Comparing several equations that predict peak VO 2 using the 20-m multistage-shuttle run-test in 8 10-year-old children

Transcription

1 DOI /s z ORIGINAL ARTICLE Comparing several equations that predict peak VO 2 using the 20-m multistage-shuttle run-test in 8 10-year-old children Xavier Melo Helena Santa-Clara José Pedro Almeida Elvis Alvarez Carnero Luís Bettencourt Sardinha Paula Marta Bruno Bo Fernhall Accepted: 20 October 2010 Springer-Verlag 2010 Abstract This study compared the validity of reported equations as predictors of peak VO 2 in 8 10-year-old children. Participants (90 boys and girls aged 8 10 years) performed the multistage-shuttle-run-test (MSRT) and peak VO 2 was measured in Weld using a portable gas analyser. The equations that estimated peak VO 2 from the MSRT performance were chosen according to the age range of this study. As follows, the FITNESSGRAM reports and the equations of Leger et al. (Can J Appl Sport Sci 5: 77 84, 1988), Barnett et al. (Pediatr Exerc Sci 5:42 50, 1993), Matsuzaka et al. (Pediatr Exerc Sci 16: , 2004) and Fernhall et al. (Am J Ment Retard 102: , 1998) were used to estimate the peak VO 2 and compared with the directly measured value. The equation of Leger et al. (Can J Appl Sport Sci 5: 77 84, 1988) provided a mean diverence (d) of 4.7mlkg 1 min 1 and a 1.0 slope. The equation of Matsuzaka et al. (Pediatr Exerc Sci 16: , 2004)(a) using maximal speed (MS) showed a higher d (5.4) than the remaining using total laps d (4.2). The equation of Barnett et al. (Pediatr Exerc Sci 5:42 50, 1993)(a) that includes triceps skinfold and MS showed the highest d (6.1) but the Communicated by Klaas Westerterp. X. Melo (&) H. Santa-Clara J. P. Almeida E. A. Carnero L. B. Sardinha P. M. Bruno Faculdade de Motricidade Humana, Estrada da Costa Cruz Quebrada, Dafundo, Lisboa, Portugal xaviermelo@esdrm.ipsantarem.pt H. Santa-Clara santaclara@fmh.utl.pt B. Fernhall College of Applied Health Sciences, Louise Freer Hall, 906 South Goodwin Avenue, Urbana, IL , USA smallest range (24.1) and slope (0.6). Data from the FITNESSGRAM had the smallest d (1.8 ml kg 1 min 1 ), but also had the highest range between limits of agreement (28.6 ml kg 1 min 1 ) and a 1.2 slope. The lowest slope (0.4) and range (22.2 ml kg 1 min 1 ) were observed using the equation of Fernhall et al. (Am J Ment Retard 102: , 1998). Log transformation of the data revealed that the equations of Matsuzaka et al. (Pediatr Exerc Sci 16: , 2004)(a) (1.1*/ 1.25) and Fernhall et al. (Am J Ment Retard 102: , 1998) (1.17*/ 1.25) showed the closest agreement among all, but they still yield unsatisfactory accuracy. Keywords Directly measured peak oxygen uptake Estimated peak oxygen uptake Portable gas analyser Field exercise testing Maximal heart rate Abbreviations BM Body mass BMI Body mass index d Mean diverence (ml kg 1 min 1 ) FFM Fat-free mass HR Heart rate HR max Maximal heart rate LoA Limits of agreement MRI Magnetic resonance imaging MS Maximal speed 20-m MSRT 20-m Multistage-shuttle-run-test RER Respiratory exchange rate TL Total number of laps TT Treadmill test TricSKF Triceps skinfold VO 2 Oxygen uptake VO 2max Maximal oxygen uptake Vol Lower leg muscle volume

2 Introduction Because cardiovascular Wtness is strongly associated with the clustering of cardiovascular disease risk factors in children and youth (Anderssen et al. 2007), simple non-invasive, valid and reliable Weld tests providing an estimate of peak VO 2 have been developed for children. Such tests require the subject to either cover as much distance as possible in a set time (Cooper 1970), cover a set distance in the fastest time possible (Kline et al. 1987) or perform a 20-m multistage-shuttle-run-test (MSRT) (Leger and Lambert 1982). These running tests are maximal from the onset and require well-motivated subjects with some knowledge of pace judgment and understanding of the test requirements (Mahoney 1992; Ramsbottom et al. 1988; van Mechelen et al. 1986). The 20-m MSRT has some advantages such as a graded physiological response and the absence of individual pace control (Paliczka et al. 1987; Ramsbottom et al. 1988). It can be administered in a relatively small space either indoors or outdoors and is therefore practical for schools without extensive facilities. In addition, children are known for their frequent stops, starts, and turns in their daily physical activity, therefore, it may be a more speciwc test than a continuous directional run (Liu et al. 1992). The 20-m MSRT provides a valid estimation of peak VO 2 for children and youth (Barnett et al. 1993; Boreham et al. 1990; Falgairette et al. 1994; Fernhall et al. 1998; Leger et al. 1988; Liu et al. 1992; Mahoney 1992; McVeigh et al. 1995; Pitetti et al. 2002; van Mechelen et al. 1986). There are several issues in predicting peak VO 2 from the 20-m MSRT. Some suggest including skinfold measurements in the estimation of peak VO 2 from the 20-m MSRT to improve the predictive power of the test (Barnett et al. 1993; McVeigh et al. 1995). Others suggest gender and age avect the predictive power of the 20-m MSRT. For instance, Leger and Gadoury (1989) in a sample of year old males and 24 females, and Leger and Lambert (1982) using 91 adults (32 females and 59 males, aged 27 and 25 years, respectively), showed that gender had little evect on predicting peak VO 2 estimated by the retroextrapolation method (extrapolation to time zero of recovery of the exponential least squares regression of the Wrst four 20-s recovery VO 2 values). However, Barnett et al. (1993) studying 55 Hong Kong Chinese students (27 boys and 28 girls, ages years) found that an inclusion of gender in the regression equations used to predict peak VO 2 determined in laboratory improved the accuracy of prediction. Boreham et al. (1990) indicated a higher correlation between the 20-m MSRT and peak VO 2 in a group of 24 schoolgirls (15.4 years) than in a group of 24 schoolboys (15.6 years) (r =0.90 vs. r = 0.64, respectively), and McVeigh et al. (1995) found that the plot of peak VO 2 measured in laboratory against the 20-m MSRT performance showed distinct gender groups and divergent regression lines for boys and girls (15 boys, 18 girls). Barnett et al. (1993) and Leger et al. (1988) also identiwed age as a critical variable. However, it is unclear which regression equation is most accurate when used in the same population of children. Whilst a recent study by Ruiz et al. (2009) has addressed the criterion validity of the 20-m MSRT to predict peak VO 2 in year olds, it is currently unclear which equation is most appropriate for younger children. To our knowledge, this is the Wrst study to assess the cardiovascular Wtness by the performance in the 20-m MSRT using a portable gas analyser in 8 10-year-old children. This study compared the validity of previously reported equations as predictors of peak VO 2 in 8 10-year-old children. The estimated peak VO 2 provided from several 20-m MSRT equations for children and youth were chosen according to the age range of this study. As follows, the Wtness reports of FITNESSGRAM and the equations of Leger et al. (1988), Barnett et al. (1993), Matsuzaka et al. (2004) and Fernhall et al. (1998) were used to estimate the peak VO 2 and then compared with the directly measured value. Methodology Participants Ninety children, 45 boys and 45 girls, aged 8 10 years from an elementary school of the Lisbon district were selected to participate. Based on their medical history, none of the participants had cardiovascular or metabolic diseases that limited their aerobic performance. The ethics committee of the university approved the study and children gave verbal consent whilst their parents or guardians provided written consent. Anthropometric measures Height was measured using a tape measure to the nearest 0.1 cm. Light upward pressure was applied under the jaw and occipital bone with two hands to provide maximum extension of the spine. Weight was measured using a standard scale (Seca, Hamburg) to the nearest 100 g. Triceps and calf skinfolds were measured using a skinfold caliper, the triceps skinfold (TricSKF) was measured on the back of the right arm over the triceps muscle, midway between the elbow and the acromion process of the scapula (The Cooper Institute for Aerobics Research 1994). The calf skinfold was measured on the inside of the right leg at the level of maximal calf girth. Each skinfold was measured three times

3 and the median value was recorded and used in the FIT- NESSGRAM software. 20-m MSRT test protocol The 20-m MSRT (Leger and Lambert 1982; Leger et al. 1988) is progressive in intensity, it is easy at the beginning (8.0 km h 1 ) and then speed is increased 0.5 km h 1 every stage after the Wrst minute. The progressive nature of the test provides a built-in single beep at the end of the time for each lap and a triple beep at the end of each minute. The test audio CD was calibrated previously to each trial and was consistently within the 0.5 s interval. The protocol was conducted in an outdoor playground of a school of the Lisbon district during the summer months. The test was clearly explained to each participant and all participants had also performed the test in their physical education classes. The test was terminated when a student failed for the second time to reach the 20 m line in the time allotted (by the beep). The recorded score was the total number of laps (TL) completed by the participant. Oxygen uptake measures Consistent with the original development of the 20-m MSRT (Leger and Lambert 1982) and for greater ecological validity, we measured peak VO 2 directly whilst the participants performed the 20-m MSRT. Oxygen uptake can be directly measured whilst performing Weld tasks, such as running, which allows for the more direct measurement of a number of physiological variables (Ruiz et al. 2009). Although Leger and Lambert (1982) used a retroextrapolation method, previously been found to be valid (r = 0.92) and accurate (SEE = 3.21 ml kg 1 min 1 ) for the determination of VO 2 max after maximal 20-m MSRT (Leger and Boucher 1980), we analysed inspired and expired gases continuously in Weld, breath by breath, through a portable gas analyser (K4b 2, Cosmed, Rome, Italy), which had been previously validated (McLaughlin et al. 2001). Before each test, the O 2 and CO 2 analysers were calibrated using ambient air and standard calibration gases of known concentration (16.7% O 2 and 5.7% CO 2 ). The calibration of the turbine Xowmeter of the K4b 2 was performed using a 3-l syringe (Quinton Instruments, Seattle, WA, USA) according to the manufacturer s instructions. Heart rate was continuously monitored (Polar T31 coded TM transmitter). This compact device was easy to attach without constricting the children s movements. The K4b 2 weights 475 g and was not expected to signiwcantly avect the energy demands of the subjects (Flouris et al. 2005), and none of the subjects had any negative remarks on the system s weight or on their mobility and vision during the test. The 20-m MSRT performed was considered a maximal evort when at least two of the following criteria were obtained: (1) voluntary exhaustion; (2) a peak HR greater than 90% of the maximal HR predicted (220-age); (3) a peak RER greater than 1.0; or (4) subjective opinion of experienced testers (Barnett et al. 1993). Estimation of peak VO 2 Prediction equations for peak VO 2 from the 20-m MSRT in children and youth are presented in Table 1. These equations were chosen among those whose age range of the participants of its original article was most similar to that of the present study. To calculate peak VO 2 from the 20-m MSRT, weight, height, triceps and calf skinfold were introduced in the FITNESSGRAM in order to allow an automatic estimation of peak VO 2. The last stage number completed was used to predict peak VO 2 (ml kg 1 min 1 ) from the speed (km h 1 ) corresponding to that stage using the estimation equation by Leger et al. (1988). In the equations of Barnett et al. (1993) and Matsuzaka et al. (2004), maximal speed (MS) was calculated from the total number of laps (TL) attained. In all other equations (Fernhall et al. 1998; Matsuzaka et al. 2004), the TL attained was used in the respective equations without further treatment. Statistical analysis Shapiro Wilk test was used to test the distribution normality. SigniWcant values were considered when P < The Friedman test was used to evaluate diverences among previously reported peak VO 2 predictive equations since the data were not normally distributed. If the Friedman test revealed signiwcant diverences between the equations, Dunnett s procedure was used to Wnd out which equations were diverent from measured directly peak VO 2 in this study. Independent Student s t test and the Mann Whitney test were used to compare mean anthropometric, spirometric and performance values between boys and girls when normal distribution was conwrmed, or not, respectively. We also evaluated mean bias (d) and limits of agreement (LoA) and slopes using Bland Altman plots. Large and signiwcant slopes are present in all the prediction equations presented in this study. Given that this systematic error is present, the constructed 95% LoA are diycult to interpret and potentially misleading. To this end, we followed the recommendations of Nevill and Atkinson (1997) and used log transformations to construct LoA in their ratio form. The statistical analyses were computed and analysed using the SPSS Statistics 17.0 and MedCalc

4 Table 1 Prediction equations for peak VO 2 from the 20-m MSRT in children and youths Author Participants Equation Leger et al. (1988) 200 M & F VO 2peak = (Speed) 3.248(Age) (Age)(Speed) 8 19 years Barnett et al. (1993) 27M+28F years (a) VO 2peak = (gender: M = 0; F = 1) 0.7(TricSKF) + 2.6(MS); (b) VO 2peak = (gender: M = 0; F = 1) 0.2(weight) + 3.2(MS); (c) VO 2peak = (gender: M = 0; F = 1) 0.8(age) + 3.4(MS) Fernhall et al. (1998) 13M+38F VO 2peak = 0.35(TL) 0.59(BMI) 4.61(gender: M = 1; F = 2) years Matsuzaka et al. (2004) 62M+70F (a) VO 2peak = (gender: M = 0; F = 1) (age) 0.831(BMI) (MS); 8 17 years (b) VO 2peak = (gender: M = 0; F = 1) 0.462(age) 0.862(BMI) (TL). Results Means and standard deviations (SD) of body composition, performance, metabolic and cardiovascular variables assessed during the 20-m MSRT are presented in Table 2. There were diverences between boys and girls in percent body fat (%Fat) (P < 0.001) and the number of laps completed (P < 0.001). Also, mean absolute (l min 1 ) and relative peak VO 2 (ml kg 1 min 1 ) values and mean peak RER were signiwcantly diverent between boys and girls (P <0.05). Mean diverences (d), LoA and range (upper LoA - lower LoA) are presented in Table 3. The FITNESSGRAM software reports yielded the smallest d (ml kg 1 min 1 ) between measured and estimated peak VO 2 values. However, it also had the highest range and the highest slope meaning that the equation over-predicted peak VO 2 in lower VO 2 participants and under-predicted peak VO 2 in participants with higher peak VO 2. The equation of Fernhall et al. (1998) yielded the lowest slope (P < ) and the lowest range. The equation of Leger et al. (1988) provided a reasonably large d (P < 0.05), one of the highest range and slope (P < ). The equation of Matsuzaka et al. (2004)(a) using MS showed a higher d (P < 0.05) than the Table 2 Means ( SD) of body composition, performance, metabolic and cardiovascular variables assessed during the 20-m MSRT Boys (n =45) Girls (n = 45) Total (n =90) BMI (kg/m 2 ) %Fat ** Total laps (no) ** Maximal speed (km h 1 ) ** VO 2peak (l min 1 ) * VO 2peak (ml kg 1 min 1 ) ** RER peak ** HR max (bpm) *DiVerence between the means of boys and girls (P <0.05); ** diverence between the means of boys and girls (P <0.001) other equation of Matsuzaka et al. (2004)(b) using TL (P < 0.05), however, the Wrst equation demonstrated a smaller range and slope (P < ). The equation of Barnett et al. (1993)(a) that includes the triceps skinfold and MS showed the highest d (P < 0.05) but the smallest range and slope (P < ). Only FITNESSGRAM software reports were not signiwcantly diverent from directly measured peak VO 2 (P >0.05). The Bland Altman plots are presented in Fig. 1. The plots provide a visual indication of both the systematic bias and the random error between predicted peak VO 2 and measured directly peak VO 2 in the present sample. There was a positive linear distribution of the values from the FIT- NESSGRAM software reports (Fig. 1a), in the equation of Leger et al. (1988) (Fig. 1b) and in the equation of Matsuzaka et al. (2004)(b) (Fig. 1h). A more random dispersion of scores are shown in the equation of Barnett et al. (1993)(a) (Fig. 1c), Fernhall et al. (1998) (Fig. 1f) and in the equation of Matsuzaka et al. (2004)(a) (Fig. 1g). The data from the Bland Altman approach based on gender are presented in Table 4. In girls, the FITNESS- GRAM software reports yielded the smallest d (ml kg 1 min 1 ) but a high slope meaning that the equation overpredicted peak VO 2 in most of the girls with lower peak VO 2 and under-predicted peak VO 2 in those with higher peak VO 2 girls. The lowest slope was yielded by the equation of Barnett et al. (1993)(c), though with a moderate d, but with a low range. In boys, the lowest d was seen using the equation of Barnett et al. (1993)(c), but had also the highest range and slope. The lowest slope and range were yielded by the equation of Fernhall et al. (1998). Large and signiwcant slopes were present in all the prediction equations presented in this study. Given that this systematic error is present, the constructed 95% LoA are diycult to interpret and potentially misleading. To this end, we proceed with the log transformation of the data. The correlation between the absolute diverences and the mean was positive in all equations (median r = 0.75), all being signiwcant (P < 0.01). From the eight equations analysed

5 Table 3 Bland Altman approach to compare directly measured peak VO 2 (ml kg 1 min 1 ) and estimated peak VO 2 (ml kg 1 min 1 ) from previously reported equations, together with the correlation between the absolute diverences and the mean d LoA Range Slope r FITNESSGRAM ** Leger et al. (1988) 4.7* ** Barnett et al. (1993)(a) 6.1* ** Barnett et al. (1993)(b) 4.2* ** Barnett et al. (1993)(c) 3.5** ** Fernhall et al. (1998) 7.3* ** Matsuzaka et al. (2004)(a) 5.4* ** Matsuzaka et al. (2004)(b) 4.2* ** Mean diverences (d), limits of agreement (LoA) and range (upper LoA lower LoA) are reported in ml kg 1 min 1 *DiVerence between directly measured peak VO 2 and estimated peak VO 2 from each equation (P < 0.05); ** diverence between directly measured peak VO 2 and estimated peak VO 2 from each equation (P <0.01) Prediction equations for VO 2peak Leger et al. (1988): VO 2peak = (Speed) 3.248(Age) (Age)(Speed) Barnett et al. (1993): (a) VO 2peak = (gender: M = 0; F = 1) 0.7(TricSKF) + 2.6(MS); (b) VO 2peak = (gender: M = 0; F=1) 0.2(weight) + 3.2(MS); (c) VO 2peak = (gender: M = 0; F = 1) 0.8(age) + 3.4(MS) Fernhall et al. (1998): VO 2peak = 0.35(TL) 0.59(BMI) 4.61(gender: M = 1; F = 2) Matsuzaka et al. (2004): (a) VO 2peak = (gender: M = 0; F = 1) (age) 0.831(BMI) (MS); (b) VO 2peak = (gender: M = 0; F = 1) 0.462(age) 0.862(BMI) (TL) only Barnett et al. (1993)(a) had a correlation greater than the equivalent correlation using log transformed measurements (median r =0.69). By observing the agreement ratios in Table 5, the equation of Matsuzaka et al. (2004)(a) showed the greatest agreement, with some bias, 1.11, but a good agreement ratio */ A quite similar situation was found with the peak VO 2 estimated by the equation of Fernhall et al. (1998) with the highest bias, given as 1.17, but the lowest agreement ratio (*/ 1.25). The worst agreement was found in the peak VO 2 estimated by FITNESSGRAM reports. Although the bias was 1.02, the agreement ratio (*/ 1.34) implies that 95% of ratios will lie between 35% of the mean bias ratio. Discussion The main Wnding of our study was that the prediction of peak VO 2 from the 20-m MSRT varies greatly among existing prediction equations in 8 10-year-old children. In general, although reasonable estimations of group peak VO 2 can be made from most of the existing equations tested, the accuracy of individual peak VO 2 predictions is unacceptable for all equations. Furthermore, most of the equations exhibited a systematic bias, resulting in an over-prediction in less Wt participants, but an under-prediction in more Wt participants. Several authors (Leger et al. 1988; Matsuzaka et al. 2004; McVeigh et al. 1995) suggest the best estimations of peak VO 2 are from those equations that include BMI or skinfold thickness, age and gender (Barnett et al. 1993; Fernhall et al. 1998; Matsuzaka et al. 2004) rather than those that included age only (Leger et al. 1988). It has been suggested that, in addition to chronological age, biological maturation inxuences the relationship between peak VO 2 and 20-m MSRT performance (Leger et al. 1988). Longitudinal studies provide a better insight of ontogenic changes in peak VO 2 in children. Armstrong and Welsman (2001) used multilevel modelling to assess serial changes in peak VO 2 between ages 11 and 17 years. Fat-free mass (FFM) was the major factor inxuencing growth of peak VO 2. The multilevel regression models indicated that even when body mass (BM) and composition were considered, peak VO 2 increased with age and maturation in both genders. However, BM may not adequately rexect the metabolically active skeletal muscle mass, due to heterogeneity of body composition across the sample (Tolfrey et al. 2006). Moreover, the use of BM as a gross surrogate of active muscle mass may be especially problematic in humans (compared with scaling studies in quadrupeds) because, in bipeds, running at maximal oxygen uptake (VO 2max ) does not involve whole body musculature (Hoppeler 1990). Theoretically, the confounding factor of variability in body composition may potentially be addressed, in part, by scaling the data by estimated FFM. However, scaling by FFM relies on the assumption that the involved metabolically active musclemass at VO 2max represents a constant proportion of FFM across the sample which is not the case in children (Tolfrey et al. 2006).

6 Fig. 1 Bland Altman plot of measured directly peak VO 2 and the estimated peak VO 2 From: a Fitnessgram; b Leger et al. (1988); c, d, e Barnett et al. (1993); f Fernhall et al. (1998); g, h Matsuzaka et al. (2004) Welsman et al. (1997) reported a correlation coeycient of 0.84 between total thigh muscle volume and peak VO 2 in prepubertal girls, and Zanconato et al. (1994) found that peak VO 2 related allometrically to calf muscle cross-sectional area by the scaling exponent Tolfrey et al. (2006) critically examined the inxuence of body size on VO 2max in boys and men using BM, estimated FFM, and estimated lower leg muscle volume (Vol) assed by sequential axial plane magnetic resonance imaging (MRI) scans of the lower right leg, as the separate scaling variables. The authors demonstrated clearly that an estimate of a proportion of the involved musculature (VO 2max /Vol 0.64 ) is a more valid allometric scaling denominator than either BM (VO 2max /BM 0.79 ) or FFM (VO 2max /FFM 1.0 ) for properly partitioning out the inxuence of body size on VO 2max. They point out that scaling by BM is confounded by heterogeneity of body composition and potentially substantial diverences in the mass exponent between boys and men. The FFM is precluded as an index of involved musculature because Vol did not represent a constant proportion of FFM in boys (unlike the men). It therefore seems likely that some parameters related to maturation, as well as chronological age, should potentially be included in the prediction equations, but this has not been done to date. Although peak HR (195 9) was adequate, the peak RER ( ) obtained during the 20-m MSRT was low. This is perplexing since the 20-m MSRT is known to be a maximal test. As a matter of fact, Voss and Sandercock (2009) found that in a naturalistic setting, the 20-m MSRT elicited a maximal evort in most of the 208 children tested, evidenced by peak HR. We show similar HR results, but the RER data suggest the children may not have produced maximal evort during the test. Lack of motivation at high exercise intensities, unfamiliarity with maximal evort are possible explanations for this Wnding. A diverent explanation may be the validity of criteria for establishing peak VO 2. Poole et al. (2008) using eight healthy male subjects hypothesized that veriwcation of VO 2max using RER or HR max variables is fundamentally Xawed in that their use could either allow underestimation

7 Table 4 Gender based Bland Altman approach to compare directly measured peak VO 2 (ml kg 1 min 1 ) and estimated peak VO 2 (ml kg 1 min 1 ) from previously reported equations Boys Girls d LoA Range Slope d LoA Range Slope FITNESSGRAM Leger et al. (1988) Barnett et al. (1993)(a) Barnett et al. (1993)(b) Barnett et al. (1993)(c) Fernhall et al. (1998) Matsuzaka et al. (2004)(a) Matsuzaka et al. (2004)(b) Mean diverences (d), Limits of agreement (LoA) and range (upper LoA lower LoA) are reported in ml kg 1 min 1 Prediction equations for VO 2peak Leger et al. (1988): VO 2peak = (Speed) 3.248(Age) (Age)(Speed) Barnett et al. (1993): (a) VO 2peak = (gender: M = 0; F = 1) 0.7(TricSKF) + 2.6(MS); (b) VO 2peak = (gender: M = 0; F=1) 0.2(weight) + 3.2(MS); (c) VO 2peak = (gender: M = 0; F = 1) 0.8(age) + 3.4(MS) Fernhall et al. (1998): VO 2peak = 0.35(TL) 0.59(BMI) 4.61(gender: M = 1; F = 2) Matsuzaka et al. (2004): (a) VO 2peak = (gender: M = 0; F = 1) (age) 0.831(BMI) (MS); (b) VO 2peak = (gender: M = 0; F = 1) 0.462(age) 0.862(BMI) (TL) Table 5 Bland Altman approach to compare the log transformed (ln) directly measured peak VO 2 (Mean 1), estimated peak VO 2 from previously reported equations (mean 2), the ratio limits of agreement, together with the correlation between the absolute diverences and the mean Mean 1 SD 1 Mean 2 SD 2 Ratio Limits Range (ml kg 1 min 1 ) r FITNESSGRAM */ ** Leger et al. (1988) */ ** Barnett et al. (1993)(a) */ ** Barnett et al. (1993)(b) */ ** Barnett et al. (1993)(c) */ ** Fernhall et al. (1998) */ ** Matsuzaka et al. (2004)(a) */ ** Matsuzaka et al. (2004)(b) */ ** Mean 1 and 2, ratio limits of agreement (LoA) and correlation (r) are log transformed (ln). Range (upper LoA lower LoA) is reported in ml kg 1 min 1 ** SigniWcant correlation (P < 0.01) Prediction equations for VO 2peak Leger et al. (1988): VO 2peak = (Speed) 3.248(Age) (Age)(Speed) Barnett et al. (1993): (a) VO 2peak = (gender: M = 0; F = 1) 0.7(TricSKF) + 2.6(MS); (b) VO 2peak = (gender: M = 0; F=1) 0.2(weight) + 3.2(MS); (c) VO 2peak = (gender: M = 0; F = 1) 0.8(age) + 3.4(MS) Fernhall et al. (1998): VO 2peak = 0.35(TL) 0.59(BMI) 4.61(gender: M = 1; F = 2) Matsuzaka et al. (2004): (a) VO 2peak = (gender: M = 0; F = 1) (age) 0.831(BMI) (MS); (b) VO 2peak = (gender: M = 0; F = 1) 0.462(age) 0.862(BMI) (TL) of VO 2max, or alternatively preclude subjects from recording a valid VO 2max. Using the most widely used criterion values of RER (1.10 and 1.15), VO 2max as determined during the ramp test ( l min 1 ) could be undermeasured by 27 and 16%, respectively. This Wnding is not consistent with the early work of Issekutz and Rodahl (1961) who, based upon a series of 5-min intermittent exercise tests, found that each participant reached VO 2max at an RER of The authors also concluded that even using the generous allowance of HR max, 10 b min 1, as the criterion for VO 2max there will be individuals excluded not because they could not reach VO 2max but because they exhibited a low

8 HR max. Analogous Wndings have been reported in 13 children (9 10 years) by Barker et al. (2009). Compared to the peak VO 2 during the ramp test, a signiwcantly lower peak VO 2 was recorded at an RER of 1.00 (1.3 vs. 1.7 l min 1 ), and at a HR of 195 beats min 1 (1.6 vs. 1.7 l min 1 ) and at 85% of age predicted maximum (1.4 vs. 1.7 l min 1 ). The authors concluded that the use of secondary criteria to verify a maximal evort in young people during ramp cycling exercise may result in the acceptance of a submaximal VO 2max. Based on the peak HR it is clear that the children in our study produced similar evort as reported in previous studies, and it is likely they produced a peak evort, even if that cannot be dewned as VO 2max. The few studies to date that have measured directly peak VO 2 during the 20-m MSRT indicate that there were no diverences between peak VO 2 during the 20-m MSRT and peak VO 2 measured in a treadmill test (TT) in laboratory (Leger and Lambert 1982), or that peak VO 2 is actually slightly higher during the 20-m MSRT (Flouris et al. 2005). In the study of Leger and Lambert (1982) data on the 25 subjects who performed the modiwed Balke TT protocol indicated similar results (r = and SEE = 3.3 ml kg 1 min 1 ) for the retroextrapolated ( ml kg 1 min 1 ) and the last minute of exercise VO 2max ( ml kg 1 min 1 ). VO 2 max measured directly during the last stage of the modiwed Balke TT protocol ( ml kg 1 min 1 ) was similar (r = and SEE = 4.16) to that measured by the retroextrapolation method at the end of the 20-m MSRT ( ml kg 1 min 1 ). This in fact the larger study to date comparing the measured directly VO 2 consumption in both Weld and laboratorial environment and the results are conclusive in showing no diverence between the peak VO 2 measured during the TT and the 20-m MSRT. In a more recent study, Flouris et al. (2005) assessed VO 2max in 110 males (age ; BMI ) using a stationary gas analyzer in a TT (modiwed Bruce treadmill test to exhaustion) and in 40 of these subjects using a portable gas analyzer in the 20-m MSRT. The results suggest increased energy demands during 20-m MSRT compared to TT ( vs ml kg 1 min 1, respectively). Thus, we are conwdent that the measured directly peak VO 2 during the 20-m MSRT is a reasonable and accurate representation of the peak VO 2 of our subjects. However, our data need to be cautiously interpreted given the above discussion regarding secondary criteria for determination of peak VO 2. Ruiz et al. (2009) also assed the agreement between VO 2max directly measured when performing the 20-m MSRT and estimated VO 2max from Wve diverent equations in year olds. Four of these equations were used in the present study allowing for comparisons: Barnett et al. (1993)(b), (c); Leger et al. (1988); Matsuzaka et al. (2004)(b). In the study of Ruiz et al. (2009), the 20-m MSRT was performed by 26 girls ( year; BM = kg) and 22 boys ( year; BM = kg) all of whom wore the K4b 2 Cosmed portable gas analyser to measure VO 2max in the Weld. Measured VO 2max for the 20-m MSRT was ml kg 1 min 1, slightly lower than the values in the present study, with corresponding means of estimated VO 2max of ml kg 1 min 1 for the equation of Leger et al. (1988), ml kg 1 min 1 for Barnett et al. (1993)(b), ml kg 1 min 1 for Barnett et al. (1993)(c) and ml kg 1 min 1 for Matsuzaka et al. (2004)(b). These estimations were found to have a higher range compared to that obtained in the present study as the d in the study of Ruiz et al. (2009) ranged from 1.3 ml kg 1 min 1 (Barnett et al. 1993)(b) to 5.5 ml kg 1 min 1 (Leger et al. 1988) whereas in the present study, and for the same restrict number of equations, the d ranged from 3.5 ml kg 1 min 1 (Barnett et al. 1993)(c) to 4.7 ml kg 1 min 1 (Leger et al. 1988). The authors concluded that all the equations signiwcantly underestimate directly measured VO 2max, except the equation of Barnett et al. (1993)(c). We did not Wnd such pattern as the equation of Barnett et al. (1993)(c) over-predicted the measured peak VO 2 by 4.2 ml kg 1 min 1. In addition, the peak VO 2 estimated by the FIT- NESSGRAM reports was the only equation that was not signiwcantly diverent from the directly measured peak VO 2 (P > 0.05). If we only take into account the d, the equation reported by Leger et al. (1988) proved to be the least accurate of the selected equations for estimating cardiorespiratory Wtness of children, similar to the Wndings of Ruiz et al. (2009). The diverences observed between the equation of Leger et al. (1988) and the other 3 equations may in part be explained by the test protocols and the gas analysis procedures used for the tests. Leger et al. (1988) used maximal speed calculated from the last stage completed. This might be a source of error, since participants stopping just short of completing a full stage would be ascribed to the previously completed stage. Leger et al. (1988) developed their equation when the Weld techniques to measure peak VO 2 directly through light weight portable metabolic systems were not available. Another possible cause for the discrepancy in estimating peak VO 2 might be the test protocol used, which is the case for the equations computed with peak VO 2 directly measured from treadmill-based protocols (Barnett et al. 1993; Flouris et al. 2005; Matsuzaka et al. 2004; Stickland et al. 2003). However, peak VO 2 measured during the 20-m MSRT is more ecologically valid and the heart rates reached during the 20-m MSRT are similar to or greater than those measured in the laboratory (Guerra et al. 2003; Voss and Sandercock 2009).

9 Matsuzaka et al. (2004) reported that the prediction equation by Leger et al. (1988) yielded little mean diverence (boys: d = 1.7 ml kg 1 min 1 ; girls: d = 0.3 ml kg 1 min 1 ) and non signiwcant slopes in either male ( 0.1) and female (0.1) children and adolescents. In the present study, the prediction equation by Leger et al. (1988) yielded a considerably higher mean diverence (boys: d =6.8mlkg 1 min 1 ; girls: d =2.5mlkg 1 min 1 ) and higher slopes in boys (1.2) and girls (1.0). The second equation considered by Matsuzaka et al. (2004) was that of Barnett et al. (1993)(c) including gender, age and MS. It yielded the lowest d (boys: 0.4 ml kg 1 min 1 ; girls: 0.3 ml kg 1 min 1 ) but the highest negative slopes in male and female ( 0.7) children and adolescents, meaning that the equation overpredicted peak VO 2 in participants with lower peak VO 2 and under-predicted peak VO 2 in participants with higher peak VO 2. In the present study, the same equation of Barnett et al. (1993)(c) showed a higher d (boys: 3.7 ml kg 1 min 1 ; girls: 3.3 ml kg 1 min 1 ) and steeper slopes in boys (1.4) and girls (1.5) corroborating the conclusion that the equation over-predicted peak VO 2 in participants with higher peak VO 2 and under-predicted peak VO 2 in participants with lower peak VO 2. Matsuzaka et al. (2004) also stated that the equation of Fernhall et al. (1998) over-predicted peak VO 2, especially for Wt male children and adolescents (d = 10.0 ml kg 1 min 1 ; slope = 0.5). In the present study, this Wnding was not conwrmed. Instead, the equation of Fernhall et al. (1998) over-predicted peak VO 2 in participants with lower peak VO 2 and an underprediction of the peak VO 2 in participants with higher peak VO 2 in boys (d =6.8mlkg 1 min 1 ; slope = 0.7), and girls (d =7.8mlkg 1 min 1 ; slope = 0.8). Overall, due to the small d, and small LoA and slope, the FITNESSGRAM software reports and the equation of Fernhall et al. (1998), respectively, showed the closest agreement among the equations tested. However, prediction of individual peak VO 2 is less accurate. For example, equation of Fernhall et al. (1998), an estimated peak VO 2 of 50.0 ml kg 1 min 1 actually means that we can be 95% sure the peak VO 2 is between 37.5 and 62.6 ml kg 1 min 1. Obviously, such a large conwdence interval means that individual prediction of peak VO 2 would not be advisable. The precise form that the 95% LoA takes depends on whether a positive relation exists between the diverences in measurement methods (errors) and the size of the measurements, i.e. heteroscedastic errors. If a positive and signiwcant relation exists, the recommended procedure is to report the ratio limits of agreement using log transformed measurements (Nevill and Atkinson 1997). Thus, is not surprising that when modelling measurements such as peak VO 2 as the range of all variables, recorded on a ratio scale (variables that cannot be negative and have a natural zero point) is forced to remain non-negative at the lower end of the scale but is theoretically unbounded at the other. This will naturally lead to heteroscedastic errors when two measurement methods, both on the same ratio scale, are to be compared and assessed for measurement agreement (Nevill et al. 1992). In order to better interpret such data, Nevill et al. (1992) recommend taking logarithms of both measurement methods and proceed as before by reporting the limits of agreement but using a log scale. In the present study, the correlation between the absolute diverences and the mean was positive in all equations (median r = 0.75), all being signiwcant (P < 0.01). From the eight equations analysed only Barnett et al. (1993)(a) had a correlation greater than the equivalent correlation using log transformed measurements (median r =0.69). As shown by the agreement ratios in Table 5, the equation of Matsuzaka et al. (2004)(a) shows the greatest agreement, with some bias, 1.11, but with a good agreement ratio */ 1.25, i.e. 95% of ratios are constrained between approximately 25% of the mean bias ratio, 1.11/1.25 = 0.89 and 1.11*1.25 = A quite similar situation is found with the peak VO 2 estimated by the equation of Fernhall et al. (1998) with the highest bias, given as 1.17, but with the lowest agreement ratio (*/ 1.25) which implies that 95% of ratios will lie between 25% of the mean bias ratio, i.e. from (1.17/1.25 = 0.93) to (1.17*1.25 = 1.46). The worst agreement was found in the peak VO 2 estimated by FITNESSGRAM reports. Although the bias was 1.02, the agreement ratio (*/ 1.34) implies that 95% of ratios will lie between 35% of the mean bias ratio, i.e. from (1.02/1.34 = 0.76) to (1.02*1.34 = 1.38). Consequently, if a new subject from the studied population presented with a measured directly peak VO 2 of 30 ml kg 1 min 1, there is a 95% probability that their predicted peak VO 2 from the FITNESSGRAM reports could be as low as = ml kg 1 min 1 or as high as = ml kg 1 min 1. Cooper et al. (2005) investigated the criterion related validity of the 20-m MSRT for VO 2max in 30 active young men ( years) who performed a continuous incremental laboratory test to volitional exhaustion to determine VO 2max and the 20-m MSRT. The results for each subject were expressed as a predicted VO 2max (ml kg 1 min 1 ) obtained by cross-referencing the Wnal level and shuttle number (completed) at which the subject volitionally exhausted with that of the VO 2max table provided in the instruction booklet accompanying the 20-m MSRT. The authors showed that the 20-m MSRT signiwcantly underpredicted VO 2 max (d = ml kg 1 min 1 ; P = 0.004), in fact the same d we reported with the estimations of peak VO 2 by FITNESSGRAM software reports. LoA were 6.3 ml kg 1 min 1 and heteroscedasticity 0.08 in opposition to the 14.1 ml kg 1 min 1 and 0.91, respectively, for the FITNESSGRAM software reports. Log transformation

10 reduced heteroscedasticity to 0.05 (P = 0.81) with LoA The signiwcant systematic bias was not eliminated (d = ; p = 0.003). Antilogs gave a mean bias of with random error */ In the present study, log transformation barely reduced heteroscedasticity (0.90) and antilogs gave a mean bias of 1.02 with random error */ In agreement with Cooper et al. (2005), we think that considering that the 20-m MSRT is a Weld test, the ratio limits for the lower performing subject (30 ml kg 1 min 1 ) are probably just on the border of acceptability, whilst the ratio limits for the higher performer (70 ml kg 1 min 1 ) are too wide to be acceptable for most sports scientists. These examples, plus additional examples provided by Nevill and Atkinson (1997) provide strong evidence that when assessing measurement agreement of variables recorded on a ratio scale in sports medicine and sports science, heteroscedastic errors are the norm and, as such, advocate the use of the log transformation when assessing measurement agreement. Considering the above discussion, the fact that a signiwcant systematic bias was identiwed in these data indicates that these equations cannot be considered valid predictors of measured directly peak VO 2. Conclusions The data reported in the present study indicate that the FIT- NESSGRAM software reports and the equations of Leger et al. (1988), Barnett et al. (1993), Fernhall et al. (1998) and Matsuzaka et al. (2004) might be unsuitable to estimate the cardiorespiratory Wtness of healthy 8-10 year old children. The equations of Fernhall et al. (1998) and Matsuzaka et al. (2004)(a) showed the closest agreement, but they still yield unsatisfactory accuracy for individual prediction of peak VO 2. Thus, published equations predicting peak VO 2 from MSRT may provide reasonable group estimates, but poor individual predictions of peak VO 2. Future recommendations The results from present study suggest that a regression equation speciwc for 8 10 year old children is needed, especially if individual prediction of peak VO 2 is desirable. The inclusion of skinfold measurement, maturational variables and biomechanical ratios in the regression equation may help to account for diverences in running economy and running styles between individuals with diverent body composition or reduce the requirement of repeat testing, and this should be explored in future studies. The secondary criteria to verify a maximal evort in young people during Weld exercise should be considered in future studies. References Anderssen SA, Cooper AR, Riddoch C, Sardinha LB, Harro M, Brage S, Andersen LB (2007) Low cardiorespiratory Wtness is a strong predictor for clustering of cardiovascular disease risk factors in children independent of country, age and sex. Eur J Cardiovasc Prev Rehabil 14: Armstrong N, Welsman JR (2001) Peak oxygen uptake in relation to growth and maturation in 11- to 17-year-old humans. Eur J Appl Physiol 85: Barker AR, Williams CA, Jones AM, Armstrong N (2009) Establishing maximal oxygen uptake in young people during a ramp cycle test to exhaustion. Br J Sports Med Barnett A, LYS C, IC B (1993) A preliminary study of the 20-m multistage shuttle run as a predictor of peak VO 2 in Hong Kong Chinese students. Pediatr Exerc Sci 5:42 50 Boreham C, Paliczka VJ, Nichols AK (1990) A comparison of the PWC170 and 20-MST tests of aerobic Wtness in adolescent schoolchildren. J Sports Med Phys Fitness 30:19 23 Cooper K (1970) The new aerobics. Bantam Books Inc, New York Cooper SM, Baker JS, Tong RJ, Roberts E, Hanford M (2005) The repeatability and criterion related validity of the 20 m multistage Wtness test as a predictor of maximal oxygen uptake in active young men. Br J Sports Med 39:e19 Falgairette G, Bedu M, Fellmann N, Spielvogel H, Van Praagh E, Obert P, Coudert J (1994) Evaluation of physical Wtness from Weld tests at high altitude in circumpubertal boys: comparison with laboratory data. Eur J Appl Physiol Occup Physiol 69:36 43 Fernhall B, Pitetti KH, Vukovich MD, Stubbs N, Hensen T, Winnick JP, Short FX (1998) Validation of cardiovascular Wtness Weld tests in children with mental retardation. Am J Ment Retard 102: Flouris AD, Metsios GS, Koutedakis Y (2005) Enhancing the eycacy of the 20 m multistage shuttle run test. Br J Sports Med 39: Guerra M, Pitetti K, Fernhall B (2003) Cross validation of the 20-meter shuttle run test for adolescents with Down syndrome. Adapt Phys Act Q 20:70 79 Hoppeler H (1990) The diverent relationship of VO 2max to muscle mitochondria in humans and quadrupedal animals. Respir Physiol 80: Issekutz B Jr, Rodahl K (1961) Respiratory quotient during exercise. J Appl Physiol 16: Kline GM, Porcari JP, Hintermeister R, Freedson PS, Ward A, McCarron RF, Ross J, Rippe JM (1987) Estimation of VO 2max from a one-mile track walk, gender, age, and body weight. Med Sci Sports Exerc 19: Leger L, Boucher R (1980) An indirect continuous running multistage Weld test: the Universite de Montreal track test. Can J Appl Sport Sci 5:77 84 Leger L, Gadoury C (1989) Validity of the 20 m shuttle run test with 1 min stages to predict VO 2max in adults. Can J Sport Sci 14:21 26 Leger L, Lambert J (1982) A maximal multistage 20-m shuttle run test to predict VO 2 max. Eur J Appl Physiol Occup Physiol 49:1 12 Leger L, Mercier D, Gadoury C, Lambert J (1988) The multistage 20 metre shuttle run test for aerobic Wtness. J Sports Sci 6: Liu NY, Plowman SA, Looney MA (1992) The reliability and validity of the 20-meter shuttle test in American students 12 to 15 years old. Res Q Exerc Sport 63:

11 Mahoney C (1992) 20-MST and PWC170 validity in non-caucasian children in the UK. Br J Sports Med 26:45 47 Matsuzaka A, Takahashi Y, Yamazoe M, Kumakura N, Ikeda A, Wilk B, Bar-Or O (2004) Validity of the multistage 20-m shuttle-run test for Japanese children, adolescents, and adults. Pediatr Exerc Sci 16: McLaughlin J, King G, Howley E, Bassett D, Ainsworth B (2001) Validation of the COSMED K4 b2 portable metabolic system. Int J Sports Med 22: McVeigh S, Payne A, Scott S (1995) The reliability and validity of the 20-meter shuttle test as a predictor of peak oxygen uptake in Edinburgh school children, age 13 to 14 years. Pediatr Exerc Sci 7:69 79 Nevill AM, Atkinson G (1997) Assessing agreement between measurements recorded on a ratio scale in sports medicine and sports science. Br J Sports Med 31: Nevill AM, Ramsbottom R, Williams C (1992) Scaling physiological measurements for individuals of diverent body size. Eur J Appl Physiol Occup Physiol 65: Paliczka VJ, Nichols AK, Boreham CA (1987) A multi-stage shuttle run as a predictor of running performance and maximal oxygen uptake in adults. Br J Sports Med 21: Pitetti KH, Fernhall B, Figoni S (2002) Comparing two regression formulas that predict vo2peak using the 20-m shuttle run for children and adolescents. Pediatr Exerc Sci 14: Poole DC, Wilkerson DP, Jones AM (2008) Validity of criteria for establishing maximal O 2 uptake during ramp exercise tests. Eur J Appl Physiol 102: Ramsbottom R, Brewer J, Williams C (1988) A progressive shuttle run test to estimate maximal oxygen uptake. Br J Sports Med 22: Ruiz JR, Silva G, Oliveira N, Ribeiro JC, Oliveira JF, Mota J (2009) Criterion-related validity of the 20-m shuttle run test in youths aged years. J Sports Sci 27: Stickland MK, Petersen SR, BouVard M (2003) Prediction of maximal aerobic power from the 20-m multi-stage shuttle run test. Can J Appl Physiol 28: The Cooper Institute for Aerobics Research (1994) FITNESS- GRAM the Prudential FITNESSGRAM test administration manual. Dallas, TX, USA Tolfrey K, Barker A, Thom JM, Morse CI, Narici MV, Batterham AM (2006) Scaling of maximal oxygen uptake by lower leg muscle volume in boys and men. J Appl Physiol 100: van Mechelen W, Hlobil H, Kemper HC (1986) Validation of two running tests as estimates of maximal aerobic power in children. Eur J Appl Physiol Occup Physiol 55: Voss C, Sandercock G (2009) Does the twenty meter shuttle-run test elicit maximal evort in 11- to 16-year-olds? Pediatr Exerc Sci 21:55 62 Welsman JR, Armstrong N, Kirby BJ, Winsley RJ, Parsons G, Sharpe P (1997) Exercise performance and magnetic resonance imagingdetermined thigh muscle volume in children. Eur J Appl Physiol Occup Physiol 76:92 97 Zanconato S, Riedy G, Cooper DM (1994) Calf muscle cross-sectional area and peak oxygen uptake and work rate in children and adults. Am J Physiol 267:R720 R725