates, analysis of variance and modified t tests were used to compare the effect of different work intensities within each group and the differences between the groups. To compare the group mean average curves, we used nonparametric analyses. Results are presented as means t SD.

RESULTS
The VO 2 max per kilogram body weight in children (43 t 6 ml 0, l min-' . kg-') was not statistically different from that in adults (40 t 10 ml 0,. min-' . kg-'). The average work rate per kilogram body weight in those children who performed the serial protocol was 0.83 t 0.12,1.60 t 0.09, 1.78 t 0.27, and 2.02 t 0.31 W/kg for 80%AT, 25%A, 5O%A, and 75%A, respectively. In adults these values were 0.84 t 0.26,1.52 t 0.37,2.02 t 0.38, and 2.49 t 0.52 W/kg, corresponding to 80%AT, 25%A, 5O%A, and 75%A. In comparing adults and children, a statistically significant difference in work rate per kilogram was found only at the highest work intensity (P < 0.05). We used the chance observation that 5O%A in adults represented precisely the same work rate increment as 75%A in children (2.02 W/kg) and constructed a graph showing the VO, normalized per kilogram for the two groups at these respective work intensities (Fig. 2).

O2 Drift
In adults, positive slopes were found in some subjects at all work rates (in 57% of the adults at 80% AT, in 83% at 25%A, and in 100% at 5O%A and 75%A). The slopes increased with increasing work intensities in adults (Fig.  3). The difference between the mean slope at 80% AT (0.17 t 0.11 ml 0, l kg-l. min2) was not significantly less than the slope at 25%A (0.48 t 0.42). However, the 25%A value was significantly less than both 50%A (1.27 t 0.50, P < 0.05) and 75%A (1.76 t 0.63, P < 0.01). In addition, linear regression of above-AT slopes as a function of work rate per kilogram revealed a correlation coefficient of 0.91 (P < 0.0001).
In the children who performed exercise at the four work rates, positive slopes were found in half of them at the 80% AT work intensity, in 75% at 25%A, in 50% at 5O%A, and only in 25% at the highest work intensity. There were no differences between the slopes at 80% AT in the 16 children who performed a single high-intensity test.
Analysis of variance and modified t tests revealed marked differences between the adults and children. At both 75%A and 5O%A the slopes in adults were significantly greater than in children (P < 0.01). The difference in the VO, response between children and adults is demonstrated graphically in Fig. 4, showing group mean responses of VO, per kilogram for the lowest (80% AT) and highest (75%A) work intensities.
The mean values were not affected by the work intensity in either the children or the adults.
In addition, the pattern of the responses was different (Fig. 5). During high-intensity exercise, the 0, cost of exercise was lower in adults during the first 3 min of exercise, but by the last 3 min, there was no statistical difference between the two groups (12.56 t 1.33 ml 0, l min-' l W-l for 75%A in children compared with 11.4 t 1.30 in adults). Statistical analysis of the group mean curves revealed significant differences between children and adults at all four work intensities (P < 0.0001). For the highest work intensity (75%A, where both groups reached the same value by the 6th min of exercise), the t,,, was 20 s in children and 45 s in adults (Fig. 5). Finally, we compared the mean value of the 0, cost of the last minute of exercise at the different work intensities. Analysis of variance and modified t tests revealed that in adults the 0, cost was significantly higher at the end of the highest-intensity exercise (12.4 t 1.38 ml 0, l min-l l W-l) than at the end of the 80% AT work rate (10.88 t 2.25, P < 0.05). In children, the last-minute 0, cost of the highest-intensity exercise (12.68 t 1.52 ml 0,. min-l . W-l) was not statistically different from that of the below-AT exercise (13.12 + 1.30). For both low-and high-intensity exercise, average O2 cost was significantly higher in children. Note difference in response time between the 2 groups at high-intensity exercise. Adults achieve the children's value only toward the end of exercise. During the last minute of exercise, O2 cost in adults was significantly higher at 75%A than at 80% AT exercise.

Modeling
The pattern of the 00, response to 6 min of exercise in children was different from that in adults. Although a single-exponential function could be used to accurately describe the low-intensity as well as the above-AT exercise in children (Fig. 6), the monoexponential model was not an appropriate description of the response in adults for most work rates above the AT. The addition of a linear term (Eq. 2), however, provided a good fit for the VO, response in the high-intensity range in adults (Fig.  7). In fact, when a model consisting of the sum of two exponentials was tried for the group mean responses, the time constant for the second-exponential component became so large that it acted virtually as a linear term.
The value of the linear coefficient (Eq. 2) correlated well with the slope of the 0, drift (r = 0.94, P < 0.0001).
The value of the linear term was 0 in all adults in the 80% AT work intensity, while positive values were found for the above-AT protocols in all but one subject (this at 25%A). In children, the linear term was 0 for all below-AT protocols but was positive in 73% of the above-AT exercises. Finally the linear coefficients in children for all above-AT exercise intensities were significantly lower than the mean values in adults (children: mean 0.24 ml 0, l kg-l l minV2, range 0.21-0.55; adults: mean 0.97, range 0.27-1.73; P < 0.01). DISCUSSION These data demonstrate that the pattern of the $70, response to high-intensity exercise in children is qualitatively and quantitatively different from that in adults. Fewer children than adults develop an 0, drift during high-intensity exercise, and in children the magnitude of the drift both in absolute terms and when normalized to body weight is smaller and not correlated with work intensity. However, despite a smaller or absent 0, drift, the 0, cost of exercise (as vo2 per watt) is greater in children than in adults. Finally, children demonstrate faster ad- justments of VO, during high-intensity exercise than do adults.
Scaling metabolic rate to body size is of critical importance when attempting to assess growth of the cardiorespiratory system (7,15,19). To determine whether a particular response truly differs between adults and children, it is first necessary to minimize the effects of size alone on the response in question. For example, metabolic rates (VO,) in adults and children are determined in large part by the mass of metabolizing tissue. Although direct noninvasive measurements of muscle mass are not available, body weight is an accurate and easily obtained index. Some aspects of physiological function in mature mammals of different sizes are known to scale to body weight to a power cl.0 by use of the allometric equation (19) metabolic function cc body massb (3) where b is the scaling factor. However, our studies of exercise responses in children demonstrated the scaling factor to be 1.01 (95% confidence interval, 0.90-1.11) for the VO, max and 0.92 (95% confidence interval, 0.80-1.02) (11). The data cited above support our use of unscaled body weight (i.e., massl) and suggest that growth in children is a unique case of the relationship between size and function in biology. The relationship between body size and metabolic function during growth in a species may not be regulated by the same mechanisms that apply to the relationship between size and function in mature animals of different species. Moreover, even if our data had been scaled to a factor ~1, the effect would be to amplify the differences in drift between adults and children.
A number of different mechanisms have been proposed to explain the 0, drift. The increasing VO, could represent the increased energy expenditure associated with dissipation of heat generated during exercise (15). The larger surface area-to-body mass ratio in children than in adults would facilitate passive heat exchange to a greater extent in children and might have contributed to a relatively smaller increment in the temperature-related \io,.
However, recent studies in adults demonstrated that fitness training lowered the magnitude of the 0, drift without affecting the increase in core temperature during exercise (4). These investigators concluded that heat exchange was not a major determinant of the 0, drift in high-intensity exercise. Alternative explanations center on the high correlation between the increase in lactate concentration during exercise and the magnitude of the 0, drift. Current theory holds that the rate of ATP degradation with increasing concentration of ADP in contracting muscle cells stimulates mitochondrial oxidative phosphorylation (5). For low-intensity exercise, 0, is sufficiently available so that ATP regeneration is supplied solely by oxidative phosphorylation.
Consequently, VO, and concentrations of ADP, PCr, Pi, and 0, at the cellular level reach a new steady state. During heavy exercise, insufficient availability of cellular 0, leads to anaerobic regeneration of ATP and net lactate release. Increasing lactate concentrations could stimulate metabolism of lactate to glucose via the Cori cycle (an O,-dependent process) or the oxidation of lactate in muscle cells (18,30). Lactate levels are known to be lower at comparable levels of work intensity in children than in adults (13,22). Thus, our finding of smaller 0, drifts in children is consistent with the hypothesis that the 0, drift reflects metabolism of lactic acid.
The 0, cost measured in the present study represents the integrated response of the organism to a known increase in the cellular energy requirement.
Thus, any observed differences in the 0, cost between adults and children or among the different work rates reflect differences in the way the cellular energy requirement is supplied. In adults, we observed the tendency [also seen previously (14)] for higher 0, costs of high-vs. low-intensity exercise. The drift in Vo2 during heavy exercise did not simply reflect a longer time required by the organism to achieve the same value of VO, per watt reached at lower work rates. Rather, by the 6th min of exercise, adults were utilizing more VO, per watt during heavy exercise than they did during below-AT work.
The 0, cost pattern was different in children. As noted, the values in children were higher than in adults at all work rates, and children demonstrated less tendency to increase VO, during high-intensity exercise. It was only by the last minute of the heaviest exercise that adults and children achieved comparable 0, costs. In a previous study, we calculated the work efficiency from the slope of the relationship between VO, and work rate measured during progressive exercise tests (ramp-type protocols) (II), and we found a small but statistically significant decrease in the 0, cost (measured as an increase in the work efficiency) with increasing body weight in girls. There was a small decrease of 0, cost with increasing weight in boys as well, but the slope of the regression was not statistically significant. The discrepancy in results could be due to several factors: first, in our previous study the 0, cost was measured from ramp protocols and not from constant-work-rate tests. Nonlinearities in the nature of the &work rate relationship could affect the resulting calculation of 0, cost. In addition, the design of the present study (i.e., comparing a group of younger subjects with a group of older ones rather than a continuum of ages and body sizes) was fundamentally different from that of the previous study. The current design would make it easier to detect subtle differences in 0, cost between adults and children.
Do higher values of vo2 per watt indicate a more effective cardiorespiratory response to exercise in children than in adults, or, alternatively, do the higher values result from a less developed ability of children to support anaerobic ("O,-sparing") mechanisms of ATP metabolism? Bar-Or and others (1, 12, 22) have suggested that children are less able than adults to sustain exercise in the supramaximal range, indicating a reduced anaerobic "potential" in children. Lower lactate levels found during exercise in children are consistent with relatively less ability to utilize anaerobic glycolytic processes for ATP turnover. In one investigation of glycolytic enzymes in children during exercise (13), lower levels of phosphofructokinase were found in 11-to 13-yr-old children than in young adults.
In contrast to this hypothesis of reduced anaerobic po-tential in children, lactate levels alone cannot be used to determine overall lactate metabolism, because blood lactate concentration is determined by the balance between production and removal. Children may metabolize lactate more quickly than adults. The latter mechanism is not inconsistent with our results. The higher VO, per watt seen in children at all work intensities could indicate a greater ability to oxidize the lactate produced during exercise. In this context, it is noteworthy that the neonatal heart muscle is less sensitive to hypoxia than the adult myocardium, specifically because of the neonates' greater capacity for anaerobic glycolysis and more rapid removal of lactate (6, 28).
The time constant of the exponential term (Eq. 2) was not affected by the work intensity in either children or adults, but the linear coefficient increased with increasing work rates in adults. This suggests that the underlying physiological adjustments of VO, responsible for the exponential phase of the response were independent of the magnitude of the input. Significant differences in time constants were observed between the adults and children at both low-and high-intensity exercise.
[This was somewhat surprising to us because in two previous independent studies we observed that the time constants for VO, for below-AT exercise were slightly smaller in children than in adults but not significantly so (9, 26). Methodological differences in the data analysis and population (subjects' age) between the previous and present studies may have contributed to the different results.] The faster kinetics in children during high-intensity exercise may represent less dependence on anaerobic responses early in exercise in children than in adults. Alternatively, there is evidence that VO, kinetics are faster after fitness training, presumably because of the increase in mitochondrial density and capillarization that occurs in muscles (17,20,24). The VO, max per kilogram was the same in adults and children, but to the extent that "fitness" can be gauged by faster kinetics, children may be generally more fit than adults.
Our data demonstrate that the response to short periods of constant-work-rate exercise can be modeled in a simple way by using the sum of an exponential and linear term. The coefficient of the linear term correlated well with the magnitude of an independently measured slope of the 0, drift in both adults and children. This approach may be useful in a number of clinical instances where both the exponential and linear components of the response may be affected. For example, in adult patients with severe cyanotic congenital heart disease, the dynamic Vo2 response is markedly slowed (25). Similarly, the irO, time constant is prolonged in patients with chronic obstructive lung disease (21). This approach could prove useful for investigations of growth and development in healthy children, as well as in a variety of childhood diseases, as a tool to gauge pathophysiological responses.
This study adds to the growing body of evidence demonstrating maturation of energy metabolism during growth in children. We speculate that the mechanism of the greater 0, cost in children than in adults reflects the kinetics of high-energy phosphate metabolism in muscle cells. Noninvasive techniaues. such as magnetic reso-nance spectroscopy, may prove useful in resolving the development of the "anaerobic potential" that occurs during growth in humans.