Journal Logo

SPECIAL COMMUNICATIONS: Methodological Advances

Translating Ramp V˙O2 into Constant Power Output: A Novel Strategy that Minds the Gap


Author Information
Medicine & Science in Sports & Exercise: September 2020 - Volume 52 - Issue 9 - p 2020-2028
doi: 10.1249/MSS.0000000000002328


Ramp incremental (RI) exercise tests are extensively used to assess aerobic fitness levels in a wide variety of healthy and clinical populations. These tests, characterized by a continuous and linear increase in exercise intensity (e.g., power output [PO] in cycling), have gained increasing popularity since their introduction in the early 1980s (1,2). In contrast to constant work rate (CWR) or step incremental tests, the main advantage of RI tests is time efficiency, thanks to the short duration (i.e., typically ~8 to 12 min) and the possibility to contemporarily evaluate both the submaximal and the maximal performance capacity (3).

Despite the benefits, we should also be mindful about features that complicate the practical use of RI tests. One particular problem is that none of the physiological parameters reach a steady state because of the continuously changing metabolic demands. This means that the measured V˙O2 will lag the true metabolic needs (i.e., the actual steady-state V˙O2) for any given PO. As a result, every PO imposed during CWR exercise will elicit a higher V˙O2 than measured at the same PO during RI exercise. The magnitude of this “gap” between the V˙O2/PO relationship obtained during RI versus CWR exercise depends on the individual V˙O2 kinetics and on different methodological factors (e.g., ramp slope, pedal rate) (2,4).

The practical relevance of the above is that, to accurately prescribe a task-specific PO based on the RI V˙O2 response, scientists and coaches should “mind the gap.” Failing to account for this discrepancy leads to the prescription of exercise intensities that elicit a higher metabolic intensity than intended. This could influence the subject’s response and adaptation to exercise and has a bearing for the variability in the responsiveness to training (5) and the interpretation of data, which might therefore result in distorted conclusions.

In the moderate-intensity domain, the lag in the V˙O2 response during RI exercise has been defined as the mean response time (MRT) and reflects the circulatory transit delay between the muscles and the lungs and the time constant (τ) inherent to the V˙O2 kinetics (3). The classic way to quantify the MRT is to identify the intersection of two regression lines (i.e., baseline V˙O2 and the first slope of the V˙O2/time relationship) (i.e., MRTLIN) (3,6). The intersection point then corresponds to the time interval before the systematic rise in V˙O2. However, more recently, the MRT has been quantified using a new approach in which the RI test is preceded by an additional CWR bout at 100 W (7). The MRT (expressed in PO) was then calculated as the difference in PO between 100 W and the ramp-identified PO (POramp) that elicited the same V˙O2 as obtained during the 100-W bout (i.e., MRTSS). In general, left-shifting RI V˙O2 data by a time interval that is equal to the MRT has been demonstrated to be a valid approach for PO selection below the gas exchange threshold (GET) (4). This method is consistently applied in most exercise physiology laboratories with a focus on oxidative metabolism (6,8,9). However, the use of the MRT correction in practice (e.g., by coaches or physiotherapists) is less common.

On the contrary, because of the appearance of the V˙O2 slow component (V˙O2sc) (10,11), accounting for the MRT seems to be inadequate for PO selection outside the moderate-intensity domain (4). During CWR exercise within the heavy-intensity domain, the V˙O2sc will drive the V˙O2 response to a higher and delayed steady state after ~10 to 15 min due to a loss of mechanical efficiency (12). However, because of the non–steady-state circumstances during RI exercise, the V˙O2sc does not have the ability to fully manifest. Boone and Bourgois (3) reported that only ~50% of the individuals performing RI exercise display a steeper slope of the V˙O2/PO relationship above GET (i.e., a V˙O2 excess). As a consequence, the dissociation in the V˙O2/PO relationship between RI and CWR exercise becomes amplified, especially when steeper ramp slopes are used. Indeed, Iannetta et al. (13) demonstrated that the visibility of a V˙O2 excess during RI exercise strongly depends on the ramp slope as the slope of the V˙O2/PO relationship increased in slower RI protocols. This supports the idea that slow RI protocols may result in a better approximation of the true V˙O2/PO relationship (3,14).

A recent review from Keir et al. (4) presented a conceptual model suggesting that the gap between the RI and the CWR V˙O2 response grows in a nonlinear fashion outside the moderate-intensity domain. However, no previous studies have attempted to quantify the discrepancy in V˙O2 in terms of PO nor to develop a translation model that accounts for the loss of mechanical efficiency within the heavy-intensity domain. As a consequence, the accurate translation of the RI V˙O2 response into a PO for heavy CWR exercise remains elusive. In this study, we aimed to model the dissociation in the V˙O2/PO relationship between RI and CWR exercise by comparing the V˙O2 profile of two RI tests with a different ramp slope (i.e., 30 and 15 W·min−1) to the end-exercise V˙O2 of CWR bouts at different exercise intensities. We hypothesized that within the moderate-intensity domain, the dissociation in PO between RI and CWR exercise would correspond to the MRT, regardless of the MRT calculation method that was used (i.e., MRTLIN or MRTSS). However, we hypothesized that once within the heavy-intensity domain, the MRT correction would be too small to compensate for the loss of mechanical efficiency. In this context, two new translation strategies that compensate for the loss of mechanical efficiency within the heavy-intensity domain are presented and tested.



Nine healthy young men (26 ± 2 yr, 1.78 ± 0.06 m, 77.5 ± 15.5 kg) volunteered to engage in this study. On the basis of a medical examination, the subjects were declared to be in good health and were free of any medical conditions that could influence the cardiovascular, cardiorespiratory, or metabolic response to exercise. All participants signed a consent form after they were informed about the purpose and the design of the experiment. The study was approved by the Ethical Committee of the Ghent University Hospital (Ghent, Belgium).

Experimental Procedure and Protocols

In total, subjects completed two RI tests and seven or more CWR tests. All tests were conducted in the laboratory on an electromagnetically braked cycle ergometer (Excalibur sport, Lode, Groningen, the Netherlands) of which the saddle and the handlebar position were adjusted individually. Participants were instructed to avoid strenuous exercise and standardize their food and drink intake the day before and on the day of a test and restrain from caffeine intake during test days. Test sessions were performed at a fixed time of the day with a minimum of 48 h of recovery in between. During all tests, heart rate (HR) was monitored (HRM-Dual, Garmin, Olathe), and gas exchange (V˙O2, V˙CO2, and V˙E) was measured breath by breath using a portable metabolic system (K5, Cosmed, Rome, Italy) that was calibrated before each test.

RI tests

In random order, subjects performed a 30-W·min−1 and a 15-W·min−1 RI test to exhaustion. Each protocol started with 15 min of cycling at 100 W (i.e., to determine the MRTSS, see Data Analysis section), 2 min of seated rest, and an additional 4 min of baseline cycling at 50 W before the onset of the ramp. Participants were asked to choose a fixed cadence between 70 and 90 rpm. This self-selected cadence was imposed during all subsequent test sessions. The test was stopped when subjects could no longer keep up their preferred cadence (±5 rpm) for more than five consecutive seconds despite strong verbal encouragement.

CWR tests

Participants performed several CWR tests to determine the end-exercise V˙O2 at seven different intensities: 1) GET −10 W, 2) GET, 3) GET +10 W, 4) halfway between GET and maximal lactate steady state (MLSS) (∆GET-MLSS), 5) MLSS −10 W, 6) MLSS, and 7) MLSS +10 W. Each test started with a 4-min warm-up at 50 W, immediately followed by an abrupt increase to the appropriate PO, which was sustained for 15 min at GET −10 W and GET (i.e., moderate intensity), for 30 min at GET +10 W, ∆GET-MLSS, MLSS −10 W, and MLSS (i.e., heavy intensity), and to exhaustion at MLSS +10 W (i.e., severe intensity). At the end of every trial, subjects were asked to rate their perceived exhaustion (RPE) using the original 6–20 Borg scale (15).

The PO at MLSS was initially predicted using a mathematical equation from Iannetta et al. (16) and set as the intensity for the first CWR test (see Data Analysis section). At the end of the warm-up and from then every 5 min, capillary blood samples were taken from the fingertip to measure the blood lactate concentration ([La]) (Biosen C-Line, EKF blood Diagnostics GmbH, Berlin, Germany). If the [La] between the 10th and the 30th min of the test increased >1 mmol·L−1, the PO for the next test was decreased by 10 W. If the [La] increased ≤1 mmol·L−1, the next test was performed 10 W higher. This procedure was repeated until the end-exercise V˙O2 at MLSS −10 W, MLSS, and MLSS +10 W could be determined. In this context, MLSS was identified as the highest PO associated with a stable [La] response (i.e., an increase ≤1 mmol·L−1 during the final 20 min of the test) (17). After determining the PO at GET (see Data Analysis section), subjects completed the remaining four tests (i.e., at GET −10 W, GET, GET +10 W, and ∆GET-MLSS) in a randomized order.

Data Analysis

Raw V˙O2 data were filtered by removing data points that lay three or more SD from the local mean, linearly interpolated on a second-by-second basis (18,19), and then averaged into 10-s bins (Origin 2019b; OriginLab, Northampton, MA).

RI tests

Peak PO (POpeak) and peak HR (HRpeak) were defined as the highest values achieved at the end of each test. V˙O2peak and peak respiratory exchange ratio (RERpeak) were expressed as the highest 30-s average throughout.

GET (expressed in V˙O2) was determined using the V-slope method (i.e., the intersection of a two-line regression of V˙CO2 vs V˙O2) (20) and the secondary criteria (i.e., the first departure from the linear increase in V˙E, an increase in V˙E/V˙O2 without a simultaneous increase in V˙E/V˙CO2, and the first rise in PETO2) (21). The respiratory compensation point (RCP) (expressed in V˙O2) was determined as the point where both V˙E/V˙O2 and V˙E/V˙CO2 increased, the second departure from the linear increase in V˙E, and the deflection point of PETCO2 (21,22).

For each individual, the V˙O2/PO relationship for the 30- and 15-W·min−1 RI protocol was modeled using two linear regression lines (Fig. 1) (3):

Visualization of the group mean V˙O2/PO relationship for RI and CWR exercise. Gray and white dots display the modeled V˙O2 response to the 30-W·min−1 and 15-W·min−1 RI protocol, respectively. Both V˙O2 profiles are drawn up to the Ppeak that was reached by the weakest subject. V˙O2peak is indicated by separate dots at the top right. Dark dots show the end-exercise V˙O2 at nine different PO, including 50 W, 100 W, GET −10 W, GET, GET +10 W, ∆GET-MLSS, MLSS −10 W, MLSS, and MLSS +10 W. Two linear regression lines were fit through these data points to model the V˙O2/PO relationship for CWR exercise in the moderate- and the heavy-intensity domain (black lines). Until the level of GET, the V˙O2 profiles of both RI and CWR exercise evolve in parallel with a delay that approximates the MRT, but once above, the gap in V˙O2 further increases. This is illustrated by a higher s 2 − CWR than s 1 − CWR (mL·min−1·W−1).

where y is the RI V˙O2, x is the PO, b is the y-intercept, and s1 − ramp and s2 − ramp correspond to the ramp slopes of the first and second portion, respectively. On the basis of a visual inspection, data points that lay before the start of the systematic rise in V˙O2 or that were related to the plateau in V˙O2 toward the end of the test were not included in the analysis.

Two different approaches were used to compute the MRT for both RI protocols:

  • 1. The traditional method using a segmented linear model (i.e., MRTLIN) (3):
  • MRTLIN (s) was defined as the time interval between the onset of the RI test and the intersection of the forward extrapolation of the baseline V˙O2 and the backward extrapolation of the V˙O2/time relationship below GET. Baseline V˙O2 was defined as the average V˙O2 during the last 2 min of baseline cycling at 50 W. Regression line [1] was modified and used to plot the RI V˙O2 response as a function of time.
  • 2. An alternative method using the steady-state V˙O2 of an additional CWR bout at 100 W (i.e., MRTSS) (7):
  • MRTSS (W) was defined as the discrepancy between 100 W and the POramp associated with the steady-state V˙O2 at 100 W. The steady-state V˙O2 at 100 W was calculated as the average V˙O2 during the last 5 min of the 15-min work bout at 100 W preceding the RI test. To retrieve POramp, equation 1 was solved for x by entering this V˙O2.

All MRT values were expressed in both time (s) and PO (W) by making the appropriate conversion (i.e., 1 W = 2 s for the 30-W·min−1 protocol and 1 W = 4 s for the 15-W·min−1 protocol). To determine the PO that would elicit a steady-state V˙O2 corresponding to GET, the POramp at GET was calculated using equation 1 and then corrected by means of the MRTSS.

CWR tests

The predictive equation that was used to predict MLSS (16) and set the PO for the first CWR test was:

with V˙O2peak and RCP derived from the 30-W·min−1 RI protocol.

For all CWR tests, the end-exercise V˙O2 was calculated as the mean V˙O2 during the final 5 min of exercise. However, in case of the MLSS +10 W trial, the end-exercise V˙O2 was defined as the highest 30-s value at the end of the test. The POramp that was associated with the end-exercise V˙O2 of each trial was calculated by solving equation 1 if the end-exercise V˙O2 ≤ GET or equation 2 if the end-exercise V˙O2 > GET. In this way, the discrepancy in PO between RI and CWR exercise (i.e., the error made during translation) could be determined for all intensities.

Analogous to the RI tests, the V˙O2/PO relationship for CWR exercise was established using two linear regression lines (Fig. 1):

where y is the CWR V˙O2, x is the PO, b is the y-intercept, and s1 − CWR and s2 − CWR correspond to the CWR slopes of the first and second portion, respectively. Data points that were included in the first portion were the average baseline V˙O2 at 50 W and the steady-state V˙O2 at 100 W (i.e., derived from the RI tests), as well as the end-exercise V˙O2 at GET −10 W and GET. The second portion consisted of the end-exercise V˙O2 at GET +10 W, ∆GET-MLSS, MLSS −10 W, and MLSS. In this way, s1 − CWR and s2 − CWR correspond to the gain in V˙O2 (i.e., ∆V˙O2/∆PO) within the moderate- and heavy-intensity domain, respectively.

Translation strategies

Each approach presented below aimed to translate a specific RI target V˙O2 into a PO for CWR exercise. In this light, the PO and the associated end-exercise V˙O2 of the CWR trials were considered as the target values (i.e., POtarget and V˙O2target). Thus, to elicit a desired metabolic intensity (V˙O2target), the goal of each translation strategy was to yield PO estimates (POest) that predict POtarget. In the present study, three different approaches were used to calculate POest within the moderate- and the heavy-intensity domain. Whereas strategy 1 comprised a simple left shift of the RI V˙O2/PO relationship (i.e., accounting for the MRT), we developed two novel methods (i.e., strategies 2 and 3) in which an additional adjustment is made for PO selection within the heavy-intensity domain. The rationale is that, supplementary to the traditional MRT correction, these strategies also account for the individual loss of mechanical efficiency that occurs at higher intensities during CWR exercise. Because strategies 2 and 3 apply a correction that differs between both intensity domains, all CWR trials were recategorized in three intensity zones based on the relative occurrence of V˙O2target to GET and RCP. If V˙O2target ≤ GET, the trial was classified in zone 1 (i.e., moderate-intensity domain); if GET < V˙O2target ≥ RCP, the trial was classified in zone 2 (i.e., heavy-intensity domain); and trials for which V˙O2target > RCP were classified in zone 3 (i.e., severe-intensity domain). Trials in zone 3 were not further included because in theory, any PO selection for a V˙O2target above RCP would eventually drive the V˙O2 to its maximum value (4).

  • Strategy 1: simple MRT correction

where POramp corresponds to the ramp-identified PO for any given metabolic load. This strategy uses an identical correction for PO selection in both intensity zones. In zone 1, this approach was performed twice, once using MRTLIN and once using MRTSS. Only the method that was associated with the smallest error was selected for further use in the translation strategies. In case both methods displayed a similar error, MRTLIN was retained as the standard MRT procedure (see Results section).

  • Strategy 2: simple MRT correction + extra correction based on the difference between s2 − CWR and s2 − ramp:

The equation for zone 2 can be simplified as follows:

This strategy applies a simple MRT correction in zone 1 but makes an additional correction for PO selection in zone 2. This extra adjustment is intended to account for the loss of mechanical efficiency based on the difference between s2 − CWR and s2 − ramp. In fact, we assume that s2 − CWR will be larger than s2 − ramp because of the lack of decrease in mechanical efficiency during RI exercise. By accounting for this difference, strategy 2 “minds the gap” that exists between the RI and the CWR V˙O2/PO relationship in the heavy-intensity domain.

  • Strategy 3: simple MRT correction + extra correction based on the ratio s2/s1:

The equation for zone 2 can be simplified as follows:

In accordance with strategy 2, this strategy applies a normal MRT correction in zone 1 and adds an extra correction in zone 2. However, the nature of the adjustment is different. Instead of considering the gap in V˙O2 that exists within the heavy-intensity domain, this approach corrects for the loss of mechanical efficiency by comparing the mechanical efficiency in the heavy-intensity domain relative to the moderate-intensity domain. The difference in the ratio s2/s1 for both the RI and CWR V˙O2/PO relationship will then determine the size of the extra correction that is needed.

Statistical Analysis

All statistical analyses were performed with SPSS Statistics 24 (IBM Corp., Armonk, NY). Paired-samples t-tests were used to detect differences between the 30-W·min−1 and the 15-W·min−1 RI protocol and between the slopes of the RI and the CWR V˙O2/PO relationship. A two-way repeated-measures ANOVA (determination method–intensity zone) was conducted to compare POtarget, POramp, and POest in zone 1 and zone 2. For all translation strategies, Bland–Altman plots were created to determine the bias and the limits of agreement (LoA) between POest and POtarget. All data are presented as mean ± SD. Statistical significance was accepted at P < 0.05.


An overview of the mean performance and the cardiorespiratory response to both RI tests is presented in Table 1. Time to exhaustion (TTE) was shorter and POpeak was higher for the 30-W·min−1 than the 15-W·min−1 protocol (P < 0.001). Although, s2 − ramp was higher in the 15-W·min−1 protocol (P = 0.003), at the group level, no differences between S1 and S2 within the same RI protocol were found (P > 0.05).

Overview of the mean performance and the mean cardiorespiratory response to both RI tests.

Table 2 shows the mean PO, end-exercise V˙O2, HR, and RPE of all CWR trials. The end-exercise V˙O2 at GET did not differ from the V˙O2 at GET as visually determined from the 30-W·min−1 (P = 0.161) and the 15-W·min−1 (P = 0.058) protocol. Furthermore, the end-exercise V˙O2 at MLSS also did not differ from the ramp V˙O2 at RCP in both RI protocols (P = 0.652 and P = 0.753). The mean TTE at MLSS +10 W was 37 ± 6 min with an end-exercise V˙O2 that was significantly lower than the average V˙O2peak of both RI tests (P = 0.002).

Mean PO, end-exercise V˙O2, HR, and RPE for seven different CWR intensities, supplemented with the ramp-identified PO (POramp) that correspond to each end-exercise V˙O2.

Without any correction, all POramp values that corresponded to the end-exercise V˙O2 response were consistently overestimated (P < 0.001) (Table 2). However, these differences were smaller for the 15-W·min−1 compared with the 30-W·min−1 protocol (P = 0.012), except for the trial at GET −10 W (P = 0.121).

Figure 1 gives a complete overview of the experimental data and visualizes the group mean discrepancy in V˙O2 between RI and CWR exercise. S2 − CWR (14.2 ± 2.4 mL·min−1·W−1) was significantly higher than s1 − CWR (10.0 ± 1.22 mL·min−1·W−1) (P = 0.004). S1 − CWR did not differ from S1 − ramp30 (P = 0.297) and S1 − ramp15 (P = 0.994), but S2 − CWR was significantly higher than S2 − ramp30 (P < 0.001) and S2 − ramp15 (P = 0.009). At the group level, no differences between s1 and s2 within the same RI protocol were found (P > 0.05).

Translation strategies

Strategy 1

Figure 2 shows no bias between POtarget and POest when MRTSS (P = 0.459) and MRTLIN (P = 0.786) were applied in zone 1. Moreover, the bias that was associated with the MRTLIN correction (0.4 ± 7.3 W, LoA: lower = −14.0 W, upper = 14.7 W) did not differ from the MRTSS correction (0.7 ± 4.9 W, LoA: lower = −8.9 W, upper = 10.2 W) (P = 0.800). Because both MRT methods generated a similar error, MRTLIN was chosen as the only MRT approach to be incorporated in all further translation strategies.

Bland–Altman plots showing the agreement between POtarget and POest for zone 1 using two different MRT calculation methods (i.e., MRTLIN and MRTSS). Black and white dots correspond to data points from the 30-W·min−1 (n = 16) and the 15-W·min−1 (n = 14) protocol, respectively. For each MRT calculation method, the differences in PO (y-axis) are plotted against the average PO (x-axis), showing the mean bias (black line) and the 95% limits of agreement (dashed lines).

When strategy 1 was applied to zone 2, a substantial bias of 17.1 ± 15.9 W (LoA: lower = −14.1 W, upper = 48.3 W) was revealed (P < 0.001) (Fig. 3A). A comparable result was visible when the Bland–Altman analysis was performed on combined data for zones 1 and 2. Across both intensity zones, strategy 1 resulted in an overall bias of 12.5 ± 15.9 W (LoA: lower = −18.7 W, upper = 43.6 W) (P < 0.001).

Bland–Altman plots showing the agreement between POtarget and POest for zone 2 and for combined data of zone 1 and 2 using three different translation strategies (i.e., strategies 1, 2, and 3). For each strategy, the differences in PO (y-axis) are plotted against the average PO (x-axis), showing the mean bias (black line), the 95% limits of agreement (dashed lines), and the regression line (gray lines). Black and white dots correspond to data points in zone 2 from the 30-W·min−1 (n = 37) and the 15-W·min−1 (n = 41) protocol, respectively. Gray dots show all data points in zone 1 (n = 30). Only strategy 2 was found to accurately predict POtarget across both intensity zones.
Strategy 2

This approach showed a strong agreement between POtarget and POest (Fig. 3B). The average error of 2.9 ± 16.1 W in zone 2 was not different from zero (LoA: lower = −28.5 W, upper = 34.4 W) (P > 0.110). This nonsignificant bias was also retained when strategy 2 was applied across both intensity zones (2.2 ± 14.2 W, LoA: lower = −25.6 W, upper = 30.0 W) (P > 0.107).

Strategy 3

The mean bias for strategy 3 averaged 5.1 ± 18.8 W (LoA: lower = −31.7 W, upper = 41.9 W) in zone 2 and 3.8 ± 16.5 W when it was applied to both zones (LoA: lower = −28.6 W, upper = 36.1 W) (Fig. 3C). Although small, these deviations were found to be significant (P = 0.019).

A general comparison of the mean values for POtarget, POramp and POest is presented in Table 3. POramp consistently overestimated POtarget in both intensity zones (P < 0.001), but the application of each translation strategy in zone 1 was able to eliminate these differences (P > 0.05). In zone 2, POest and POtarget were different using strategy 1 (P < 0.001) and strategy 3 (P = 0.019), but strategy 2 generated POest values that did not differ from POtarget (P = 0.11).

Comparison of the target PO (POtarget) and the estimated PO (POest) in zone 1 and 2 using no correction strategy (i.e., POramp), strategy 1 (i.e., simple MRT correction), strategy 2 (i.e., correction for the MRT and the loss of mechanical efficiency via the difference between s 2 − CWR and s 2 − ramp), and strategy 3 (i.e., correction for the MRT and the loss in mechanical efficiency via the ratio s 2/s 1).


The main goals of this study were to investigate the gap in the V˙O2/PO relationship between RI and CWR exercise and to develop a translation strategy in an attempt to resolve this discrepancy. In accordance with our hypothesis, it was demonstrated that correcting for the MRT (i.e., strategy 1) is an appropriate method to define a PO for moderate CWR exercise. The calculation method of this MRT (i.e., MRTLIN or MRTSS) did not influence the accuracy of estimation. However, for PO selection within the heavy-intensity domain, we found that a simple MRT correction resulted in an overestimation of POtarget and, thus, was no longer sufficient. In an attempt to eliminate this gap, two new translation strategies that correct for the individual loss of mechanical efficiency (i.e., strategies 2 and 3) were tested. It was shown that only strategy 2 was successful to accurately translate the RI V˙O2 response into a PO for CWR exercise, irrespective of the target intensity zone (i.e., moderate or heavy) or the RI protocol that was used (i.e., 30 or 15 W·min−1).

The V˙O2sc is traditionally considered as an increased O2 cost that results predominantly from a decreased efficiency of muscle contractions (11). Although its appearance during CWR is explicit, the manifestation of the V˙O2sc during RI exercise (in the form of a V˙O2 excess) is less univocal. In the present study, s2 − ramp was higher than s1 − ramp in the 15-W·min−1 protocol (11.21 vs 9.99 mL·min−1·W−1), but this was not the case for the 30-W·min−1 protocol (9.53 vs 9.54 mL·min−1·W−1). Although this seems remarkable at first sight, previous research already showed that the V˙O2sc during RI exercise may be difficult to detect. Boone and Bourgois (3) reported that only 47% of a group of physically active subjects displayed a steeper slope in the V˙O2/PO relationship above GET, and that in 18% of the subjects, this slope was even less steep. However, just because the V˙O2sc is not detectable does not mean that a loss of mechanical efficiency is not present. Presumably, the rate of the PO increase in the 30-W·min−1 protocol was too fast to enable the V˙O2sc to become visible in our subject group. This is further supported by the fact that an excess in V˙O2 was more visible in the slower 15 W·min−1 protocol. On an individual basis, s2 − ramp was higher in, respectively, three and seven subjects during the 30- and 15-W·min−1 protocol. Given this individual variability together with the apparent variation in V˙O2 excess between protocols, a relationship between the actual V˙O2sc and the individual loss of mechanical efficiency during RI exercise seems rather unlikely, regardless their common physiological meaning.

Moderate-intensity domain

For all intensities, we found significant differences in PO between RI and CWR exercise. At the level of GET, the magnitude of this dissociation was 35 ± 15 W for the 30-W·min−1 protocol and 23 ± 11 W for the 15-W·min−1 protocol. These differences are quite substantial (i.e., ~20%), and therefore, it stands to reason that sport and health practitioners can make large mistakes in the prescription of training and rehabilitation programs if they do not correct for this mismatch in metabolic load.

Initially, two different methods were chosen to quantify the MRT. In addition to the traditional linear approach (i.e., MRTLIN), Iannetta et al. (7) recently suggested a novel method that is based on the steady-state V˙O2 at 100 W (i.e., MRTSS). When strategy 1 (i.e., correcting for the MRT) was applied for PO selection in zone 1, we observed a high accuracy in estimating POtarget, irrespective of the MRT calculation method that was used (Fig. 2). In particular from a practical point of view, these results might question the added value of using the MRTss because its use requires the performance of an additional CWR bout that prolongs the total duration of the RI test with ~10 min. For this reason, it was decided to utilize MRTLIN and not MRTSS as the standard method for MRT calculation in the translation strategies presented in this study.

Heavy-intensity domain

Although strategy 1 enables an accurate PO selection in zone 1, the present study results made clear that this strategy is insufficient to work in zone 2. Without any correction, the differences in PO between RI and CWR exercise continued to raise with increasing intensities, reaching a discrepancy of 55 ± 12 W for the 30-W·min−1 protocol and 34 ± 17 W for the 15-W·min−1 protocol at the level of MLSS. It was expected that this difference would be smaller for the 15-W·min−1 protocol. Indeed, Iannetta et al. (13) already reported that the dissociation between the RI and the CWR V˙O2 response becomes smaller during protocols with a slower ramp because these protocols provide more time for the V˙O2sc to develop. Consequently, the use of a simple MRT correction in zone 2 did not suffice and was associated with an average estimation error of 17.1 ± 15.9 W (Fig. 3). The positive slope of the regression line in Figure 3A shows that the mean bias tends to rise with increasing intensities and, thus, supports the idea that strategy 1 is not accurate for PO selection within the heavy-intensity domain.

To eliminate the gap between the V˙O2/PO relationships at higher intensities, Iannetta et al. (13) proposed the use of an extremely slow ramp slope (i.e., 5 W·min−1). Although we agree with the authors that such RI protocol could resolve the discrepancy between RI and CWR exercise, we doubt the practical feasibility of this test, considering its very long duration (i.e., ~1 h in moderately trained individuals). Even more importantly, such an approach would jeopardize the very reasons to perform an RI test, being the validity and efficiency in assessing both submaximal and maximal fitness. It should also be acknowledged that the imposition of this 5 W·min−1 protocol somehow forces sports practitioners to abandon the protocols that they are working with and have expertise with. These protocols are usually population specific in terms of the chosen ramp slope to elicit exhaustion within 8 to 12 min. For these reasons, in the present study, we aimed to introduce a translation strategy that does not extend the RI test duration and works for protocols with different ramp slopes.

In contrast to strategy 1, we observed that strategies 2 and 3, which comprised an extra correction based on the individual loss of mechanical efficiency, resulted in a much better estimation of POtarget in zone 2 (Table 3). However, despite the fact that both approaches displayed an almost neglibible deviation in terms of PO (i.e., 2 to 4 W), statistical analysis revealed that only strategy 2 was effective to accurately predict POtarget in zone 2. When all three translation strategies were applied to the complete data set (i.e., combined data of zone 1 and 2), similar results were encountered.

Practical implications

We acknowledge that the use of translation strategies 2 and 3 is associated with some limitations. To implement these methods in practice, several parameters need to be derived from the RI test, including the V˙O2 at GET and RCP, the RI V˙O2/PO relationship, and the MRT. Furthermore, knowledge of s1 − CWR and/or s2 − CWR is mandatory. Although we believe that the rationale for developing strategy 3 makes sense, we advocate the use of strategy 2 as this approach was shown to be more accurate and requires to know only s2 − CWR. Still, it remains a problem that this single parameter needs be determined experimentally, and performing extra tests to quantify s2 − CWR would complicate the test protocol. The use of a population average, rather than an individualized slope, could be done to overcome this problem. In the current subject group, the mean s2 − CWR was 14.2 ± 2.4 mL·min−1·W−1. We acknowledge that this approach is less accurate because individual differences are not taken into account anymore. Moreover, it is difficult to say to what extent the average value of one subject group can be transferred to a broader population since V˙O2 kinetics are influenced by individual characteristics (12).

Additional analyses showed that the use of strategy 2 by means of an average s2 − CWR did not reveal a significant bias between POest and POtarget in the present subject group (P = 0.398). However, to investigate the applicability of the present translation strategy to other populations, we tested the model to a different subject group using existing data of our own laboratory. The subject group consisted of eight (post)menopausal women (54.1 ± 2.6 yr, 1.62 ± 0.6 m, 57.2 ± 8.5 kg) with a mean V˙O2peak of 36.5 ± 5 mL·min−1·kg−1. The subjects completed either a 10-W·min−1 or a 15-W·min−1 RI protocol together with several CWR tests at different intensities. A total of 25 data points in zone 1 and 2 were eligible to include in the Bland–Altman analysis. Figure 4 displays a mean bias of −0.9 ± 11.4 W (LoA: lower = −23.3 W, upper = 21.5 W) (P = 0.704), demonstrating that strategy 2 was able to accurately predict POest in a subject group of the opposite gender and with a different age and physical fitness level. Although these results are promising, we suggest that future studies examine the CWR V˙O2/PO relationship more thoroughly in a wide range of populations (e.g., elite athletes, sedentary people, clinical patients) and in a broader range of ramp slopes to obtain population-specific values of s2 − CWR and further validate the model.

Bland–Altman plot showing the agreement between POtarget and POest in a validation group of eight middle-age women. Translation strategy 2 was applied to a total of 25 data points in zone 1 (n = 9) and 2 (n = 16). The differences in PO (y-axis) are plotted against the average PO (x-axis), showing the mean bias (black line) and the 95% limits of agreement (dashed lines).

In this study, the CWR V˙O2/PO relationship was modeled using a double linear fitting. By doing so, it was assumed that the gain in V˙O2 in the heavy-intensity domain evolves in a linear fashion. Keir et al. (4) presumed that the V˙O2 gain would increase in a nonlinear way but did not have evidence to support this hypothesis. Additional analysis on the present data showed that, on an individual basis, the use of a linear regression analysis resulted in a better curve fitting compared with an exponential regression analysis (RMSE = 71 ± 44 mL·min−1 vs 85 ± 48 mL·min−1) (P = 0.036). Although this difference is rather small, we believe that it is appropriate to linearly model the V˙O2/PO relationship within the heavy-intensity domain.


Until now, defining a constant PO for heavy CWR exercise based on the RI V˙O2 response has been doubtful. This study is the first to provide a practical solution by introducing a new and comprehensive translation strategy that can be used for PO selection within both the moderate- and the heavy-intensity domain and for RI protocols with different ramp slopes. Taking into account the variation of estimation, strategy 2 offers an easy and practical method that can be used within a population of physically active young people. In this light, we have provided a supplemental spreadsheet that includes a simple calculator tool to perform this novel translation strategy (see spreadsheet, Supplemental Digital Content 1,

The authors thank the subjects for their commitment to the study. Funding for this project was received from the “Special Research Fund” of the Ghent University.

Results of the present study do not constitute endorsement by the American College of Sports Medicine and are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. No conflicts of interest, financial or otherwise, are declared by the authors.


1. Whipp BJ, Davis JA, Torres F, Wasserman K. A test to determine parameters of aerobic function during exercise. J Appl Physiol Respir Environ Exerc Physiol. 1981;50(1):217–21.
2. Davis JA, Whipp BJ, Lamarra N, Huntsman DJ, Frank MH, Wasserman K. Effect of ramp slope on determination of aerobic parameters from the ramp exercise test. Med Sci Sports Exerc. 1982;14(5):339–43.
3. Boone J, Bourgois J. The oxygen uptake response to incremental ramp exercise: methodological and physiological issues. Sports Med. 2012;42(6):511–26.
4. Keir DA, Paterson DH, Kowalchuk JM, Murias JM. Using ramp-incremental VO2 responses for constant-intensity exercise selection. Appl Physiol Nutr Metab. 2018;43(9):882–92.
5. Weatherwax M, Harris NK, Kilding AE, Dalleck LC. Incidence of VO2max responders to personalized versus standardized exercise prescription. Med Sci Sports Exerc. 2019;51(4):681–91.
6. Boone J, Koppo K, Bouckaert J. The VO2 response to submaximal ramp cycle exercise: influence of ramp slope and training status. Respir Physiol Neurobiol. 2008;161(3):291–7.
7. Iannetta D, Murias JM, Keir DA. A simple method to quantify the VO2 mean response time of ramp-incremental exercise. Med Sci Sports Exerc. 2019;51(5):1080–6.
8. Fontana FY, Keir DA, Bellotti C, De Roia GF, Murias JM, Pogliaghi S. Determination of respiratory point compensation in healthy adults: Can non-invasive near-infrared spectroscopy help? J Sci Med Sport. 2015;18(5):590–5.
9. Caen K, Vermeire K, Bourgois JG, Boone J. Exercise thresholds on trial: are they really equivalent? Med Sci Sports Exerc. 2018;50(6):1277–84.
10. Whipp BJ, Wasserman K. Oxygen uptake kinetics for various intensities of constant-load work. J Appl Physiol. 1972;33(3):351–6.
11. Jones AM, Grassi B, Christensen PM, Krustrup P, Bangsbo J, Poole DC. Slow component of VO2 kinetics: mechanistic bases and practical applications. Med Sci Sports Exerc. 2011;43(11):2046–62.
12. Jones AM, Poole DC. Oxygen uptake kinetics in sport, exercise and medicine. J Sports Sci Med. 2005;4(1):84.
13. Iannetta D, Azevedo RA, Keir DA, Murias JM. Establishing the VO2 versus constant-work-rate relationship from ramp-incremental exercise: simple strategies for an unsolved problem. J Appl Physiol (1985). 2019;127(6):1519–27.
14. Grassi B, Rossiter HB, Zoladz JA. Skeletal muscle fatigue and decreased efficiency: two sides of the same coin? Exerc Sport Sci Rev. 2015;43(2):75–83.
15. Borg GA. Psychophysical bases of perceived exertion. Med Sci Sports Exerc. 1982;14(5):377–81.
16. Iannetta D, Fontana FY, Maturana FM, et al. An equation to predict the maximal lactate steady state from ramp-incremental exercise test data in cycling. J Sci Med Sport. 2018;21(12):1274–80.
17. Heck H, Mader A, Hess G, Mücke S, Müller R, Hollmann W. Justification of the 4-mmol/L lactate threshold. Int J Sports Med. 1985;6(3):117–30.
18. Lamarra N, Whipp BJ, Ward SA, Wasserman K. Effect of interbreath fluctuations on characterizing exercise gas exchange kinetics. J Appl Physiol. 1987;62(5):2003–12.
19. Keir DA, Murias JM, Paterson DH, Kowalchuk JM. Breath-by-breath pulmonary O2 uptake kinetics: effect of data processing on confidence in estimating model parameters. Exp Physiol. 2014;99(11):1511–22.
20. Beaver WL, Wasserman K, Whipp BJ. A new method for detecting anaerobic threshold by gas exchange. J Appl Physiol. 1986;60:2020–7.
21. Binder RK, Wonisch M, Corra U, et al. Methodological approach to the first and second lactate threshold in incremental cardiopulmonary exercise testing. Eur J Cardiovasc Prev Rehabil. 2008;15(6):726–34.
22. Wasserman K, Mcilroy MB. Detecting the threshold of anaerobic metabolism in cardiac patients during exercise. Am J Cardiol. 1964;14:844–52.


Supplemental Digital Content

Copyright © 2020 by the American College of Sports Medicine