Reliability and validity of the 3-min all-out running test

Confiabilidade e validade do teste de 3 minutos máximo

Fiabilidad y validez de la prueba de 3 minutos máximos

Critical power (CP) represents, at least theoretically, the highest power output that can be sustained from an aerobic energy supply, and has been suggested to demarcate the heavy and severe intensity domains (Poole, Ward, Gardner, & Whipp, 1988). In this context, CP represents a good indicator of performance in long-duration events, and is a useful parameter for training prescription (Bosquet, Leger, & Legros, 2002; Vanhatalo, Jones, & Burnley, 2011). The concept of CP was proposed originally for use with small muscle groups (Monod & Scherrer, 1965) but was later adapted for whole body exercises, including running, cycling, and swimming (Ettema, 1966; Hill, 1993).

Linear regression of total work performed during a series of exhaustive exercise bouts relative to time (Monod & Scherrer, 1965; Moritani, Nagata, deVries, & Muro, 1981) can be used to determine CP and the finite work capacity above CP (W’). Whipp et al. (1982) subsequently determined that the linear regression of power output relative to the inverse of time to exhaustion also could determine CP and W’. These linear models are known as the work-time limit (Linear W-T_lim) and power output-inverse time limit (P-T_lim) models, respectively (Gaesser, Carnevale, Garfinkel, Walter, & Womack, 1995). CP is expressed as the slope the linear W-t_lim model, and theoretically, represents the maximum power output that can be maintained for an extended period without exhaustion (Wakayoshi et al., 1992). The parameter W’ is given by the y-intercept of the W-T_lim model, and is defined as the finite energy reserve from the phosphagen pool, anaerobic glycolytic component, and O₂ stores (Poole et al., 1988). Additionally, W’ also may relate to the attainment of a critical concentration of metabolites linked to muscle fatigue (Jones, Wilkerson, DiMenna, Fulford, & Poole, 2008). When applying the concept to running and swimming activities, power is replaced by speed, CP by critical speed (CS) and W’ by maximal capacity to displace the body (D’) (Jones, Vanhatalo, Burnley, Morton, & Poole, 2010).

Traditionally, use of the CP concept required a subject to exercise to exhaustion at several constant power outputs on separate days, and a early debate centered on the number of bouts that should be performed for valid modeling (Hill, 1993). Therefore, in order to reduce the bias in the research outcomes caused by the combination of different mathematical models and predictive trials (amount and duration), a single 3-min all-out test (3MT) was developed to determine CP (Burnley, Doust, & Vanhatalo, 2006) and W’ (Vanhatalo, Doust, & Burnley, 2007). The 3 MT was based on the concept of CP, which implies that at the time point where W’ becomes wholly depleted, the highest achievable power output is CP (Burnley et al., 2006; Coats et al., 2003; Vanhatalo et al., 2007). Namely, if there was a method to completely deplete W’ (e.g. long all-out exercise bout), the remaining power output should equal CP. Indeed, Vanhatalo et al. (2007) observed that end-test power from 3 min of all-out exercise [i.e., average power output over final 30-s (EP)], and the work above EP (WEP) were not significantly different with CP and W’ determined using repeated exhaustive exercise tests, respectively. Estimation of CP from EP was correlated positively and has a low standard error of estimation (SEE), whereas estimation of W’ from WEP had a greater SEE.

The all-out concept has been adapted subsequently for isokinetic knee-extension exercise (Burnley, 2009), rowing (Cheng, Yang, Lin, Lee, & Wang, 2012), and running (Broxterman, Ade, Poole, Harms, & Barstow, 2013; Pettitt, Jamnick, & Clark, 2012). Pettitt et al. (2012) utilized the two-component model equations from the running 3 MT to predict performances of 800, 1600, and 5000 m outdoor races. Runners with higher D’ values in that study ran at faster speeds relative to their CS. Subsequently, Broxterman et al. (2013) compared estimates of CS and D’ from the 3 MT with the “traditional” model using exhaustive treadmill running bouts. The CS and D’ estimates from the 3 MT and those modeled using the speed-inverse t_lim (S-inverse T_lim) model did not differ. When interpreting these results; however, it is worthy to note that estimations of CS between treadmill and track running performances can differ (Kranenburg & Smith, 1996).

The aforementioned studies demonstrated that the 3 MT can be used to determine CS and D’; however, the validation was associated with laboratory tests and no study has reported on the reliability of these variables. In that regard, the objectives of the study were 1) to determine the test-retest reliability of the running 3 MT and 2) to examine if a 3MT could be used to accurately determine CS and D’ derived from track performances.

Mean values of S _max and VO_2max obtained from incremental test were 4.36 ± 0.22 m.s^-1 and 56.6 ± 4.1 ml kg^-1·min^-1, respectively. Performance times for the 800, 1600, and 2400 m distances were 158 ±12 s, 364 ± 34 s and 599 ± 75 s. Summary statistics of the estimated parameters from the 3 MT trials and the linear models are presented in Table 1. For all subjects, the data obtained from the two linear models displayed with a high goodness of fit (R² of 0.955 ± 0.045). The standard errors as a percentage of the parameter estimates of CS and D’ were 1.0 ± 0.7% and 20.7 ± 9.9%, for the linear D-T_lim model and 3.3 ± 2.4% and 13.9 ± 7.4% for the S-inverse T_lim model, respectively.

Figure 1 depicts the determination of CS and D’ from two trials of the 3 MT (left panel) along with linear D-T_lim model (top right panel), and the speed-inverse T_lim model (bottom right panel) for representative subject. Test retest reliability of the CS and D’ estimates were strong (Table 2).

Analysis of CS and D’ revealed a significant main effect (F = 6.925, P=0.045 and F = 10.48, P=0.023, respectively). CS differed between the linear D-T_lim and S-inverse T_lim models (P=0.031), but no differences for CS were observed between the 3 MT trials, or between 3 MTs and the linear models (P>0.2) (Table 1). D’ for the linear models did not differ (P=0.058), but they were significantly higher than D’ estimates from the 3 MT (P<0.05). Small and moderate variations were observed for CS and D’, respectively, between the 3 MT and linear models (Table 2). Moreover, CS and D’ presented high and significant (P<0.01) intraclass correlation coefficients between 3MT and linear methods.

To our knowledge, this is the first study that determined the reliability of the running 3 MT, as well its validity for estimating the parameters of linear relationships from track running performances. The high test-retest reliability observed in the present study demonstrates that the parameters estimated from the running 3MT have minimal variability. Furthermore, our results also indicate that the 3 MT is able to estimate CS, however, underestimates D’. Taken together, our results suggest that 3 MT might be useful tool to assess and monitor training-induced changes in aerobic power, but caution is warranted when using 3 MT for measuring anaerobic capacity through the estimate of D’.

The 3 MT for cycling has been reported to have good predictive validity and reliability for CP, and moderate reliability for W’ (Burnley et al., 2006; Johnson, Sexton, Placek, Murray, & Pettitt, 2011; Vanhatalo et al., 2007; Simpson et. al., 2015) when performed under linear factor resistance applied. Nonetheless, the validity is doubtful, when the resistance model was changed (e.g. isokinetic mode) (Karsten B et al., 2014). For rowing, Cheng et al. (2012) reported that the 3 MT provided moderate test-retest reliability for assessing CP (ICC = 0.788) and W’ (ICC = 0.628). In running, our results also demonstrate that 3 MT is sufficiently robust and not significantly affected by biological and technical variations of the test, since CS and D’ estimated from running 3 MT was reliable. These results are impactful because good reliability implies good precision of single measurements, and better tracking of changes in measurements in research or practical settings.

The CS metric has been described as a good indicator of performance in long-duration events, and a useful parameter for training prescription (Bosquet et al., 2002; Vanhatalo et al., 2011); however, the traditional linear models were time-consuming, and might therefore be a problem for experimental studies and training programs. The 3MT has emerged as a solution to this problem (Broxterman et al., 2013; Cheng et al., 2012; Vanhatalo et al., 2007). For running, Broxterman et al. (2013) reported that the CS determined from the 3 MT was positively correlated (r = 0.92) and not significantly different from the CS determined from three exhaustive treadmill running bouts using the S-inverse T_lim model. Further, they observed steady state VO₂ during a subsequent square-wave bout below CS, whereas VO₂ was driven to VO_2max with exhaustion occurring prior to 15 min in a square-wave bout exceeding CS, as determined from the 3 MT. We used track running performances to determine CS and our results are similar to those of Broxterman, et. al. (2013). Moreover, we observed lower typical errors and coefficient of variations between CS estimated from 3 MT and linear models. Collectively, these studies indicated that the running 3 MT may reduce the bias in the research outcomes caused by different mathematical models, and can be applied successfully to determine the CS in both laboratory and field settings.

A finite energy reserve from anaerobic sources and myoglobin O₂ stores (Miura, Sato, Whipp, & Fukuba, 2000; Monod & Scherrer, 1965; Moritani et al., 1981), and/or the attainment of a critical concentration of one or more metabolites linked to muscle fatigue (Jones et al., 2008), are suggested as mediating factors for W’ and D’. In cycling exercise, Vanhatalo et al. (2007) reported that the W’ from 3 MT was predictive of W’ from the linear models; where others have reported lower reliability for W’ (Johnson et al., 2011) and lower predictive validity W’ and D’ (Broxterman et al., 2013; Cheng et al., 2012). In our study, D’ from the 3 MT correlated positively with D’ estimated from the linear models, yet yielded lower estimates. Similarly, Broxterman et al. (2013) reported that the 3 MT underestimated D’ for five of the seven subjects measured, despite not being statistically different. In addition, Galbraith A. et al., 2014 found higher D’ value in the treadmill protocol compared to the field protocols. Differences in energetic demand between methods may account for the discrepancy between parameters. The high speeds sustained in sprint running, cost of acceleration and aerodynamic resistance are important components of the total energy demand during all-out bouts, which is reduced during middle and long performances relying on the linear models, though still higher than treadmill running (di Prampero, Atchou, Bruckner, & Moia, 1986; di Prampero et al., 1993). In this context, Arsac and Locatelli (2002) reported that the energy required to overcome aerodynamic resistance and changes in kinetic energy were 13% and 24% of the total energy demand of 100 m running performance, respectively. Therefore, D’ might have been lower due to the higher energy required by non-locomotor elements during the 3 MT.

A few studies have shown correlations between W’ and D’ with metabolic measures of anaerobic capacity (i.e. maximally accumulated oxygen deficit and work performed in a 30-s all-out cycle ergometer test (Chatagnon, Pouilly, Thomas, & Busso, 2005; Hill & Smith, 1993; Jenkins & Quigley, 1991). However, poor test-retest reliability of W’ and D’ has been observed (Gaesser & Wilson, 1988; Johnson et al., 2011; Taylor & Batterham, 2002) which may complicate its validity to determine anaerobic capacity. Conversely, D’ estimates in our study had good reliability, probably due to the lack of influence of the high variability of time to exhaustion at a given intensity or performance time inherent in the traditional CS model (Whipp & Ward, 2009). Additional studies are needed to analyze the relationships between D’ estimated from different methods and measures of anaerobic capacities to determine the validity and physiological meaning of D’ from 3 MT.

Our study was unique to prior studies in that we determined time and speed of the 3 MT manually every 20-m, whereas Broxterman et al. (2013) and Pettitt et al. (2012) utilized accelerometer strapped to the right foot and GPS systems to record their data, respectively. Manual timekeeping appears to represent an inexpensive and reliable alternative for estimating CS and D’ from the 3 MT.

In summary, the results our study indicates that estimated parameters from 3 MT have good test-retest reliability. Furthermore, despite that 3 MT underestimated D’, CS was similar between estimates from the 3 MT and traditional linear models. Therefore, we conclude a running 3 MT utilizing manual timekeeping is inexpensive and noninvasive method to evaluate aerobic fitness in running; however, further studies are needed for analyze the validity of D’ from 3 MT as a measure of anaerobic capacity.

Based on the present study, we suggest that the CS from 3 MT is a valid and reliable means to determine aerobic capacity in running. Therefore, a running 3 MT utilizing manual timekeeping can be used as an inexpensive, noninvasive and suitable method for prescribing aerobic training and monitoring training-induced performance changes in aerobic capacity; however, we recommend caution when using D’ from 3 MT as a measure of anaerobic capacity.