It is well known that endurance training results in performance improvements in previously sedentary and recreationally active individuals (1). Furthermore, it is known that athletes who are already trained need to incorporate high-intensity interval training into their training programs to improve their performance in preparation for competition (12,17). A number of authors have investigated the effects of high-intensity training (HIT) on athletic performance (16,27,28). These studies demonstrate that high-intensity, short-duration interval training can result in further increases in peak power output (Wmax) and o2max in well-trained athletes.
Although the importance of HIT is widely accepted, the method of applying the training load varies. For example, there are 2 commonly used methods for applying workload during high-intensity cycle training. These are based on prescribing training intensity either according to heart rate or power output.
During training, a number of factors can alter the submaximal heart rate/workload relationship (15). However, a longitudinal study of 13 professional road cyclists found that under controlled conditions, the heart rate coinciding with various physiological markers was relatively constant (2-3 bpm) during the course of a season. In contrast, power output, which may be a more direct measure of workload, varied significantly (19).
Despite the popularity and accuracy of using power meters to monitor performance during training and racing, no studies have validated their use to prescribe training. Specifically, the use of power in preference to heart rate to prescribe training has not been justified. Accordingly, the aim of this study was to compare the training adaptations and performance characteristics of well-trained cyclists after 4 weeks of HIT using either heart rate or power output to prescribe the training intensity. We used a 40-km time trial and o2max test as measures of performance (21). We hypothesized that there would be no measurable differences between the 2 modes of training.
Experimental Approach to the Problem
The minimal sample size for this study was determined using the 40-km time trial data of Palmer et al. (20). Assuming that the smallest meaningful difference is 1.0% with an SD of 0.5%, the sample size required for this study to achieve a statistical power of 80% and a significance level of 5% was n = 5 for each group (3). The study was conducted during 18 months. Inclusion criteria for the study were a minimum of 6 training hours per week during the 6-week period before the trial and a minimum competitive cycling background of at least 3 years. All participating cyclists were randomly assigned to 1 of 3 groups: a power-based training group (GPOWER), a heart rate-based training group (GHEART), or a control group (GCONTROL). Two subjects were unwilling to participate in the GCONTROL group because of concerns about possible reductions in training status. These subjects were subsequently randomized to either the GPOWER or GHEART group.
Twenty-one well-trained men cyclists (13) (mean ± SD; age = 31 ± 6 years; stature = 1.82 ± 0.07 m; mass = 74.9 ± 8.8 kg) volunteered for this study. After being fully informed of the risks and stresses associated with the study, all subjects completed a Physical Activity Readiness Questionnaire (4), had their training logs analyzed, and had a personal interview about their cycling history. All subjects gave their written informed consent to participate in the study. The experimental protocol was approved by the ethics and research committee of the Faculty of Health Sciences of the University of Cape Town.
Each subject completed a 40-km familiarization time trial (TT), a peak aerobic capacity test (o2max), and a repeat 40-km TT test at 7 days, 3 days, and 1 day, respectively, before the start of the 28-day training period.
Subjects reported to the laboratory, which had stable climatic conditions (22.4 ± 1.4° C, 53 ± 5% relative humidity, 100.5 ± 8 kPa), at the same time of day for all tests and training sessions. During each performance trial, subjects were blinded to all cues other than completed distance to prevent the adoption of a pacing strategy or performance bias.
Subjects were asked not to eat for at least 2 hours before each of the performance tests or supervised training sessions. Each subject was asked to refrain from training for 24 hours before the o2max test and to perform a 90-minute submaximal recovery ride (at an intensity below the “lactate turnpoint” determined during o2max testing ) 24 hours before the 40-km TT familiarization and 40-km TT tests. They also were asked to refrain from consuming any caffeine for at least 3 hours before each performance test or training session. Before each testing session, subjects were questioned to confirm that they had adhered to these instructions.
After the 28-day training period, subjects completed a 10-day washout period of training, during which time they were asked not to participate in any racing or prolonged or high-intensity exercise. After the washout period, the o2max test, 40-km TT test, and anthropometric measurements were repeated. Subjects were verbally questioned before the testing procedure to ensure that they had adhered to the training protocol.
To confirm that subjects adhered to the prescribed training intensity, each subject was asked to keep a detailed training diary and to record all heart rate data during all training sessions performed outside of the laboratory.
Body Composition Measurements
Stature, body mass, and sum of 7 skinfolds (triceps, biceps, suprailiac, subscapular, calf, thigh, and abdomen) were measured immediately before the 40-km TT test according to the methods described by Ross and Marfell-Jones (24). In addition, body mass was measured before each testing and training session.
Apparatus and Testing Procedure
All testing and training was performed on an electronically braked cycle ergometer (Computrainer Pro 3D, RacerMate, Seattle, Wash). The subject's bicycle was attached to the ergometer by the rear axle quick release mechanism and supported under the front wheel by a plastic support. Rear tire pressure was inflated to 800 kPa before calibration. Calibration and load generator contact pressure was adjusted until the calibration value measured between 0.88 and 0.93 kg. The average calibration value recorded during the study was 0.914 ± 0.02 kg, as recommended by Davison et al. (9,10). Each subject completed a self-paced warm-up for 15 minutes before each testing and training session.
The o2max and Wmax tests were performed at a starting workload of 2.5 W·kg−1 body mass. Load was increased incrementally at a rate of 20 W every minute until the subject could not sustain a cadence of greater than 70 rpm or was volitionally exhausted. During the progressive exercise test, ventilation volume (e), o2, and co2 were measured for 15-second intervals using an online breath-by-breath gas analyzer and pneumotach (Oxycon, Viasis, Hoechberg, Germany). Calibration of this device was performed during the 15-minute self-paced warm-up according to the manufacturer's instructions. o2max was recorded as the highest o2 reading recorded for 30 seconds during the test. Peak power output was calculated by averaging the power output for the final minute of the o2max test.
Subjects completed the self paced maximal 40-km TT on a simulated flat 40-km course. Subjects were allowed to consume water ad libitum throughout the test and were asked to produce the fastest possible 40-km TT time.
After the testing period, each subject reported to the laboratory on 8 occasions for a supervised training session (2 sessions each week for 4 weeks). Subjects performed the training sessions on their own bicycles mounted to a Computrainer ergometer. After a 15-minute self-paced warm-up period, each subject completed a high-intensity interval training session, consisting of 8 intervals of 4 minutes' duration, interspersed with 90-second self-paced recovery periods.
Each GCONTROL subject completed a 40-km self-paced training ride twice each week at an intensity below 70% Wmax (determined during the progressive exercise test). In addition, GCONTROL subjects were asked to perform the same training as the HIT groups outside of the laboratory.
Subjects in the GHEART group completed each interval at a fixed workload corresponding to 80% Wmax, and subjects in the GHEART group were asked to increase and then maintain a target heart rate equivalent to the heart rate recorded at 80% Wmax. These values were chosen with the aim of ensuring that both HIT groups trained at the same intensity. Because of the slow half-life of heart rate response (2), subjects in the GHEART group were asked to achieve the target heart rate within 3 minutes during the first interval, within 2 minutes during the second interval, and within 1 minute for all subsequent intervals.
Between each laboratory training session, subjects were asked to adhere to the following training protocol:
-90-minute recovery session (at an intensity below the “lactate turnpoint” determined during o2max testing)
-90-minute recovery session
Analysis of training data (power output and heart rate) were performed using Cyclingpeaks analysis software, WKO+ edition, version 2.1, 2006 (Peaksware Inc., Lafayette, Colo). Data from each training session were filtered using the “Fast Find” function. “Leading Edge” and “Trailing Edge” values as well as minimum and maximum duration values were assigned according to each subject's absolute training intensity. Filtered data subsequently were inspected visually and corrected before analysis.
Data were analyzed for statistical significance using STATISTICA version 7.0 (Stat-soft Inc., Tulsa, Okla). A 2-way analysis of variance (ANOVA) with repeated measures was used to examine differences in performance measures between the groups before and after HIT intervention or control training. Where a significant difference was found (p ≤ 0.05) for either the main effect of group or the interaction, a Tukey post hoc analysis was performed. A significant interaction (group × time) was interpreted as meaning that the groups responded differently during the training period for that variable. A 1-way ANOVA was performed for the percentage change in each group for all the performance variables. All data are expressed as mean ± SD.
Further analysis of changes in performance and physiological variables after training were assessed using magnitude-based inferences following the procedure described by Batterham and Hopkins (5). Mean effects of training and their 90% confidence intervals (CIs) were estimated using an Excel spreadsheet (www.sportsci.org/jour/05/ambwgh.htm) with values obtained from a t-test for each independent variable between groups. The spreadsheet computes the chance that the true effect is substantial when a value for the smallest worthwhile change is entered. A value of 1% was defined as a meaningful difference for the performance measures, as used in previous cycling studies (8,21,22). Although we acknowledge the controversy regarding multiple comparisons, we have not tried to correct for any bias using a Bonferroni-adjusted p value because the advantages of this procedure may be outweighed by the disadvantages of this adjustment (23).
One cyclist from GPOWER fractured his wrist in a cycling accident, and 1 cyclist from GHEART contracted a viral illness; therefore, their data were excluded from further analysis. On analysis, 2 subjects (1 from GHEART and 1 from GPOWER) were found to have pretraining o2max scores that fell outside 2 SDs from the group average. The data from these subjects were therefore excluded from further analysis. Subsequently, 6 cyclists remained in each experimental group (GPOWER and GHEART), and 5 cyclists were in the GCONTROL group. The sample size criteria of 5 subjects per group were met by all groups. Descriptive measures of the cyclists in the 3 groups before the start of the study are shown in Table 1. There were no statistical differences in any of these measurements between groups.
The subjects' o2max, Wmax, and 40-km TT performance are shown in Table 2. There were no significant differences in o2max between groups either before or after training (Figure 1). An ANOVA with repeated measures for Wmax showed a significant interaction between groups over time (p < 0.05; Figure 1). On further analysis, the HIT groups (GPOWER and GHEART) increased Wmax scores significantly by 3.5 and 5.0%, respectively, relative to the GCONTROL group (p < 0.05; Table 2). There was a significant interaction between groups over time for 40-km TT performance (p < 0.05; Figure 1). Relative improvement in 40-km TT performance was significant for both of the HIT groups (GPOWER and GHEART) relative to the GCONTROL group, with improvements of 2.3 and 2.1%, respectively (p < 0.05; Table 2). There was no significant interaction between the 2 HIT groups for any of the dependent variables.
The mean changes in physiological and performance variables for all groups and magnitude-based statistics for the differences in the changes are shown in Table 3. For o2max scores, only the GHEART group showed a statistical likelihood of a beneficial effect from training. There were clear-cut beneficial effects on Wmax and 40-km TT performance for both the HIT groups (GPOWER and GHEART), with the greatest probability of a beneficial effect being demonstrated for the GHEART group in all of the parameters. For both o2max and Wmax, there was a statistically “likely” benefit of GHEART over GPOWER for improvements in these parameters. In addition, there was a significant correlation (r = 0.70; 95% CI: 0.36-0.87) between changes in Wmax and 40-km TT performance for all groups combined (Figure 2).
The objective of prescribing identical average training intensities for each HIT group was achieved during this study. Average power outputs during each interval for groups GPOWER and GHEART expressed as percentages of Wmax were 79 ± 1 and 78 ± 3% Wmax, respectively. The average training intensity of the GPOWER and GHEART groups did not differ significantly. This is in accordance with the similar relative heart rate intensity (GPOWER = 88 ± 3% HRmax; GHEART = 88 ± 2% HRmax). However, the power profile during intervals was markedly different between these groups. Specifically, the average 5-second maximum power output recorded for each interval differed significantly (p < 0.001) between the GPOWER and GHEART groups. The average 5-second maximum power output for all intervals for GHEART was 498 ± 89 W, whereas the average 5-second maximum power output for all intervals for GPOWER was 401 ± 63 W. Representative examples of 2 randomly chosen subjects are shown in Figure 3. Although the average power outputs and heart rates did not differ, intervals prescribed using heart rate were characterized by a high starting power output followed by a marked decline in power, which is sustained toward the end of the interval. Heart rate increased throughout these intervals until the target heart rate was achieved, after which heart rate remained stable. During power-based intervals, power remained relatively constant throughout, as dictated by the prescribed method, whereas heart rate increased continuously throughout each interval period, with a plateau toward the end of the interval (Figure 3).
The present study has shown that HIT performed at either a fixed heart rate or fixed power elicited improvements in Wmax and 40-km TT performance in already well-trained cyclists. The improvements in Wmax (range 3.5-5.0%) and 40-km TT performance (range 2.1-2.3%) were similar to changes previously reported using similar HIT protocols and conducted over similar time periods (11).
Although we cannot state unequivocally that, for this specific HIT protocol, heart rate-based training elicited greater improvements in performance and related parameters than power-based training, there is evidence in our data to suggest that this may be so. Analyses inferring about magnitudes (5) show that there is a “likely” beneficial effect of heart rate-based training in comparison with power-based training, particularly with reference to improvements in Wmax and o2max (Figure 1). Given that the performances of elite athletes are separated by small margins that are indiscernible and that cannot be detected with null-hypothesis testing (5,8), it is appropriate to infer about the magnitude of the differences between groups. This approach suggests that there may be advantages in using heart rate compared with power output prescribed training for this specific protocol. However, further research is needed to confirm these findings.
The average power output (79 ± 1 vs. 78 ± 3% Wmax for GPOWER vs. GHEART) and heart rate (88 ± 3 vs. 88 ± 2% HRmax for GPOWER vs. GHEART) during training intervals did not differ between the 2 HIT groups. However, the GHEART group had a different power profile during intervals, which may, therefore, have resulted in short periods of training above o2max and Wmax values (supramaximal training; Figure 3). Previous studies have shown that training at or above o2max intensity may be the most effective means of eliciting additional improvements in already highly trained athletes (6,7,14,18,25).
Further research is needed to compare responses to heart rate- and power-based training in the field and to compare the training effects at differing workloads and interval durations. In addition, a comparison of training responses using either fixed or variable power outputs within intervals may provide more valuable data for the optimal use of power meters to prescribe training. Power-based training may be more beneficial for training protocols such as T-max intervals (16,18), but evidence for other power-based sessions is still lacking.
In summary, HIT prescribed using either heart rate or power improved performance (Wmax and 40-km TT) similarly. However, magnitude-based inferences suggest that there may be an advantage of using target heart rate over power output for this particular HIT protocol.
The prescription of training based on heart rate has the potential to be inaccurate if the conditions under which the training is prescribed are not controlled. Although power is a more direct method of prescribing training, a potential limitation of this method is that power output training zones change significantly over a relatively short time period. This necessitates frequent maximal testing to update these values. In contrast, heart rate-based training zones remain relatively stable throughout prolonged periods of time, which makes this method easier and less dependent on frequent testing than power-based training. In conclusion, although power meters have proven to be accurate and useful tools to monitor training, exercise prescription by power has not proven to have additional benefits over heart rate when prescribing HIT. When coaches are unable to monitor training progress frequently, we recommend prescription using heart rate-based training. Coaches also should consider that heart rate-based intervals performed under stable conditions may provide an additional advantage over power-based intervals.
The authors thank the cyclists who participated in this study. We also would like to express our gratitude to Christel Rösemann for her assistance in data collection. The Medical Research Council, Discovery Health, and the Nellie Atkinson and Harry Crossley research funds of the University of Cape Town supported this study.
1. Abernethy, PJ, Thayer, R, and Taylor, AW. Acute and chronic responses of skeletal muscle to endurance and sprint exercise. A review. Sports Med
10: 365-389, 1990.
2. Achten, J and Jeukendrup, AE. Heart rate monitoring: applications and limitations. Sports Med
33: 517-538, 2003.
3. Altman, DG. Practical Statistics for Medical Research
. London: Chapman and Hall, 1991.
4. American College of Sports Medicine. Preparticipation health screening and risk stratification. In: ACSM's Guidelines for Exercise Testing and Prescription
. Whaley M.H., Brubaker P.H., Otto R.M., eds. Baltimore: Lippincott Williams & Wilkins, 2007.
5. Batterham, AM and Hopkins, WG. Making meaningful inferences about magnitudes. Int J Sports Physiol Perform
1: 50-57, 2006.
6. Billat, LV. Interval training for performance
: a scientific and empirical practice. Special recommendations for middle- and long-distance running. Part I: aerobic interval training. Sports Med
31: 13-31, 2001.
7. Billat, LV and Koralsztein, JP. Significance of the velocity at o2
max and time to exhaustion at this velocity. Sports Med
22: 90-108, 1996.
8. Currell, K and Jeukendrup, AE. Validity, reliability and sensitivity of measures of sporting performance
. Sports Med
38: 297-316, 2008.
9. Davison, RCR, Corbett, J, and Ansley, L. Influence of temperature and protocol on the calibration of the Computrainer electromagnetically braked cycling ergometer. Paper presented at: British Association of Sport & Exercise Science, Annual Conference, Wolverhampton, 2006.
10. Davison, RCR, Corbett, J, and Ansley, L. Influence of temperature and protocol on the calibration of the Computrainer electromagnetically braked cycling ergometer. J Sports Sci
25: 257-258, 2007.
11. Durnin, JVGA and Womersley, J. Body fat assessed from total body density and its estimation from skinfold thickness measurement on 481 men and women aged 16-72 years. Br J Nutr
32: 77-97, 1974.
12. Hawley, JA, Myburgh, KH, Noakes, TD, and Dennis, SC. Training techniques to improve fatigue resistance and enhance endurance performance
. J Sports Sci
15: 325-333, 1997.
13. Jeukendrup, AE. High-Performance Cycling
. Champaign: Human Kinetics, 2002.
14. Klausen, K. Cardiac output in man in rest and work during and after acclimatization to 3,800 m. J Appl Physiol
21: 609-616, 1966.
15. Lambert, MI, Mbambo, ZH, and St Clair Gibson, A. Heart rate during training and competition for long-distance running. J Sports Sci
16: S85-S90, 1998.
16. Laursen, PB, Blanchard, MA, and Jenkins, DG. Acute high-intensity interval training improves Tvent and peak power output in highly trained males. Can J Appl Physiol
27: 336-348, 2002.
17. Laursen, PB and Jenkins, DG. The scientific basis for high-intensity interval training: optimising training programmes and maximising performance
in highly trained endurance athletes. Sports Med
32: 53-73, 2002.
18. Laursen, PB, Shing, CM, Peake, JM, Coombes, JS, and Jenkins, DG. Interval training program optimization in highly trained endurance cyclists. Med Sci Sports Exerc
34: 1801-1807, 2002.
19. Lucia, A, Hoyos, J, Perez, M, and Chicharro, JL. Heart rate and performance
parameters in elite cyclists: a longitudinal study. Med Sci Sports Exerc
32: 1777-1782, 2000.
20. Palmer, GS, Dennis, SC, Noakes, TD, and Hawley, JA. Assessment of the reproducibility of performance
testing on an air-braked cycle ergometer. Int J Sports Med
17: 293-298, 1996.
21. Paton, CD and Hopkins, WG. Tests of cycling performance
. Sports Med
31: 489-496, 2001.
22. Paton, CD, and Hopkins, WG. Combining explosive and high-resistance training improves performance
in competitive cyclists. J Strength Cond Res
19: 826-830, 2005.
23. Perneger, TV. What's wrong with Bonferroni adjustments. BMJ
316: 1236-1238, 1998.
24. Ross, WD and Marfell-Jones, MJ. Kinanthropometry. In: Physiological Testing of the High Performance Athlete
. MacDougall J.D., Wenger H.A., and Green H.S., eds. Champaign: Human Kinetics, 1991.
25. Smith, TP, McNaughton, LR, and Marshall, KJ. Effects of 4-wk training using Vmax/Tmax on o2
max and performance
in athletes. Med Sci Sports Exerc
31: 892-896, 1999.
26. Solberg, G, Robstad, B, Skjønsberg, OH, and Borchsenius, F. Respiratory gas exchange indices for estimating the anearobic threshold. J Sports Sci Med
4: 29-36, 2005.
27. Stepto, NK, Hawley, JA, Dennis, SC, and Hopkins, WG. Effects of different interval-training programs on cycling time-trial performance
. Med Sci Sports Exerc
31: 736-741, 1999.
28. Weston, AR, Myburgh, KH, Lindsay, FH, Dennis, SC, Noakes, TD, and Hawley, JA. Skeletal muscle buffering capacity and endurance performance
after high-intensity interval training by well-trained cyclists. Eur J Appl Physiol Occup Physiol
75: 7-13, 1997.