A three-minute all-out test performed in a remote setting does not provide a valid estimate of the maximum metabolic steady state

The maximal metabolic steady state (MMSS) demarcates the transition from heavy to severe-intensity exercise (Jones et al. 2019). During severe-intensity exercise, blood lactate concentration, whole-body oxygen consumption (\({{\dot{V}O}}_{2}\)), muscle [H⁺], and muscle [PCr] cannot stabilise, and task failure is characterised by consistent perturbations in these values and attainment of peak \({{\dot{V}O}}_{2}\) (\({{\dot{V}O}}_{2}\)peak). In the heavy domain, these muscle and whole-body parameters attain a delayed steady state (Black et al. 2017). The MMSS has been assessed using the critical power model (Jones et al. 2019). Using ~ 3–5 time-to-task failure severe-intensity trials lasting ~ 2–15 min, critical power is identified as the power-asymptote of the hyperbolic power–duration relationship, and the finite work capacity above critical power as the curvature constant (W′) (Jones and Vanhatalo 2017; Jones et al. 2019). Critical power measured in this manner has been shown to discriminate heavy and severe-intensity exercise responses, and therefore estimate the MMSS (Jones et al. 2008; Black et al. 2017). Work output at the MMSS is used in training intensity regulation, training load monitoring, and predicting endurance performance (Coyle et al. 1988; Maunder et al. 2021).

The requirement for multiple time-to-task failure severe-intensity trials led to development of the three-minute all-out test (3MT) for identification of the MMSS (Burnley et al. 2006; Vanhatalo et al. 2007). In the 3MT, an athlete works all-out for three minutes without pacing, depleting W′ in the initial part of the test to ensure the work output is eventually limited to the critical power. Accordingly, average power output during the final 30 s is used to estimate critical power, and the total work performed above the end-test power is used to estimate W′. The 3MT was designed for use on an electromagnetically braked laboratory ergometer in a linear mode, where power output is the product of the linear factor (flywheel resistance) and the square of the cadence; the linear factor is typically applied, such that the power output achieved at the individual cyclist’s preferred cadence is the gas exchange threshold power output (GET) plus 50% of the interval between GET and \({{\dot{V}O}}_{2}\)peak (Burnley et al. 2006; Vanhatalo et al. 2007, 2008a). The athlete is also supervised and verbally encouraged by a team of researchers, and blinded to power output and time (Burnley et al. 2006; Vanhatalo et al. 2007, 2008a). The 3MT performed in this manner has been shown to produce valid (Burnley et al. 2006; Vanhatalo et al. 2007, 2016) and reliable (Burnley et al. 2006; Johnson et al. 2011) estimates of MMSS and W′ in trained populations.

Since development of the laboratory-based 3MT in 2006, the availability of accurate and reliable power-measuring devices for use during indoor cycle training has increased dramatically (Hoon et al. 2016; Zadow et al. 2016, 2018). As the 3MT only requires accurate measurement of power output and time, it is theoretically possible that the 3MT could be used by athletes in remote settings using personal equipment without supervision for quantification of MMSS and W′. However, it is unknown if the 3MT provides reliable and valid estimates of the MMSS when performed by unsupervised endurance athletes remotely, using typical indoor training set-ups, where there is opportunity to shift gears and therefore resistance to pedalling, and view elapsed time (3MT_R). Addressing this gap in the literature is pertinent for remote endurance coaches operating primarily without face-to-face communication with athletes, and for endurance athletes without regular access to laboratory facilities.

Therefore, the primary aim of the present investigation was to determine the reliability and validity of the 3MT for estimation of the MMSS when performed in remote settings by unsupervised athletes using their own indoor cycling setup. It was hypothesised that the 3MT_R would produce reliable and valid estimates of the MMSS.

The primary aim of the present investigation was to determine the reliability and validity of the 3MT when performed in remote settings by unsupervised athletes using their own indoor cycling setup (3MT_R). The primary outcomes were that end-test power in the 3MT_R was reliable, but overestimated the MMSS. These results suggest that the 3MT_R, as performed in the present investigation, should not be used to estimate the MMSS in endurance-trained cyclists.

The reliability of the 3MT_R is evidenced by the low CV and high Pearson’s correlation coefficients for the primary outcome metrics (Fig. 2, Table 1). The CV values reported for EP_remote (~ 4.0–5.5%) are similar to what has been reported for laboratory-based measurements of end-test power (~ 3–7%) (Burnley et al. 2006; Johnson et al. 2011). The WEP_remote was less reliable than EP_remote (CV, ~ 13.9–16.2%), which is in line with reliability data for laboratory-based estimates (CV, ~ 20–28%) (Johnson et al. 2011). Therefore, these data suggest that the 3MT_R produces similarly repeatable outcomes as the laboratory-based 3MT. However, despite the strong reliability of the 3MT_R, we present robust evidence that EP_remote overestimated the MMSS; this is shown by the severe-intensity physiological responses and short time-to-task failure (11.4 ± 6.5 min; range, 5.0–21.5 min) during constant-work rate cycling at EP_remote, and the consistency of differences between EP_remote and MMSS-\({{\dot{V}O}}_{2}\) and MMSS-[La⁻] at an individual level (Fig. 3). Additionally, we feel the magnitude by which EP_remote overestimated the power outputs at MMSS-\({{\dot{V}O}}_{2}\) (18 ± 11%) and MMSS-[La⁻] (20 ± 12%) permits this conclusion, even in the context of recent data demonstrating the transition from the heavy to severe-intensity domain is a phased transition (Pethick et al. 2020). Therefore, it is unlikely that the overestimation of MMSS by EP_remote is attributable to variability in the results of remote tests.

In the 3MT_R, total work completed above the end-test power was substantially lower (10.3 ± 4.8 kJ) than has typically been observed during laboratory-based trials (~ 14–17 kJ) (Burnley et al. 2006; Vanhatalo et al. 2007, 2008a, 2008b, 2016). Importantly, the sub-group of participants included in the validity aspect of this study had similarly low WEP_remote (9.4 ± 2.6 kJ). In these participants’ work above, the MMSS-\({{\dot{V}O}}_{2}\) power output was similar to work above laboratory-based end-test power observed elsewhere (17.4 ± 3.7 kJ) (Burnley et al. 2006; Vanhatalo et al. 2007, 2008a, 2008b, 2016). Accordingly, it is likely participants in the present investigation failed to fully deplete work output above MMSS in the initial 150 s of the test, and therefore, that EP_remote was supplemented by work output above MMSS. This would explain why subsequent constant-work rate trials at EP_remote produced clear severe-intensity responses, and therefore, why EP_remote significantly overestimated MMSS.

The reason why work above MMSS was not fully depleted in the initial 150 s of the 3MT_R could be attributable to several factors. First, laboratory-based trials are supervised and strong verbal encouragement is provided throughout the test (Burnley et al. 2006; Vanhatalo et al. 2007, 2008a, 2008b). Therefore, in the 3MT_R where participants were unsupervised and strong verbal encouragement was not provided by the researchers, it is possible the absence of social facilitation resulted in a sub-maximal or paced effort, and therefore failure to fully deplete work above MMSS in the first 150 s. Given that the work above the MMSS-\({{\dot{V}O}}_{2}\) power output was similar to work above laboratory-based end-test power observed elsewhere (Burnley et al. 2006; Vanhatalo et al. 2007, 2008a, 2008b, 2016), it is possible that the 3MT_R was paced but still a maximal effort overall (i.e., that total work done across the 180 s was maximal). Had the 3MT_R been performed as a sub-maximal overall effort, it is likely that work above MMSS-\({{\dot{V}O}}_{2}\) would have also been noticeably low. The possibility of pacing within the 3MT_R is supported by the lowest 6 and 30 s power outputs being substantially lower than EP_remote (91 ± 5 and 97 ± 3% of EP_remote, respectively). Periods of cycling below MMSS during the 3MT_R may have allowed partial recovery of the finite capacity for work above MMSS during the test (Skiba et al. 2012), and thus could have inflated the end-test power value and contributed to the overestimation of MMSS. Pacing may have been made more likely by participants being able to view elapsed time during the 3MT_R, which is a key difference compared to laboratory 3MTs (Burnley et al. 2006; Vanhatalo et al. 2007, 2008a, 2008b).

A further key difference was the opportunity for altering the resistance to pedalling in the remote test, compared to the fixed linear factor used in previous work (Burnley et al. 2006; Vanhatalo et al. 2007, 2008a). Peak cadence achieved in the 3MT_R in the present investigation (110 ± 21 revs^.min⁻¹) was substantially lower than those typically achieved in the laboratory test (~ 140–155 revs^.min⁻¹), although end-test cadence was similar (Burnley et al. 2006; Vanhatalo et al. 2008b, a). Therefore, given the similarity in peak power output between our remote and previous studies of laboratory-based 3MTs, it is likely participants in the present investigation self-selected a greater pedalling resistance in the initial part of the test. Differences in the cadence profiles of the 3MT_R investigated here and the laboratory-based 3MT reported elsewhere may also contribute to why work above MMSS was not fully depleted in the initial 150 s of the remote test. It has previously been shown that adjusting the linear factor to produce a higher peak cadence (155 ± 12 vs. 148 ± 15 revs^.min⁻¹) resulted in a significant reduction in end-test power and total work done in a laboratory-based 3MT (Vanhatalo et al. 2008a), time-to-task failure at a constant-work rate in the severe domain was reduced when cadence was experimentally increased by 20 revs^.min⁻¹ (Nielsen et al. 2004), and mean power output during a 30-s all-out sprint was reduced by ~ 15% when performed isokinetically (100 revs^.min⁻¹) compared to isoinertially with a higher peak cadence (117 ± 14 revs^.min⁻¹) (Fuentes et al. 2013). As the rate of metabolic energy expenditure is increased at a given power output at higher pedalling cadences (Umberger et al. 2006; Brennan et al. 2019), it is possible the higher cadences achieved during laboratory trials may be necessary to fully deplete work above MMSS in the initial period of a 3MT, and therefore for end-test power to produce a valid estimate of the MMSS. This may explain why other studies have reported lower end-test power output during laboratory-based 3MTs performed at higher than preferred cadences (Wright et al. 2019), and that critical power is greater when cycling at 60 vs.100 revs^.min⁻¹ (Barker et al. 2006).

Future research may seek to determine if alteration to the 3MT_R instructions used in the present investigation would facilitate full depletion of work above MMSS in the initial component of the test, and therefore provide a valid estimate of MMSS. Speculatively, video conferencing in which researchers or a coach view the test and provide verbal encouragement in real time may help to reduce pacing and would also obviate the need for athletes to be able to view elapsed time. Another strategy may be to partially deplete work above MMSS prior to a 3MT_R with a planned severe-intensity effort. This approach is similar to what has been investigated recently in a laboratory setting, whereby a ramp test is performed prior to a 3MT (Goulding et al. 2021).

In summary, the present investigation suggests that whilst the three-minute all-out test can be performed reliably by endurance-trained cyclists in remote settings using typical indoor training set-ups, these tests overestimated the MMSS, likely due to failure to fully deplete work above end power in the initial 150 s of the test. Therefore, the 3MT_R protocol adopted in the present investigation should not be used for identification of the MMSS. Given that many individual athletes do not have regular access to laboratory facilities (and the on-going possibility of restricted travel to, and use of, laboratories during global pandemics), future research should explore if alterations to the 3MT_R protocol utilised here can produce valid estimates of the MMSS.