Comparison of V̇O2-Kinetic Parameters for the Management of Heart Failure

Objective: The aim of this study was to analyze whether V̇O2-kinetics during cardiopulmonary exercise testing (CPET) is a useful marker for the diagnosis of heart failure (HF) and to determine which V̇O2-kinetic parameter distinguishes healthy participants and patients with HF. Methods: A total of 526 healthy participants and 79 patients with HF between 20 and 90 years of age performed a CPET. The CPET was preceded by a 3-min low-intensity warm-up and followed by a 3-min recovery bout. V̇O2-kinetics was calculated from the rest to exercise transition of the warm-up bout (on-kinetics), from the exercise to recovery transition following ramp test termination (off-kinetics) and from the initial delay of V̇O2 during the warm-up to ramp test transition (ramp-kinetics). Results: V̇O2 off-kinetics showed the highest z-score differences between healthy participants and patients with HF. Furthermore, off-kinetics was strongly associated with V̇O2peak. In contrast, ramp-kinetics and on-kinetics showed only minimal z-score differences between healthy participants and patients with HF. The best on- and off-kinetic parameters significantly improved a model to predict the disease severity. However, there was no relevant additional value of V̇O2-kinetics when V̇O2peak was part of the model. Conclusion: V̇O2 off-kinetics appears to be superior for distinguishing patients with HF and healthy participants compared with V̇O2 on-kinetics and ramp-kinetics. If V̇O2peak cannot be determined, V̇O2 off-kinetics provides an acceptable substitute. However, the additional value beyond that of V̇O2peak cannot be provided by V̇O2-kinetics.


INTRODUCTION
The incidence and prevalence of heart failure (HF) are high and continue to increase in the developed world with aging of the population. Concomitant deaths and healthcare costs related to this syndrome are increasing (Virani et al., 2020). Accurate diagnostic and risk assessment methods for HF are essential to guide clinical decisions for therapeutic strategies with the ultimate goal of decreasing risk and improving health outcomes. Gas exchange variables obtained through cardiopulmonary exercise testing (CPET) are an established method for accurately stratifying risk in patients with HF; many CPET responses now have a substantial evidence base (Wagner et al., 2018). However, maximal CPET parameters can be difficult to interpret as they are highly dependent on subject effort (Wagner et al., 2020). As many patients with HF are not familiar with severe exercise intensities, pushing these patients to their physiological limit remains a challenge. Therefore, there have been many efforts to investigate submaximal markers such as oxygen uptake kinetics (VO 2 -kinetics) in patients with HF.
VO 2 -kinetics represent the rate at which generation of aerobic adenosine triphosphate (ATP) adjusts to changes in the exercise intensity (Poole and Jones, 2012). This parameter depends on the ability of the cardiovascular system to rapidly increase or decrease the oxygen supply to the working muscles (Kemps et al., 2009;Chatterjee et al., 2013) as well as on the ability of the muscles to rapidly utilize oxygen (Weiss et al., 2017). Therefore,VO 2kinetics can provide critical information regarding the regulating capacity of the cardiovascular system and the skeletal muscles to utilize oxygen (Chatterjee et al., 2013) and, thus, exercise intolerance and functional mobility (Sietsema et al., 1994;Pavia et al., 1999;Hummel et al., 2016).
Studies investigating whetherVO 2 -kinetics is a useful marker for risk stratification in HF have reported conflicting findings (de Groote et al., 1996;Meyer et al., 1998;Pavia et al., 1999;Nanas et al., 2001;Schalcher et al., 2003;Fortin et al., 2015;Hummel et al., 2016). While some have reported that the prognostic value ofVO 2 -kinetics is even superior toVO 2peak (Schalcher et al., 2003;Fortin et al., 2015), others have reported only moderate or minimal additional value beyondVO 2peak (de Groote et al., 1996;Pavia et al., 1999;Hummel et al., 2016). These conflicting results are likely caused by the fact that varying features ofVO 2 -kinetics were analyzed (i.e., on-kinetics or off-kinetics), and different calculation approaches were used. VO 2 -kinetics is traditionally measured by performing a constant load test. CPET using a ramp protocol is, however, the preferred method to perform an exercise test in the clinical setting (Ross et al., 2016). The current manuscript therefore focused only on the utility ofVO 2 -kinetics during a ramp protocol. The aims of the study were (1) to analyze whetherVO 2 -kinetics parameters obtained from a CPET can distinguish between healthy participants and cardiac patients with HF and between New York Heart Association (NYHA) functional classes; (2) to determine whichVO 2 -kinetic parameter and which calculation are most useful; and (3) whether the most promisingVO 2 on-anḋ VO 2 off-kinetic parameter can add additional value toVO 2peak .

Cohort and Recruitment
The COmPLETE-Study is a cross-sectional single-center study and consists of two parts, namely, COmPLETE-Health and COmPLETE-Heart. COmPLETE-Health included healthy men and women without any known exercise-limiting diseases between 20 and 90 years of age equally distributed across age decades and sex. COmPLETE-Heart included cardiac patients with stable HF with NYHA functional classes I-III, with symptoms and signs stable for at least 1 month. Diagnosis of HF was confirmed by clinical history, physical examination, assessment of natriuretic peptide (NT-proBNP), and echocardiographically documented structural heart disease or diastolic dysfunction according to the European Society of Cardiology guidelines (Ponikowski et al., 2016). Details on recruitment procedures and complete inclusion and exclusion criteria can be found in the study protocol (Wagner et al., 2019).

Setting
The study was carried out at the Department of Sport, Exercise, and Health at the University of Basel, Switzerland, and was funded by the Swiss National Science Foundation (grant no. 182815). The study complies with the Declaration of Helsinki and was approved by the Ethics Committee of Northwestern and Central Switzerland (EKNZ 2017-01451). Written informed consent was obtained from all participants before the start of the study.

Acquisition of Participant Characteristics
Resting systolic and diastolic blood pressures were measured with the participant in the supine position using a noninvasive vascular screening system (VaSera VS-1500N; Fukuda Denshi, Tokyo, Japan). Physicians assessed medical history and medications by the questionnaire onsite. Based on clinical data, structured questions, and self-reported exercise tolerance, each patient with HF was assigned to an NYHA functional class by a physician who was blinded to both CPET results and laboratory data. Blood samples were drawn via venipuncture by trained medical staff in fasting state (at least 3 h). Samples were immediately centrifuged, the plasma aliquots were frozen at a temperature of −80 • C, and all samples were analyzed together after completion of the study.

Cardiopulmonary Exercise Testing
An exercise test to maximal voluntary exertion using an electromagnetically braked cycle ergometer (Ergoselect 200; Ergoline, Bitz, Germany) was performed according to one of the following five ramp protocols: (i) a 3-min warm-up either unloaded, a load of 10 or 20 W for protocols 1-3, or a load of 50 W for protocols 4 and 5 followed by (ii) a ramp protocol with a linear workload increases of 7, 10, 15, 20, or 30 W/min for protocols 1-5, respectively, followed by (iii) a 3-min recovery phase at the same workload as the warm-up. The protocol was chosen to achieve a duration of approximately 10 min.
Gas exchange and ventilatory variables were analyzed breath-by-breath continuously using a computer-based system (MetaMax 3B; Cortex Biophysik GmbH, Leipzig, Germany). Each test was preceded by a resting period of 3 min to reach steady-state conditions. In the absence of clinical symptoms or electrocardiographic abnormalities, all tests were continued until maximal exertion. Before and during the test, the participants were verbally encouraged to reach maximal exhaustion. Before each test, the equipment was calibrated in standard fashion with reference gas and known volume.VO 2peak was defined as the highest 30 s average value during the CPET. The slope of ventilation vs. carbon dioxide consumption (VE/VCO 2 slope) was calculated from 1 min after beginning of the ramp test up to the respiratory compensation point. As recommended earlier (Gargiulo et al., 2014;Salvioni et al., 2020), we also calculateḋ VO 2peak andVE/VCO 2 slope expressed as percentage of the predicted values. For this purpose, we used our recently published data (Wagner et al., 2021), which are mainly based on the healthy cohort of the present study.
VO 2 -Kinetic Assessment Figure 1 displays the different methods used to determineVO 2kinetics. Initially,VO 2 was filtered by removing all outliers that differed more than three standard deviations from the local mean (moving average of six breaths). The filteredVO 2 values were then linearly interpolated to provide second-by-second values, as previously recommended (Benson et al., 2017).VO 2 on-kinetics was assessed from the rest to exercise transition of the 3-min constant load warm-up period. In accordance with earlier studies, we calculated the time constant ofVO 2 on-kinetics by two different approaches: (1) τVO 2 on-kinetics. A mono-exponential function was fit (see Appendix for the exact equation) from the beginning to the end of the warm-up period using non-linear leastsquares method regression analyses (Hummel et al., 2016) (see Eq. 1 in Appendix).
(2) τVO 2 on-kinetics byVO 2 -deficit. This was determined by the oxygen deficit and the steady-state increase ofVO 2 above the resting value (Sietsema et al., 1994;Schalcher et al., 2003) (see Eq. 2 in Appendix). VO 2 off-kinetics was assessed from the active recovery period that directly followed the incremental phase of the CPET. This was done using three different approaches: (1) τVO 2 off-kinetics. Determined by the time constant of a mono-exponential function that was fitted from the beginning to the end of the recovery period using nonlinear least-squares method regression analyses (de Groote et al., 1996;Pavia et al., 1999;Hummel et al., 2016) (see Eq. 3 in Appendix).
(2) Slope linearVO 2 off-kinetics. Determined by the slope of a linear function that was fitted into theVO 2 -time relationship of the first minute of recovery using linear least-squares method regression analyses (Nanas et al., 2001) (see Eq. 4 in Appendix).
(3) % relVO 2 reduction 60 s and 120 s post-test. Determined by the decrease inVO 2 from the end of the incremental phase up to the first (% relVO 2 reduction 60 s post-test) and second minute (% relVO 2 reduction 120 s post-test) expressed as percentages ofVO 2peak (Fortin et al., 2015).
Ramp test kinetics were assessed from the initial delay oḟ VO 2 at the beginning of the incremental exercise phase (mean response time; MRT), as previously described (Meyer et al., 1998;Niemeyer et al., 2020). For this purpose, the intersection between a horizontal line crossing theVO 2 of the final 30 s of the warmup phase (VO 2 warm-up) and a straight line, which was fitted into the linearVO 2 -work rate response of the incremental phase was calculated (see Eq. 5 in Appendix).

Statistical Analysis
We investigated potential differences inVO 2 -kinetics variables between healthy participants and patients with HF using linear regression models, which were adjusted for age and sex. In detail, residual diagnostics were used to see whether the model assumptions were satisfied, and some kinetic parameters were subsequently log-transformed.
To investigate the associations betweenVO 2 -kinetic parameters andVO 2peak , linear regression analyses witḣ VO 2peak as the dependent variable and age, sex, and the kinetic variables as independent variables were calculated. Separate models for each kinetic parameter were built. Therefore, we modeled age using restricted cubic splines (natural splines) with four knots included along with an interaction by sex to control for potential non-linear age progression (Harrell, 2015). For some models, the residuals exhibited heteroscedasticity, and we present robust p-values and confidence intervals for those models (HC3) (Long and Ervin, 2000).
Descriptive statistics were used to compare theVO 2 -kinetic variables between NYHA classes I, II, and III and the healthy participants. To achieve comparability, we first created a matched dataset where we matched two healthy participants to every patient with HF according to age and sex (2:1 matching). We used the R package "MatchIt" for these calculations (version 3.0.2) (Ho et al., 2011).
The age-and sex-specific quantile curves were calculated using healthy participants only and applying generalized additive models for location, scale, and shape (GAMLSS, R package version 5.1-6) (Stasinopoulos et al., 2017). The age trajectories were modeled using penalized B-splines (P-splines). We adopted the Bayesian information criterion to select the conditional distribution that offered the best compromise between model complexity and goodness-of-fit. The models were inspected using diagnostic residual plots such as worm plots (van Buuren and Fredriks, 2001) and Q-Q plots. The z-scores of the patients with HF were calculated based on the established reference curves using the healthy participants.
Proportional odds ordinal logistic regressions were used to analyze whether the kinetic parameters with the largest mean difference in the z-scores added additional predictive information for disease severity (NYHA class). We used the unitless adequacy index to quantify the predictive information contained iṅ VO 2peak , age, and sex compared to the full set of predictors including the kinetic parameters (Harrell, 2015). An adequacy index near one indicates thatVO 2peak , age, and sex contain nearly all predictive information already, and that the kinetic parameters add little predictive information. We used likelihood ratio tests to assess whether the kinetic parameters improved the model fit. R version 3.6.1 or later (R Foundation for Statistical Computing, Vienna, Austria) was used for all analyses, and p-values ≤ 0.05 were considered statistically significant. All tests were two sided. FIGURE 1 | Graphical illustration of all analyzed kinetic parameters. τVO 2 on-kinetics byVO 2 -deficit was calculated by dividing theVO 2 -deficit by the amplitude ( V O 2ss ) of theVO 2 response.VO 2 , oxygen uptake; τ, tau; MRT, mean response time; rel, relative.

Participant Characteristics
A total of 526 healthy participants and 79 patients with HF (NYHA functional classes I-III) were included in the study. All patients with HF were in stable condition; their etiologies were cardiomyopathy (n = 8), coronary artery disease (n = 60), pulmonary hypertension (n = 1), valvular regurgitation (n = 8), and valvular stenosis (n = 2). Thirty-five patients with HF had a preserved ejection fraction (≥50%), 15 patients with HF had midrange ejection fraction (40-49%), and 23 patients with HF had a reduced ejection fraction (<40%) while the data of six patients with HF were missing. Participant characteristics are presented in Table 1.

VO 2 -Kinetics in Health and Heart Failure
Group differences between healthy participants and patients with HF irrespective of their NYHA class are reported in Table 2. Six out of eightVO 2 -kinetic parameters showed evidence for a difference between the groups (p ≤ 0.007). The number of participants involved in the analysis of the respective kinetic parameter indicates the susceptibility to minor measurement difficulties during the CPET and the number of outliers due to the determination method which were excluded. Figure 2 presents violin plots of all analyzed kinetic parameters for NYHA class I, II, and III and the age-and sex-matched healthy reference group. In addition to the kinetic parameters, violin plots were presented for CPET markers known to have high predictive value (Wagner et al., 2018) includingVO 2peak , OUES, andVE/VCO 2 slope for comparison. Data are presented as mean ± standard deviation if not stated otherwise. BMI, body mass index; BP, blood pressure; HR, heart rate; P max , maximal power;VO 2max , maximal oxygen uptake; VE, volume of expiration; VCO 2 , carbon dioxide output; RER max , maximal respiratory exchange ratio; HR max , maximal heart rate; BL max , maximal blood lactate. * Two participants from the healthy cohort were matched to every patient with heart failure according to age and sex (2:1 matching).
Frontiers in Physiology | www.frontiersin.org FIGURE 2 | Comparison of VO 2 on-kinetic und VO 2 off-kinetic parameters between a healthy control group and patients with heart failure with NYHA functional classes I, II, and III.VO 2peak , peak oxygen uptake; OUES, oxygen uptake efficiency slope; VE, volume of expiration; VCO 2 , carbon dioxide output; τ, tau; MRT, mean response time; rel, relative.
The z-scores ( Table 2) show that τVO 2 on-kinetics was the bestVO 2 on-kinetic parameter to discriminate between healthy participants and patients with HF. Slope linearVO 2 off-kinetics (ml/min/s) and % relVO 2 reduction 60 s post-test performed best among theVO 2 off-kinetic parameters. These three parameters were therefore considered superior to the others, and further analyses were limited to these parameters.
Quantile curves for τVO 2 on-kinetics, slope linearVO 2 offkinetics (ml/min/s), and % relVO 2 reduction 60 s post-test are presented in Figure 3. The quantile curves based on the healthy participants tend toward pathological numbers with increasing age. For the parameter τVO 2 on-kinetics, 60% of the patients with HF were located above the 50th percentile. For the slope lineaṙ VO 2 off-kinetics (ml/min/s), 85% of the patients with HF were located below the 50th percentile, and for relVO 2 reduction 60 s post-test, 78% were located above the 50th percentile.

Association ofVO 2 -Kinetics andVO 2peak
There was strong evidence for associations betweenVO 2peak (ml/kg/min) and allVO 2 on-and off-kinetics parameter except for MRT ramp kinetics in which there was no evidence for such an association observed ( Table 2, last three columns). The direction of the association can be described as follows: the faster thė VO 2 -kinetic response (depending on the parameter, a positive or negative association) the higher theVO 2peak values are observed. By far, the largest adjusted R 2 was observed for the slope lineaṙ VO 2 off-kinetics (ml/min/s). Table 3 shows the results of several models for NYHA class prediction by kinetic parameters. All kinetic parameters improved the model when the base model contained sex and age. As indicated by the Chi 2 , slope linearVO 2 off-kinetics (ml/min/s) improved the model from the three kinetic parameters most but not to the extent thatVO 2peak did.

Additional Value of Kinetics to Predict Disease Severity
There was little evidence that any of the three kinetic parameters improved the models already containing sex, age, andVO 2peak . The adequacy index comparing the base models containing age, sex, andVO 2peak to a model additionally containing the kinetic parameters was between 0.98 and 0.99. This means that the base models without theVO 2 kinetic parameters contain nearly all the predictive information already.

DISCUSSION
To our knowledge, the current study is the first to provide detailedVO 2 kinetic results in large cohorts of healthy FIGURE 3 | Quantile curves for τ VO 2 on-kinetics, slope linear VO 2 off-kinetics (ml/min/s), and % rel VO 2 reduction 60 s post-test for males and females, separately. The quantile curves are based on the healthy participants only (light blue data points). Values of the patients with heart failure are presented in orange.VO 2 , oxygen uptake. participants and patients with HF. All previously suggested methods to calculateVO 2 on-and off-kinetics for risk stratification using a standard ramp protocol were analyzed and compared. Our results show that theVO 2 off-kinetics according to relVO 2 reduction 60 s post-test (%) or slope linearVO 2 offkinetics (ml/min/s) present an alternative to evaluate aerobic function and disease severity ifVO 2peak cannot be determined. Additional value beyond that ofVO 2peak for risk stratification of patients with HF was not provided byVO 2 on-or off-kinetics.

Differences Between Healthy Participants and Patients With Heart Failure
This study provides evidence thatVO 2 -kinetic parameters differ between healthy participants and a group of mild to moderate functionally impaired patients with HF for all kinetic calculation methods with the exception of τVO 2 on-kinetics byVO 2deficit and MRT of the ramp kinetics. The observed differences are in line with previous findings showing that patients with HF had significantly slowerVO 2 -kinetics compared to healthy volunteers (Sietsema et al., 1994;Meyer et al., 1998;Pavia et al., 1999;Hummel et al., 2016). The slowing ofVO 2 -kinetics in HF is closely related to impaired ventricular-pulmonary vascular function (Kemps et al., 2009;Chatterjee et al., 2013) and/or impaired peripheral oxygen utilization (Weiss et al., 2017). Mean differences in z-scores ( Table 2) clearly indicate thatVO 2 off-kinetics, irrespective of the calculation method, discriminate better between healthy participants and patients with HF compared toVO 2 on-kinetics.
These results are in line with previous research showing that off-kinetics can be determined with greater fidelity (Kemps et al., 2009) and higher reproducibility (Kemps et al., 2007) than on-kinetics in patients with HF. Further, irrespective of the methodological difficulties with on-kinetics, off-kinetics may discriminate patients with HF better from their healthy counterparts as has been observed in a previous study (Sietsema et al., 1994).
The comparison between the off-kinetics parameters (different calculation approaches) revealed a higher potential to distinguish healthy participants and patients with HF for % relVO 2 reduction 60 s post-test and slope linearVO 2 off-kinetics compared to % relV O 2 reduction 120 s post-test and τVO 2 off-kinetics. Interestingly, both superior off-kinetics parameters were determined from the first minute of the recovery period only, while the other parameters were calculated from the first 2 min (% relVO 2 reduction 120 s post-test) or the entire recovery duration (τVO 2 off-kinetics). This indicates that the very early phase of the off transition better distinguished between healthy participants and patients with HF.

Association WithVO 2peak
Strong significant associations betweenVO 2peak and off-kinetics were observed. Slope linearVO 2 off-kinetics explained 39% of the variation inVO 2peak among healthy participants and patients with HF. In contrast,VO 2 on-kinetics showed significant but only weak associations withVO 2peak ; the on-kinetics parameter τVO 2 (s) explained only 4% of the variation inVO 2peak . The stronger association of the off-kinetics compared to the on-kinetics can likely be explained by the methodological considerations of the on-kinetics described above.
A recent study showed that the level of exhaustion had no impact onVO 2 off-kinetics (Ichikawa et al., 2020). That the determination ofVO 2 -kinetics, unlikeVO 2peak , does not require the subject to perform the test to maximal voluntary exertion is a large advantage. Many patients lack the motivation to perform a maximal exercise test, are not familiarized with severe exercise, or may have a contraindication to maximal exertion (Green and Askew, 2018). In contrast, the successful determination oḟ VO 2peak requires either aVO 2 -plateau or a confirmation of a secondary exhaustion criteria (Wagner et al., 2020). Considering the large existing evidence base for the valuable informatioṅ VO 2 -kinetics provides coupled with the present results,VO 2 off-kinetics can be suggested as potential substitute forVO 2peak .

Predicting Disease Severity
The ability of a model to predict health status and disease severity of the patients with HF improved significantly when theVO 2 on-kinetic parameter (τVO 2 on-kinetics) and thė VO 2 off-kinetic parameter (% relVO 2 reduction 60 s post-test) were added. However, onlyVO 2 off-kinetics added substantial information to the model. Thus,VO 2 off-kinetics could be a tool to discriminate not only between healthy participants and those with mild functional impairment (NYHA class I) but also between NYHA classes as visualized by Figure 2. Our results are in line with previous studies showing the potential ofVO 2 -kinetics for risk stratification (Schalcher et al., 2003;Fortin et al., 2015) but are in contrast to others who did not demonstrate better predictive value by the addition ofVO 2 offkinetics (de Groote et al., 1996;Pavia et al., 1999;Hummel et al., 2016).
Since we could already show the association withVO 2peakconsidered the gold standard criteria for risk stratificationanother established parameter, NYHA functional class, was used to stratify risk in patients with HF. Based on the different underlining physiological aspects represented bẏ VO 2peak andVO 2 -kinetics (Sietsema et al., 1994;Chatterjee et al., 2013), some additional predictive value ofVO 2 -kinetics could be expected. However, our results showed minimal evidence of additional value ofVO 2 on-or off-kinetics. Two reasons likely explain these results: (i)VO 2peak is already a very strong risk predictor in patients with HF and the association ofVO 2peak and NYHA class is already known to be high and (ii)VO 2 -kinetics are likely to provide the same predictive information asVO 2peak , which is underscored by the association betweenVO 2 off-kinetics andVO 2peak in this study. Therefore, even though we observed thatVO 2 -kinetics has predictive value, it does not appear to have value beyonḋ VO 2peak .

Practical Applications
Our results indicate that the method of quantifyingVO 2 -kinetics is critical to its clinical application. They suggest that the determination ofVO 2 on-kinetics from rest to a light constant load phase is not optimal; rather, the results favor the analysis of off-kinetics when using a ramp protocol. The calculation of relVO 2 reduction 60 s post-test (%) or slope linearVO 2 off-kinetics (ml/min/s) is recommended to distinguish between healthy individuals and patients with HF. SinceVO 2 off-kinetics is not affected by the level of exhaustion (Ichikawa et al., 2020), these parameters might be used as a substitute forVO 2peak when maximal exhaustion is not reached or whenVO 2peak cannot be interpreted.
Using some basic spreadsheet calculation tools, the calculation of relVO 2 reduction 60 s post-test (%) and slope linearVO 2 offkinetics (ml/min/s) are quite simple (see Appendix). To facilitate the routine application ofVO 2 off-kinetics in the clinical setting, we recommend that the incorporation of these parameters in CPET application software.

LIMITATIONS
Our study has limitations. The study was cross-sectional, and therefore no hard endpoints such as mortality or hospitalization were available. Furthermore, our HF cohort is predominantly represented by male ischemic patients with mildly reduced left ventricular ejection fraction and comparatively preserved exercise capacity as suggested by their mean values ofVO 2 peak. This may not fully reflect the real-world HF population, which partly limits the transferability of our findings. To further improve the reliability and validity of theVO 2 on-and offkinetics determination, a warm-up and a cool-down phase of 5 min instead of 3 min could be applied.

CONCLUSION
Differences inVO 2 -kinetics between healthy participants and patients with HF are observed and are highly dependent on how they are calculated.VO 2 off-kinetics appears to be superior for distinguishing patients with HF and healthy participants compared withVO 2 on-kinetics and ramp-kinetics. IfVO 2peak cannot be determined,VO 2 off-kinetics provides an acceptable substitute. However, additional value beyond that ofVO 2peak cannot be provided byVO 2 -kinetics.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Ethics Committee of Northwestern and Central Switzerland (EKNZ 2017-01451). The patients/participants provided their written informed consent to participate in this study.  (1) τVO 2 on-kinetics.

AUTHOR CONTRIBUTIONS
whereVO 2rest was defined as the mean of the final 60 s of the resting period preceding the warm-up. V O 2SS was the steady-state increase ofVO 2 aboveVO 2rest and τ the time constant of the overall response. τ thereby includes the cardio-dynamic and the primary component ofVO 2 on-transient kinetics and is equivalent to that what previously has been also called mean response time (MRT) of square wave exercise.
where V O 2SS was previously calculated from the meanVO 2 of the final 30 s of the warm-up period. TheVO 2 -deficit was calculated from the difference between the consumedVO 2 and theVO 2 -demand, which was calculated by multiplying V O 2SS with the duration of the warm-up period (for further details see: Schalcher et al., 2003). VO 2 off-kinetics: (1) τVO 2 off-kinetics.V whereVO 2rec was defined as the asymptotic value of the recovery response. V O 2SS was the steady-state decrease ofVO 2 abovė VO 2rec and τ the time constant of the overall response.
(2) Slope linearVO 2 off-kinetics.V O 2 = a t + b where a represents the slope and b the intercept of the linearVO 2 -time relationship.

Mean Response Time Ramp Test
whereVO 2warm−up and P warm−up are defined as theVO 2 and the work rate of the warm-up phase preceding the incremental phase. a and b represent the slope and the intercept of theVO 2 -work rate relationship of the incremental phase. S is defined as the ramp slope. TheVO 2 work rate slope was previously calculated using linear least-squares method regression analyses. To avoid any effects of a non-linearVO 2 response due to the initial lag ofVO 2 or a potential plateau, the first minute and the last 2 min were excluded for the calculation.