A significant difference (p = 0.003) was also found between age-adjusted estimated HRmax and observed HRmax (199 ± 4 vs. 192 ± 9 b·min−1, respectively) with 76.5% of the participants having a lower observed HRmax compared with estimated HRmax. The SEE and E of age-predicted HRmax was 8.6 and 8.1 b·min−1, respectively.
The group results demonstrate that mean predicted V̇o2max using the ACSM treadmill test slightly overestimates the measured V̇o2max, and the error in accuracy of the predicted value indicated by the SEE and E were very similar for each of the 2 methods (P V̇o2max-2 and P V̇o2max-All). From an equation accuracy perspective, SEE and E are more useful in indicating the accuracy of a prediction equation than mean differences because unless deviations from the mean are all primarily positive or negative, they may have a canceling effect hence minimizing the mean difference (19,20). For example, although predicted V̇o2max was reported to slightly overestimate the measured V̇o2max for the group and this trend of overestimation occurred in 70% of the participants for both P V̇o2max-2 and P V̇o2max-All, it should be noted that M V̇o2max was underestimated in the remaining 30% of the participants. The SEE, on the other hand, indicates the accuracy of a predicted score, as does E, which takes into account the total deviations of the predicted values from the actual scores and so indicates the dispersion of scores around a regression line. A smaller SEE and E indicates a greater accuracy of prediction (3). Hence, because the SEE and E values were extremely similar for P V̇o2max-2 and P V̇o2max-All (SEE = 4.21–4.35; E = 3.98–4.08 ml·kg−1·min−1) with up to 70% of the predicted scores within 1 SEE of M V̇o2max, one can conclude that either approach can be used to predict V̇o2max with a similar accuracy.
When evaluating the magnitude of the SEE and E in this study, it is helpful to compare the errors with those of other studies that have also evaluated the ACSM treadmill running equation. The results of this study are in agreement with those of other studies in that V̇o2max is overestimated, but this study reports improved predictive accuracy over most of these studies. The SEE values for predicted V̇o2max in this study of 4.21–4.35 ml·kg−1·min−1 (P V̇o2max-2 and P V̇o2max-All) translate to an error of slightly >1 metabolic equivalent, which is better than the SEE reported in most other previous studies. For example, Foster et al. (11) found an SEE of 4.8 ml·kg−1·min−1 in an older cardiac patient population, and Mier and Gibson (23) reported an even higher SEE of 5.98 ml·kg−1·min−1; both the studies did not adhere to the exercise protocol stipulated by ACSM and used 1-minute stages of non–steady-state exercise – failure to achieve steady-state HR may have contributed to a greater overestimation in these studies. The only study reporting a better predictive accuracy was Morrow et al. (25) with an SEE of 3.7 ml·kg−1·min−1, but their data are not truly comparable because the ACSM protocol was performed on a running track instead of on a treadmill and at altitude. Unfortunately, because the total error was not calculated in any of the studies that evaluated the ACSM treadmill equation, a comparison of E values cannot be included.
However, despite there being no significant difference between predicted and measured V̇o2max for the group, and the relatively low SEE compared to other studies, a number of participants demonstrated considerable variance between predicted and measured V̇o2max, with 30% of predicted scores varying by >1 SEE of M V̇o2max. Because the procedures for estimating V̇o2max from the HR response during submaximal exercise tests are based on several assumptions, the focus of the ensuing discussion is to consider the implications of these assumptions on the predictive accuracy of the ACSM treadmill test in estimating V̇o2max.
When estimating V̇o2max using HR measurements, 1 of the assumptions of such procedures is that the aerobic demand (V̇o2) for a given submaximal WR is the same for everyone (2). However, variability does exist in measured V̇o2 for a given submaximal WR among different individuals (interindividual variability), with an SEE as high as 7% (2). Factors that may affect the V̇o2 during running (running economy) include training status, technique of movement, gender, age, and body mass. Trained runners have been shown to be more economical than untrained runners (24), men may be more economical than women (4), and adult athletes may be more economical than young athletes (6). Using the example of training status, the evaluation of submaximal aerobic demand during treadmill running (24) in elite, highly trained, and moderately trained runners and nonrunners (V̇o2max = 75.6, 70.5, 59.2, 51.4 ml·kg−1·min−1, respectively) revealed that V̇o2 varied significantly (p < 0.05) as a function of training—untrained runners were less economical than the 3 running groups and moderately trained and trained runners were less economical than elite runners. Hence, the assumption that V̇o2 for a given work load is the same for everyone, as assumed by the ACSM equation (equation 1), may influence the accuracy to which V̇o2max can be predicted for different individuals. For example, Morrow et al. (25) evaluated the ACSM equation using well-trained elite male runners (V̇o2max = 68.2 ml·kg−1·min−1), whereas this study used subjects with a lower training status (V̇o2max = 53.01 ml·kg−1·min−1), but the potentially better running economy of the elite male runners (25) compared with less trained runners as suggested by Morgan et al. (24) is not accounted for and may account for the slight variation in predictive accuracy between studies.
A further assumption made when estimating V̇o2max using HR measurements is that the HR– V̇o2 relationship is linear up to and including maximal WRs. Davies (8), however, showed that the HR–V̇o2 relationship is linear over most of its working range, but at near maximal WRs, the relationship changes, and the curve becomes nonlinear. Furthermore, Davies (8) reported that intraindividual HR variation fluctuates at different submaximal WRs influencing the linearity of the HR–V̇o2 relationship. The ACSM submaximal treadmill test stipulates that HR remains between 110 b·min−1 and 85% of age-predicted HRmax for at least 2 consecutive stages. The rationale for these recommendations is that, presumably, the HR–V̇o2 relationship remains linear and also that a steady-state HR is achievable within that HR range. Despite the linear HR–V̇o2 relationship during submaximal exercise, for any given level of submaximal work, the HR can vary independently of V̇o2 because of factors such as emotional state, hydration state, ambient temperature and humidity, body temperature and heat stress, time after previous meal and mechanical efficiency (1,13,22,30). These changes are often unaccompanied by concomitant changes in HRmax or the true V̇o2max; therefore, any intraindividual variation in submaximal HR measurements subsequently used in equations to calculate predicted V̇o2max may lead to errors in its calculation because a nonlinear increase in the HR–WR relationship directly influences the slope of the V̇o2 regression. Davies (8) reported that predicting V̇o2max using relatively low workloads may produce less accurate estimates because intraindividual variation in HR is 8% higher at lower workloads but is reduced to approximately 2% at higher exercise HRs of above 165 b·min−1, significantly improving the accuracy of predicting V̇o2max. Lamberts et al. (18) and Lamberts and Lambert (17) also reported that the lowest variation in HR occurred at higher HRs (∼85–90% HRmax), this may be because faster running speeds are more economical (22). Although the use of HR >165 b·min−1 suggested by Davies (8) is high and contradicts ACSM recommendations because it would exceed 85% age-predicted HRmax in individuals over the age of 26 years (85% of 194 b·min−1 = 165 b·min−1), it would seem reasonable to recommend that one works at higher HRs where intraindividual variation in HR is potentially lower but not so high that the HR deviates away from linearity and fails to achieve a steady state—both of which can contribute to inaccurate predictions of V̇o2max.
With respect to this study, it was noted that P V̇o2max-All and P V̇o2max-2 resulted in a similar mean predicted V̇o2max. For P V̇o2max-All, data from all stages between 110 b·min−1 and 85% HRmax were included in the HR–WR regression line, but one could argue that HR at the lower WRs may be subject to greater intraindividual variation and may instil inaccuracy into this method. However, fitting a regression line through all data points should eliminate anomalies that may occur in the data because of intraindividual HR variation at lower WRs, and furthermore, it would seem intuitive that the validity of the test would be enhanced by using more exercise stages. P V̇o2max-2, on the other hand, only uses the HR–WR data from 2 exercise stages from the exercise test and could be open to error because any deviation in HR away from linearity between the 2 exercise stages such as an unusually large or small increase in an individual's HR will profoundly affect the regression slope and subsequent calculation of predicted V̇o2max. However, because P V̇o2max-2 uses data from the last 2 stages within 85% HRmax, HR should be at a level where intraindividual HR variation is thought to be low (8). Two additional points that relate to the assumed linear relationship between HR and V̇o2 are that where the HR response to exercise is abnormal (tachycardia or bradycardia) this can lead to serious underestimation or overestimation of V̇o2max. Also, predicting V̇o2max by extrapolation of the submaximal HR–V̇o2 slope to estimated maximal HR does not allow for the nonlinear nature of curve toward maximal exercise, which may also result in predictive errors (30).
A further assumption that is made when estimating V̇o2max using the HR measurements is that all participants of a similar age attain a comparable HRmax (220 − age) (8). This assumption can lead to a large amount of error because considerable interindividual variability exists in the maximal HR achieved by people of the same age, the SD of age related HRmax is 10–12 b·min−1 (2,34). The traditional 220 − age equation has been reported to have a tendency to overestimate HRmax in many younger individuals (34). This trend was shown in the present study in which a significant difference (p = 0.003) was found between age-adjusted estimated HRmax and observed HRmax with >75% of the participants having a lower observed HRmax compared with estimated HRmax. The consequence of using estimated HRmax that is higher than the observed HRmax is an inflated predicted V̇o2max, and this is likely to be a major contributory factor to the overestimation of V̇o2max using the ACSM treadmill test. Interestingly, the use of age-predicted HRmax has recently been rejected as valid when verifying the achievement of V̇o2max during maximal exercise testing (28). It is possible, therefore, that the accuracy of predicted V̇o2max may be improved when observed HRmax is used, but it is appreciated that in most cases, a knowledge of HRmax would not be available to those using the ACSM treadmill running equations.
In conclusion, the ACSM treadmill running test slightly overestimates the measured V̇o2max and the accuracy of predicted V̇o2max was reasonably good and similar whether using data from all stages within 110 b·min−1 and 85% HRmax or just the data from the last 2 stages within 85% HRmax. Although 70% of predicted values were within 1 SEE of measured V̇o2max, nearly a third of individuals did show a larger variation between predicted and measured V̇o2max. This highlights that the use of HR as a independent variable influences the accuracy of predicted V̇o2max, and this is most likely because of the error incurred by the assumed linear relationship between HR and V̇o2, and extrapolation to estimated HRmax.
It is appreciated that a limitation of this study is the low external validity because of the relatively small sample size and the homogenous nature of the group that were tested. The results, therefore, cannot be generalized to the wider population but only to a specific group of young men. Future research needs to include participants from a wider age range, varying levels of fitness, and both genders. Although the ACSM treadmill ‘walking’ equation has been validated in healthy older men and women (age range 65–90) (27) with the conclusion that it tended to overestimate V̇o2max, particularly as aerobic fitness levels increased, the ACSM treadmill ‘running’ equation has not be validated in this group, because for many elderly participants, walking is more appropriate. However, future validation of the ACSM treadmill running equation research should include men and women between the age of 25 and 60 years.
Furthermore, Taylor et al. (33) observed that raising the grade while keeping the speed constant was the most satisfactory method of increasing work load to attain V̇o2max on a treadmill and, in fact, V̇o2max has been shown to be higher during uphill compared with during level running (26) by approximately 2 ml·kg−1·min−1 (9,32). This study used level running during the assessment of V̇o2max, but because maximal running speed may have been the limiting factor during the maximal test instead of the cardiopulmonary capacity for some individuals, it may have been better to have used an alternative treadmill protocol that used running speeds well within the individual's running capacity, and used treadmill inclination to increase WR. This may have achieved a measured V̇o2max slightly closer to predicted V̇o2max. It was noted by Kasch et al. (16), however, that there was no significant difference in V̇o2max achieved during horizontal and inclined treadmill running. Finally, the reliability of peak V̇o2 during the maximal test could have been tested by the use of a supramaximal constant WR performed immediately after the incremental test (29).
From a practical viewpoint, the ACSM equation provides a reasonable estimation of V̇o2max, overestimating measured V̇o2max by approximately 3% on average for the group, although V̇o2max was not overestimated in all individuals—it was underestimated in about a third of individuals. However, up to 70% of the participants achieved a predicted V̇o2max that was actually within just over 4 ml·kg−1·min−1 of measured V̇o2max, which is fairly close. Because the accuracy of the predicted value was very similar for both methods used (P V̇o2max-2 and P V̇o2max-All), it is recommended that the approach used to calculate predicted V̇o2max be down to personal preference although there are a few considerations that need to be taken into account when adopting an approach. When using P V̇o2max-2, it is important that the HR measurements used are taken from the latter exercise stages of the test (but within 85% of HRmax), where the HR is less subject to intraindividual variation that is more likely to occur at low WRs. When using P V̇o2max-All, it is essential that care is taken when plotting the HR against WR to ensure the most accurate extrapolation to maximal WR that is subsequently incorporated into the ACSM equation, careless data plotting will without doubt lead to an inaccurate prediction. With respect to who and how the use of the ACSM treadmill test is appropriate, it would serve as a useful tool in tracking changes in aerobic fitness in the same participant over a period of time and could be used by fitness professionals, gym instructors, and potentially coaches although it may not be suitable where a large number of players need to be assessed unless there are multiple treadmills to hand. Importantly, however, the ACSM treadmill test is not appropriate if the HR response is abnormal, perhaps because of medication, because predicted V̇o2max will be severely overestimated or underestimated. If a precise measure of V̇o2max is required though, as is necessary for some athletic groups, then this should be measured directly.
Research relating to this manuscript was not funded by a grant support. The author would like to thank all the participants who volunteered to participate in this study. The author has no conflict of interest to declare.
1. Achten, J and Jeukendrup, AE. Heart rate monitoring: Applications and limitations. Sports Med
33: 517–538, 2003.
2. American College of Sports Medicine. Guidelines for Exercise Testing and Prescription
. London, United Kingdom: Lippincott Williams & Wilkins, 2006.
3. Berg, KE and Latin, RW. Essentials of Research Methods in Health, Physical Education, Exercise Science, and Recreation
. London, United Kingdom: Lippincott Williams & Wilkins, 2008.
4. Bransford, DR and Howley, ET. Oxygen cost of running in trained and untrained men and women. Med Sci Sports
9: 41–44, 1977.
5. Buchfuhrer, MJ, Hansen, JE, Robinson, TE, Sue, DY, Wasserman, K, and Whipp, BJ. Optimizing the exercise protocol for cardiopulmonary assessment. J Appl Physiol
55: 1558–1564, 1983.
6. Bunc, V and Heller, J. Energy cost of running in young and adult female athletes. Ergonomics
37: 167–174, 1994.
7. Davies, CTM. Maximum oxygen uptake: Prediction from cardiac frequency during submaximal exercise. J Physiol
189: P77–P78, 1967.
8. Davies, CTM. Limitations to the prediction of maximum oxygen intake from cardiac frequency measurements. J Appl Physiol
24: 700–706, 1968.
9. Draper, SB, Wood, DM, and Fallowfield, JL. The effect of test protocol on V̇o2
peak and the incidence of a V̇o2
plateau. J Sports Sci
17: 31, 1999.
10. Eston, R, Williams, JG, and Faulkner, J. Control of exercise intensity using heart rate, perceived exertion and other non-invasive procedures. In: Kinanthropometry and Exercise Physiology Laboratory Manual: Test, Procedures and Data
. (Vol. 2). Physiology
. R. Eston and T. Reilly, eds. London, United Kingdom: Routledge, 2009. pp. 237–270.
11. Foster, C, Crowe, AJ, Daines, E, Dumit, M, Green, MA, Lettau, S, Thompson, NN, and Weymier, J. Predicting functional capacity during treadmill testing independent of exercise protocol. Med Sci Sports Exerc
28: 752–756, 1996.
12. Greiwe, JS, Kaminsky, LA, Whaley, MH, and Dwyer, GB. Evaluation of the ACSM submaximal ergometer test for estimating V̇o2
max. Med Sci Sports Exerc
27: 1315–1320, 1995.
13. Hamilton, MT, Gonzalez-Alonso, J, Montain, SJ, and Coyle, EF. Fluid replacement and glucose infusion during exercise prevent cardiovascular drift. J Appl Physiol
71: 871–877, 1991.
14. Heyward, VH. Advanced Fitness Assessment and Exercise Prescription
. Leeds, United Kingdom: Human Kinetics, 2002.
15. Howley, ET, Bassett, JR, and Welch, HG. Criteria for maximal oxygen uptake: Review and commentary. Med Sci Sports Exerc
27: 1292–1301, 1995.
16. Kasch, FW, Wallace, JP, Huhn, RR, Krogh, LA, and Hurl, PM. V̇o2
max during horizontal and inclined treadmill running. J Appl Physiol
40: 982–983, 1976.
17. Lamberts, RP and Lambert, MI. Day-to-day variation in heart rate at different levels of submaximal exertion: Implications for monitoring training. J Strength Cond Res
23: 1005–1010, 2009.
18. Lamberts, RP, Lemmink, KA, Durandt, JJ, and Lambert, MI. Variation in heart rate during submaximal exercise: Implications for monitoring training. J Strength Cond Res
18: 641–645, 2004.
19. Lang, PB, Latin, RW, Berg, KE, and Mellion, MB. The accuracy of the ACSM cycle ergometry equation. Med Sci Sports Exerc
24: 272–276, 1992.
20. Latin, RW, Berg, KE, Smith, P, Tolle, R, and Woodby-Brown, S. Validation of a cycle ergometry equation for predicting steady-rate V̇o2
. Med Sci Sports Exerc
25: 970–974, 1993.
21. Latin, RW and Elias, BA. Predictions of maximum oxygen uptake from treadmill walking and running. J Sports Med Phys Fitness
33: 34–39, 1993.
22. McArdle, WD, Katch, FI, and Katch, VL. Exercise Physiology, Energy, Nutrition and Human Performance
. London, United Kingdom: Lippincott Williams & Wilkins, 2001.
23. Mier, CM and Gibson, AL. Evaluation of a treadmill test for predicting the aerobic capacity of firefighters. Occup Med
54: 373–378, 2004.
24. Morgan, DW, Bransford, DR, Costill, DL, Daniels, JT, Howley, ET, and Krahenbuhl, GS. Variation in the aerobic demand of running among trained and untrained subjects. Med Sci Sports Exerc
27: 404–409, 1995.
25. Morrow, JR, Van Handel, PJ, Bradley, PW, and Kearney JT. The validity of the ACSM energy requirement equation with well-trained elite athletes at altitude. Med Sci Sports Exerc
20: S1, 1988.
26. Paavolainen, L, Nummela, A, and Rusko, H. Muscle power factors and V̇o2
max as determinants of horizontal and uphill running performance. Scand J Med Sci Sports
10: 286–291, 2000.
27. Peterson, MJ, Pieper, CF, and Morey, MC. Accuracy of V̇o2
max prediction equations in older adults. Med Sci Sports Exerc
35: 145–149, 2003.
28. Poole, DC, Wilkerson, DP, and Jones, AM. Validity of criteria for establishing maximal O2
uptake during ramp exercise tests. Eur J Appl Physiol
102: 403–410, 2008.
29. Rossiter, HB, Kowalchuk, JM, and Whipp, BJ. A test to establish maximum O2
uptake despite no plateau in the O2
uptake response to ramp incremental exercise. J Appl Physiol
100: 764–700, 2006.
30. Rowell, LB, Taylor, HL, and Yang W. Limitations to prediction of maximal oxygen intake. J Appl Physiol
19: 919–927, 1964.
31. Sherman, NW, Pivarnik, JM, Jackson, AS, and Ross, RM. The effect of estimating the energy cost of running on exercise prescription. Med Sci Sports Exerc
20: S2, 1988.
32. St Clair Gibson, A, Lambert, MI, Hawley, JA, Broomhead, S, and Noakes, TD. Measurement of maximal oxygen uptake from two different laboratory protocols in runners and squash players. Med Sci Sports Exerc
31: 1226–1229, 1999.
33. Taylor, HL, Buskirk, E, and Henschel, A. Maximal oxygen intake as an objective measure of cardio-respiratory performance. J Appl Physiol
8: 73–80, 1955.
34. Whaley, MH, Kaminsky, LA, Dwyer, GB, Getchell, LH, and Norton, JA. Predictors of over- and underachievement of age-predicted maximal heart rate
. Med Sci Sports Exerc
24: 1173–1179, 1992.
35. Yoon, B, Kravitz, L, and Robergs, R. V̇o2
max, protocol duration, and the V̇o2
plateau. Med Sci Sports Exerc
39: 1186–1192, 2007.
Keywords:Copyright © 2012 by the National Strength & Conditioning Association.
maximal heart rate; extrapolation; intraindividual variation; regression slope