The ramp and all-out exercise test to determine critical power: validity and robustness to manipulations in body position

The relationship between power and time to task failure during cycle exercise over durations spanning ~ 2–30 min is well-described by a hyperbolic function (Poole et al. 2016). This power–time relationship is defined by two parameters: critical power (CP); representing the power asymptote of the hyperbola, and W′; the curvature constant of the hyperbola representing a finite amount of work that can be performed above CP (Fukuba et al. 2003). CP represents an important parameter of aerobic function (Poole et al. 1988); separating exercise intensities where a steady state is attainable for pulmonary oxygen uptake (\(\dot{V}\)O₂) and muscle metabolites (i.e. “heavy” intensity domain) from intensities where a steady state is unattainable (i.e. “severe” intensity domain) (Poole et al. 1988; Jones et al. 2008). During sustained exercise above CP, therefore, pulmonary \(\dot{V}\)O₂ is driven towards its maximal value (\(\dot{V}\)O₂ max) (Poole et al. 1988), intramuscular phosphocreatine projects towards a nadir (Jones et al. 2008), and exercise tolerance is predictably limited (Poole et al. 2016).

CP and W′ are both sensitive to exercise training (Gaesser and Wilson 1988; Poole et al. 1990; Jenkins and Quigley 1992; Vanhatalo et al. 2008a) and offer an accurate prediction of endurance performance within the task duration range of ~ 2–30 min, underscoring the importance of these parameters as determinants of endurance performance. Moreover, elite male runners typically sustain 96% of their critical speed (analogous to CP) over the course of a marathon (Jones and Vanhatalo 2017), and critical speed is predictive of marathon performance across a range of abilities (Smyth and Muniz-Pumares 2020). CP and W′ therefore provide invaluable information to endurance performance athletes, coaches and practitioners regarding the physiological and mechanical performance capabilities of an athlete, as well as the efficacy of a given training intervention or ergogenic aid. However, the conventional approach for establishment of CP and W′ requires undertaking 3–5 constant load prediction trials to the limit of tolerance, ideally on separate days, such that confident estimates of the parameters may be obtained (Muniz-Pumares et al. 2018). Precise determination of CP is therefore both time- and labour-intensive for researchers and practitioners alike.

The power–duration relationship predicts that when W′ has been fully depleted (i.e. at task failure), the highest power output that can be sustained is CP (Coats et al. 2003; Chidnok et al. 2013). Hence, Vanhatalo et al. (2007) demonstrated that during the final 30 s of an all-out 3-min bout of cycle exercise, power output plateaued to a work rate that was not different from, and highly correlated with, CP (i.e. end-test power; EP). More recently, Murgatroyd et al. (2014) demonstrated that CP could be accurately and reliably determined from the EP attained during a single exercise test, incorporating a 3-min all-out bout of exercise performed immediately following task failure during a maximal ramp-incremental exercise test, whereas W′ was underestimated by the work above EP (WEP). This approach represents a significant advance over 1- (Clark et al. 2013) or 2-day (Vanhatalo et al. 2007; Bergstrom et al. 2012) testing procedures, because additional parameters of aerobic function (i.e., the gas exchange threshold, GET; mean response time of \(\dot{V}\)O₂ kinetics; \(\dot{V}\)O_2max) and thus the boundaries between moderate, heavy, and severe exercise intensity domains can be determined in a single visit. However, the validity of this test has not been confirmed by more than one study (Murgatroyd et al. 2014), nor have the robustness of its underlying principles been tested using alternative exercise modes. For instance, if EP from the contiguous ramp all-out test can be demonstrated to provide a valid estimate of CP and W′ in an alternative mode of exercise (e.g. supine exercise), this would provide further evidence of the robustness of the underlying principles of this approach for the determination of CP. Such an approach would considerably reduce the burden associated with determination of the power-time parameters for sports and exercise practitioners.

The data presented in the present study comprise retrospective analysis of data from five previous reports (Goulding et al. 2017, 2018a, b, 2019a, b). A total of 26 participants completed the upright exercise experiments (Goulding et al. 2017, 2018a, 2019b) and a total of 16 participants completed the supine exercise experiments (Goulding et al. 2018b, 2019a). However, several participants completed more than one of the original experiments, therefore in these instances only the data from the first experiment that the participant took part in was used for further analysis. Thus, 19 [mean ± standard deviation (SD); age: 28 ± 8 years; height: 181 ± 7 cm; body mass: 78 ± 9 kg] and 12 [age: 24 ± 5 years; height: 179 ± 9 cm; body mass: 79 ± 8 kg] healthy, recreationally active males took part in the upright and supine portions of the study, respectively, with 2 participants completing both upright and supine arms of the study. All participants provided written informed consent and all experiments received ethical approval from the Liverpool Hope University Research Ethics Committee. Participants were asked to avoid alcohol and strenuous exercise 24 h prior to each visit, not to consume caffeine 3 h prior to each visit, and to arrive 3 h postprandial. Tests were separated by at least 24 h, with each test performed at the same time of day (± 2 h).

The present study determined whether a contiguous ramp all-out exercise test could determine CP and W′ in a single laboratory visit during both upright and supine cycle exercise. The findings of the present study show that CP conventionally estimated from a series of severe-intensity prediction trials was not different from, and highly correlated with the EP from the contiguous ramp all-out exercise test during both modes of exercise, suggesting that the ramp all-out test may be used to determine CP in a single visit. However, WEP was significantly different from W′ during both modes of exercise. This suggests that the ramp all-out exercise test is not appropriate for characterising W′, and thus, for instance, predicting endurance performance of 2–20 min duration. Furthermore, the wide 95% LoA (i.e. − 14, 9% of CP for upright, and − 11, 17% of CP for supine exercise) between CP and EP suggests that the two parameters should not be used interchangeably.

Murgatroyd et al. (2014) designed and validated the contiguous ramp all-out test utilised in the present study to enable the precise characterisation of the power–duration relationship in a single test. These authors demonstrated close agreement with CP determined via gold-standard procedures: EP and CP were not significantly different, and the 95% LoA between the two measures were ± 6% of CP. During exercise 10 W below EP in validation trials, all participants were able to complete 30 min of cycling and \(\dot{V}\)O₂ attained a delayed steady state, whereas during exercise 10 W above EP, participants attained \(\dot{V}\)O₂ max and task failure occurred within ~ 19 min (Murgatroyd et al. 2014). The purpose of the present study was therefore to independently validate these findings by determining whether EP could provide similarly precise estimates of CP during both upright and supine cycling. Our results are somewhat in agreement with those of Murgatroyd et al. (2014). Specifically, we found no difference between EP derived from the ramp all-out test and CP determined via gold-standard procedures in both body positions, and these parameters were highly correlated in both groups. Furthermore, the CV between EP and CP was less than the minimal level of error with which the gold-standard procedures used to determine CP are typically associated with in both body positions, i.e. 5% (Hill 1993; Mattioni Maturana et al. 2017). These findings therefore indicate a high level of agreement between EP and CP and suggest that the ramp all-out test is a valid tool to determine CP in a single laboratory visit. However, the 95% LoA between CP and EP were somewhat wide in both body positions (upright: − 30, 19 W or − 3 ± 11% of CP, supine: − 16, 24 W or 3 ± 15% of CP), a finding inconsistent with those of Murgatroyd et al. (2014). In this group of participants, therefore, it can be said that for any given individual there was a 95% probability that the difference between EP and CP would be between − 14% and + 9% (or − 12% and + 18% for supine exercise) of the mean of both measurements (Atkinson and Nevill 1998). This degree of error is somewhat larger than that typically deemed acceptable for estimation of CP with gold-standard modelling procedures (i.e. 5%, Hill 1993; Mattioni Maturana et al. 2017; Muniz-Pumares et al. 2018), therefore these finding suggest that CP and EP should not be used interchangeably. Vanhatalo et al. (2008a) previously demonstrated that a 4-week exercise training intervention resulted in a 25 W or 10% increase in CP. As the 95% LoA between EP and CP overlaps this change in both body positions in the present study, our data suggest that the ramp all-out test-determined EP is not sufficiently sensitive to monitor training-induced changes in CP. Moreover, 4/19 and 4/16 participants in the upright and supine positions were unable to maintain their \(\dot{V}\)O₂ above 95% of their \(\dot{V}\)O₂ peak during the all-out phase. This observation highlights the fact that, due to the difficulty of the ramp all-out test, participants must be highly motivated to complete it. However, given appropriate motivation, if the goal is to prescribe exercise within the heavy or severe domains, this may now be achieved in one test compared to the previously typical five (i.e. one incremental ramp test and four constant work-rate prediction trials), provided the tests are not in extremely close proximity to EP (i.e. not within ± 9–18% EP). The reason for the discrepancy regarding the 95% LoA between the present study and (Murgatroyd et al. 2014) is presently unclear, however, as no other study has examined the validity of the ramp all-out test.

WEP systematically underestimated W′ during upright exercise, whereas WEP was greater than W′ during supine exercise. The physiological determinants of W′ are still not well understood (Poole et al. 2016), therefore the reasons for these discrepancies are not currently clear. However, Murgatroyd et al. (2014) also found that WEP during the all-out phase of the ramp all-out test was lower than W′, and the WEP during the standalone 3 min all-out test is typically lower than W′ (Vanhatalo et al. 2007). Moreover, CP and W′ derived from constant work-rate prediction trials have been shown to overestimate ramp exercise performance (Black et al. 2016), suggesting that W′ may be reduced during ramp exercise. A reduction in W′ in ramp and/or all-out exercise versus constant work-rate exercise could therefore potentially account for the underestimation of W′ by WEP in the upright group in the present study. The reasons for the overestimation of W′ by WEP in the supine position are less clear, but may simply be related to random error, given the 95% LoA between the two measurements was equivalent to 81.25% of W′. Irrespective of the reason for the overestimation of W′ by WEP in the supine position, a major attraction of the power–time relationship is that once CP and W′ are known, it becomes possible to predict endurance exercise performance with high precision for exercise durations spanning 2–30 min. However, the inability of WEP to provide accurate estimates of W′ in the present study limits the use of this test to predict endurance performance.

One potential source of error that may have contributed to the wide 95% LoA in the present study is the fact that the linear factor must be set a priori before knowledge of a participant’s GET (Bergstrom et al. 2012; Clark et al. 2013). This is crucial because CP is sensitive to manipulations in cadence during variable power exercise (Vanhatalo et al. 2008b). The chosen linear factor must therefore be selected such that the cadence produced at the end of the all-out phase (i.e., when riding close to EP) elicits a power output that is close to the as yet unknown CP. In the present study, we utilized the same procedures as Murgatroyd et al. (2014), wherein the linear factor was set according to the previously determined relationship between CP and body mass (van der Vaart et al. 2014). In this study, CP approximated three times body mass in a cohort of 60 healthy active men; however, the relationship was inherently variable (R² = 0.32). Variability in CP related to other factors, such as training status, is thus not considered in this estimation, and this likely translates to error in determination of CP from EP using the ramp all-out test. However, whilst these considerations likely explain a significant proportion of the error inherent in estimating CP from the ramp all-out test, our procedures were the same as those of Murgatroyd et al. (2014) and thus the reasons for the discrepancies between the data presented in that study and herein remain unclear. The protocol presented by (Constantini et al. 2014) would be a useful method to mitigate this concern, as these authors utilized a ramp and all-out exercise test separated by 20 min to determine CP and W′. This would allow time for estimation of the GET and thus selection of the appropriate linear factor for the all-out test. Future work in this area should validate this protocol to determine whether or not reliable and valid estimates of CP and W′ can be obtained.

CP and W′ are highly relevant for athletic performance as they conflate to determine the tolerable duration of exercise over the range of 2–30 min. CP and W′ can thus be used to prescribe training intensity, monitor the distribution of training intensity, predict competition performance, and monitor the performance of athletes over time. The ramp all-out test proposed by Murgatroyd et al. (2014), thus represents an attractive alternative to the previously typical 3–5 constant work-rate trials needed to precisely determine the power-time relationship. However, this is the first study to attempt to independently validate the ramp all-out test. The data presented herein suggest that the ramp all-out test may be used to prescribe heavy- or severe-intensity exercise for training or testing purposes, provided the intensities selected differ from EP by an appreciable margin (e.g. ± 9–18% EP using the data of the present study). For instance, it is common practice to prescribe heavy-intensity exercise as 40% of the difference between \(\dot{V}\)O₂ max and the GET. However, this method does not take into account inter-individual variation CP and is thus liable to error. Using the ramp all-out test, greater confidence that a prescribed intensity is definitively heavy could be achieved by prescribing a work rate that is 50% of the difference between the GET and EP, for example. However, the wide LoA between CP and EP, along with the inability of WEP to predict W′, suggests that the test should not be used for performance prediction and/or athlete monitoring over time due to the unacceptably large error involved in estimation of the power–time relationship from the ramp all-out test. It is therefore recommended that traditional constant work-rate prediction trials are used in favour of the ramp all-out test in applied and practical settings for these purposes. Future research should test the robustness of these findings in various populations, e.g. athletes and patient populations.