Energetic responses of head-out water immersion at different temperatures during post-exercise recovery and its consequence on anaerobic mechanical power

While exercise recovery may be beneficial from a physiological point of view, it may be detrimental to subsequent anaerobic performance. To investigate the energetic responses of water immersion at different temperatures during post-exercise recovery and its consequences on subsequent anaerobic performance, a randomized and controlled crossover experimental design was performed with 21 trained cyclists. Participants were assigned to receive three passive recovery strategies during 10 min after a Wingate Anaerobic Test (WAnT): control (CON: non-immersed condition), cold water immersion (CWI: 20 ℃), and hot water immersion (HWI: 40 ℃). Blood lactate, cardiorespiratory, and mechanical outcomes were measured during the WAnT and its recovery. Time constant (τ), asymptotic value, and area under the curve (AUC) were quantified for each physiologic parameter during recovery. After that, a second WAnT test and 10-min recovery were realized in the same session. Regardless the water immersion temperature, water immersion increased τV˙O2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau \dot{{V}}}_{{\text{O}}_{2}}$$\end{document} (+ 18%), asymptote (V˙O2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\dot{{V}}}_{{\text{O}}_{2}}$$\end{document}+ 16%, V˙CO2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\dot{{V}}}_{{\text{CO}}_{2}}$$\end{document} + 13%, V˙E\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\dot{{V}}}_{\text{E}}$$\end{document} + 17%, HR + 16%) and AUC (V˙O2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\dot{{V}}}_{{\text{O}}_{2}}$$\end{document}+ 27%, V˙CO2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\dot{{V}}}_{{\text{CO}}_{2}}$$\end{document} + 18%, V˙E\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\dot{{V}}}_{\text{E}}$$\end{document} + 20%, HR + 25%), while decreased τHR\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau {\text{HR}}$$\end{document} (− 33%). There was no influence of water immersion on blood lactate parameters. HWI improved the mean power output during the second WAnT (2.2%), while the CWI decreased 2.4% (P < 0.01). Independent of temperature, water immersion enhanced aerobic energy recovery without modifying blood lactate recovery. However, subsequent anaerobic performance was increased only during HWI and decreased during CWI. Despite higher than in other studies, 20 °C effectively triggered physiological and performance responses. Water immersion-induced physiological changes did not predict subsequent anaerobic performance.


Introduction
The specific physical properties of water impose physiological alterations in the human organism during water immersion.Under thermoneutral conditions, the hydrostatic pressure gradient acts by redistributing a significant volume of peripheral venous blood to the thoracic region (Arborelius et al. 1972;Echt et al. 1974), triggering cardiovascular responses such as a reflex reduction in heart rate (HR) (Risch et al. 1978b), reduction in plasma renin and aldosterone, and an increase in diuresis (Srámek et al. 2000).Under non-thermoneutral conditions, immersion temperature acts through the coefficient of heat conductivity, which is higher in water compared to air (Parsons 2014), challenging tissue temperature maintenance (Cannon and Nedergaard 2011).Depending on the water temperature, different physiological responses can be observed due to heat conduction and the activation of central and peripheral thermoreceptors, such as the following: peripheral vasoconstriction or vasodilation, non-shivering and/or shivering thermogenesis, altered HR, and increased minute ventilation ( VE ) and oxygen consumption ( VO 2 ) (Cannon and Keatinge 1960; Mekjavic  and Bligh 1989; Bonde-Petersen et al. 1992; Srámek et al.  2000; Kingma et al. 2012).The magnitude and speed of these responses depend on the immersed body surface and depth of immersion (Risch et al. 1978a;Kruel et al. 2014), the thermal gradient and duration of immersion (Cannon and Keatinge 1960;Mekjavic et al. 1991;Srámek et al. 2000), and HR before the immersion (Kruel et al. 2014).
Several studies have evaluated the cardiorespiratory effects of head-out water immersion during post-exercise recovery (Connelly et al. 1990;Mekjavic et al. 1991;Al Haddad et al 2010;Bastos et al 2012).However, the effect of immersion and water temperature during postexercise recovery on off-transient pulmonary VO 2 kinet- ics and excess post-exercise oxygen consumption (EPOC) remains unclear.Nevertheless, previous decrease in tissue temperature and peripheral oxygenation caused by cold water immersion (CWI) delays on-transient pulmonary VO 2 kinetics and increases VO 2 deficit in subsequent intense exercise (Ferretti et al. 1995;Stanley et al. 2014).Consequently, it is likely that CWI delays the off-transient pulmonary VO 2 kinetics and increases EPOC, which is also influenced by thermoregulatory adjustments due to water immersion duration (Cannon and Keatinge 1960;Mekjavic and Bligh 1989;Srámek et al. 2000), but there is currently no experimental support for this assumption.The same can be sustained for hot water immersion (HWI), although no study was explored this question.Furthermore, water immersion increases the VO 2 as a result of hydrostatic pres- sure, independent of immersion temperature (Mekjavic and Bligh 1989).
The increase in VO 2 during CWI recovery might be an alternative resource (Gaesser and Brooks 1984;Cochrane 2004) to accelerate the oxidation of the blood lactate [La − ] produced during anaerobic exercise (Belcastro and Bonen 1975).Though questioned (Lucertini et al. 2017), CWI immediately following exercise intensifies the reduction in blood [La − ] (Nakamura et al. 1996;Bastos et al. 2012), especially at colder temperatures (Nakamura et al. 1996).There is, however, no consensus regarding the deleterious effect of blood [La − ] on performance (Tesch et al. 1978;Nielsen et al. 2001;Robergs et al. 2004;Brooks 2018); from a traditional physiological perspective, a faster reduction in blood [La − ] could accelerate the individual's physiological recovery and favor subsequent exercise performance.Although this assumption has not been directly proven, several studies have been conducted aiming at more effective methods of blood [La − ] recovery (Belcastro and Bonen 1975;Weltman et al. 1979;Freund and Zouloumian 1981;Dood et al. 1984;Chiappa et al. 2008).
Despite accelerating the recovery, improving wellbeing and reducing injury rates (Cochrane 2004;White and Wells 2013), the use of CWI to improve subsequent exercise performance is questionable (White and Wells 2013;Stephens et al. 2017).While CWI improves the performance of intermittent and endurance exercises performed in warm environments (Marino 2002;Wegmann et al. 2012;Ross et al. 2013;Tyler et al. 2013;Ihsan et al. 2016), its effectiveness prior to anaerobic exercise remains controversial (Tyler et al. 2013;Versey et al. 2013).Different studies have reported an increase (Marsh and Sleivert 1999), maintenance (Peiffer et al. 2010), or even a decrease (Bergh and Ekblom 1979;Sargeant 1987;Ferretti et al. 1992;Schniepp et al. 2002) of anaerobic mechanical power after CWI.Thus, while post-exercise CWI may be beneficial from a physiological point of view in terms of accelerating blood [La − ] recovery, it may be detrimental to subsequent anaerobic performance.Contrarily, HWI might be an alternative to improve anaerobic performance because HWI shifts to the right the force-velocity relationship (Binkhorst et al. 1977;De Ruiter and De Haan 2000) and increases mechanical power (Sargeant 1987) and mechanical efficiency (Ferguson et al. 2002) during exercises performed at high velocities of muscle contraction such as anaerobic performance tests.
Based on assumptions previously mentioned, the central hypothesis of this study is that head-out, short-term water immersion during recovery of an anaerobic performance protocol (i.e., two times Wingate test) would delay VO 2 recovery and increase EPOC and accelerates blood [La − ] recovery during both CWI and HWI.In addition, subsequent anaerobic performance would be improved after HWI but not after CWI.Thus, the objective of this study was to compare the effects of CWI and HWI on VO 2 and blood [La − ] recovery profile and its effects on subsequent anaerobic performance.

Participants
Twenty-one healthy and well-trained male cyclists (five triathletes) (age: 26.9 ± 5.8 years, body mass: 72.3 ± 7.9 kg, stature: 176 ± 8 cm), who regularly trained on average more than 250 km per week in the past four years and participated in national competitions for at least 2 years before the study (level 4 athletes;De Pauw et al. 2013), were recruited for this research.Participants completed a medical history questionnaire form, and volunteers were excluded if they indicated regular use of medication; history of cardiovascular, metabolic, or respiratory disease; or temperature intolerance.All participants received prior individual guidance about the objectives, procedures and risks of the study, and signed an informed consent term approved by our institutional ethics committee in accordance with the Declaration of Helsinki.

Design
This study used a randomized and controlled crossover experimental design.All selected individuals attended the laboratory for four sessions.In the first session, anthropometric data were collected (body mass and height), the depth of immersion in the tank was individualized, and the participants were familiarized with the experimental protocol composed of the physical test and recovery.In the following three sessions, athletes were submitted to three different methods of recovery.They performed an anaerobic power test (i.e., Wingate Anaerobic Tests in a cycloergometer, WAnT) and recoveried for 10 min in a sitting position non-immersed considered control condition (CON), CWI at 20 °C or HWI at 40 °C.The order of trials execution was counterbalanced and participants were randomly assigned to perform the following sequence: CON/CWI/HWI (n = 4), CON/ HWI/CWI (n = 3), CWI/CON/HWI (n = 3), CWI/HWI/ CON (n = 4), HWI/CON/CWI (n = 4), HWI/CWI/CON (n = 3).After recovery, participants performed a second WAnT and recovery period in the following sequence: WAnT 1 -Recovery 1 -WAnT 2 -Recovery 2 .
All participants performed the tests between 02:00 and 06:30 p.m.An interval of at least 48 h, and maximal of 7 days was adhered between each session.The participants were instructed verbally and in writing in the first session and 24 h before each experimental session not to eat a heavy meal or to consume stimulants for at least four hours and to fast for three hours prior to data collection.Moderate consumption of water was limited to 1 hour before testing.Participants were also instructed not to perform physical exercise during the 24 h prior to the test and to allow sufficient time for at least eight hours of peaceful sleep the night before the test.In the experimental sessions, the participants were asked if they complied with the instructions.

Experimental procedures
Initially, participants were positioned in dorsal decubitus at rest for 10 min.Then, they remained in the sitting position at rest in the cycloergometer for 5 min.After this period, participants performed the first anaerobic power test for 30 s (WAnT 1 ).After the test, participants rapidly left the cycloergometer, removed their shoes, sat on a bench, and rested for 10 min (post-WAnT 1 ) either: CON or in a tank with water.The transition between the end of WAnT 1 and beginning of recovery (when the individual sat on the bench) took between 14 and 20 s.Bench adjustments were made so that all individuals were immersed at the same depth relative to the anatomical point (xiphoid process) since it is related to the greatest reductions in HR at thermoneutral temperatures (Kruel et al. 2014).After the recovery period, participants stood up, wore their shoes and sat on the cycle ergometer as soon as possible, receiving help with drying themselves when necessary.The transition time between the end of recovery and the beginning of WAnT 2 was 20 to 30 s.After the end of WAnT 2 , participants were monitored during a new recovery period lasting 10 min (post-WAnT 2 ), under the same conditions as before.The same recovery modality was repeated twice in each experimental trial (Fig. 1).
Water immersion was performed in a 250-L tank, with water temperature maintained by an electronic thermostat (± 1 ℃) (model N320, Novus, Porto Alegre, Brazil).The thermostat was connected to an adapted air conditioner (model 14R23CR,Springer Admiral,Porto Alegre,Brazil), in which the evaporator was replaced by chillers to cool or heat the water.The water circulation was maintained by a low flow immersion pump positioned in a closed-circuit system composed of the tank and chillers.The system was set up to avoid agitation and stratification of water temperature within the tank.During the different collection days, the environmental temperature was maintained between 21 and 24 ℃, and relative humidity between 65 and 80% (Fig. 2).Through this system, it was possible to determine a specific immersion temperature and keep it constant throughout the experiment.We are not aware of any other experiments that have described something similar before.

Cardiorespiratory measurements
Upon arrival at the laboratory, participants were fitted with an HR monitor (Vantage NV, Polar Electro-Oy, Kempele, Finland).Following 10 min in the dorsal decubitus position, they sat on the cycloergometer and rested for five min equipped with masks and a pneumotachograph coupled to a metabolic cart, for taking gas exchanges at pre-exercise ( PRE ).Averages were taken from the final two min for VO 2 (mL⋅min −1 and mL⋅kg −1 ⋅min −1 ), carbon dioxide production ( VCO 2 ) (mL⋅min −1 ), VE (mL⋅min −1 ), HR (beats⋅min −1 ), and Respiratory Exchange Ratio (RER = VCO 2 ⋅ VO 2 −1 ).These variables were collected continuously throughout the experimental protocol (Fig. 1).Manual calibration of the respiratory assessment system was performed prior to testing (MGC-CPX/D System, Medical Graphics Corp, St. Paul, MN).For the spirometry system, a three-liter syringe was used to simulate different ventilatory flows.Two cylinders with gas mixtures of known content were used for the gas analysis system.

Blood lactate measurements
At the end of the pre-exercise period, a 25-μl blood sample was collected through a capillary from the left earlobe.Blood samples were further collected at 2, 4, 6, 8 and 10-min recovery intervals after performing each WAnT, totaling to 11 blood samples for each protocol.The blood samples were transferred to Eppendorf tubes containing 50 μL of 1% NaF and analyzed for blood [La − ] (mmol⋅L −1 ) using a lactimeter by electroenzymatic method (Yellow Springs 1500 Sport Lactate Analyser, Ohio, USA), previously calibrated using lactate standards.

Performance measurements
The WAnT was performed on a cycle ergometer (Cybex The Bike, Cybex, Ronkonkoma, NY) by pedaling at maximum intensity against a fixed resistance at the highest possible pedaling frequency for 30 s. Braking force (N) was determined as the product of body mass by 0.075 (7.5% of body weight) and both braking force and pedaling rate (rpm) were monitored electronically.A manual calibration with known load (2 kg) was performed according to manufacturer's specifications at the beginning of each session and automatic calibrations always preceded each test.During the WAnT, the individuals were verbally motivated to guarantee that everyone reached and maintained the highest possible frequency during the test.The individuals started pedaling at the highest possible frequency without resistance three seconds before the beginning of the test.There was no previous warm-up.Mechanical power output was electronically recorded every 5 seconds, with calculation of peak (fifth second value) and mean power output (W) and mechanical work (kJ) for entire test.Relative values were also calculated by dividing absolute values by body mass (W⋅kg −1 and kJ⋅kg −1 ).The pedaling rate and estimated distance traveled were also evaluated as total (along a 30-s test) and mean values (of six 5-second intervals).Distance traveled was evaluated as the product of pedal rotations during the test by distance covered for each rim rotation (6.11 m).Fatigue index (%) was calculed using the following formula: fatigue index = ((peak power − lowest power)/peak power) × 100.

Subjective recovery sensation
At the end of the experimental protocol, the athletes were asked in which passive recovery method they felt the best and the worst subjective sensation of recovery.The evaluators were also prepared to record possible cases of malaise, nausea, and physiological tremor during the sessions.

Data analysis
The gas exchanges data were continuously recorded breathby-breath, and HR was recorded every five seconds.The data during the 10-min recovery were first interpolated for 1 second, totaling 600 points.Then a moving average of five points was performed (SPSS version 20.0, Chicago, IL).Except for RER, a monoexponential decay curve was fitted using the following equation: where y(t) is the y value in a given time (t) , y 0 is the asymp- totic value of the curve, A 1 is the amplitude response (i.e., the difference between the highest value of the curve and its asymptote), " e"is the base of natural logarithms (2.72) and τ (tau) is the time constant that represents the decay for the variable to reach 63% of its asymptote (Rossiter et al. 1999).The values of τ and the asymptotic value of the curve (asymp) were derived by nonlinear regression using the method of least squares (Origin 6.0, OriginLab Corporation, Northampton, MA, USA).Interactions of the Levenberg-Marquardt algorithm were performed to minimize the sum of squares and to obtain an intraclass correlation coefficient (ICC) and standard error (SE), which were used to evaluate the fit of the monoexponential curve (Fig. 3).ICC values ranged between r 2 = 0.90 and r 2 = 0.99, suggesting good quality adjutments.The determination of the τ and asymp aimed to evaluate the kinetics of the fast and slow phases of the cardiorespiratory recovery curve, respectively.RER 10min represents the magnitude of RER at the end of the recovery period, calculated through the average of the last minute.The average of the two final minutes of pre-exercise (1) y(t) = y 0 + A 1 e −t∕ − 1 , cardiorespiratory period ( PRE ) served as baseline of asymp and RER 10min AUC was determined to assess the total magnitude of the response of cardiorespiratory variables for each recovery period.Previously interpolated and filtered data were subtracted from the mean PRE data.AUC was determined by calculating the integral by the trapezoidal method.AUC units were expressed in liters and beats.The AUC of VO 2 during recovery was denominated EPOC.RER AUC was calculated in terms of the ratio between VCO 2 AUC and EPOC.
The analysis of the blood [La − ] behavior was performed by comparing its values over time and curve adjustments by means of a quadratic function.The total magnitude lactate response was obtained by summing the values of blood [La − ] collected during recovery, subtracted by blood [La − ] PRE and expressed in mmol (blood [La − ] overall ).

Statistical analysis
Data are expressed as mean, standard deviation (SD), and standard error (SE).The variables collected were analyzed under three different factors as follows: recovery methods (CON, CWI, HWI), tests (WAnT 1 and WAnT 2 ), and time (5,10,15,20,25 and 30 s for performance outcomes during WAnT and 2, 4, 6, 8 and 10 min for blood [La − ] throughout the recovery period).To analyze the variables with a single measurement for each recovery method, we carried out the Analysis of Variance (ANOVA) (for example: pre-exercise evaluations).Repeated-measures, two-way ANOVAs were used to analyze the variables with two to four evaluations for each recovery method in a repeated measures structure Test: y 0 is the asymptotic value of the curve, A 1 is the amplitude of the response between the highest value of the curve and its asymptote, and τ (tau) is the time constant (TC) that represents the decay for the variable to reach 63% of its asymptote (pre-exercise, WAnT 1 and WAnT 2 ), such as  VO 2 ,  VCO 2 ,  VE , VO 2 asymp , VCO 2 asymp , VE asymp , RER 10min , HR asymp EPOC, VCO 2 AUC , VE AUC and blood [La − ] overall ).Models with three fixed factors (recovery method, tests and time) were performed in situations that presented five to six evaluations per session (for example: blood [La − ], pedal rotations, distance and mechanical power output over time).In all analyses, the Mixed Model methodology was adopted, considering the effect of athletes as random and the effect of recovery method, tests, and time as fixed factors.The level of significance was set at 0.05 for all analyses.To verify the possible influence of the covariates age, body mass, height, blood [La − ] PRE and transition time, we considered the analysis of covariance model, with the model composed of the fixed factors and the covariates only when they presented significance level (P < 0.05).The complements of significant effects for recovery method, tests, and time were performed using the Tukey test.The study of model adequacy was performed by analyzing residuals versus fitted values (homogeneity) and verifying normality using the Kolmogorov-Smirnov, Anderson-Darling, and Cramer-von Mises tests.Pearson's partial or residual correlations were performed after eliminating the effect of factors to verify the association between variables.Statistical analysis was performed using the SAS statistical package (version 9.4; SAS Institute Inc, Cary, NC).

Results
Descriptive results of the physiological and performance variables are shown in Supplemental Table S1.The data obtained between the three recovery methods (CON, CWI, and HWI) in their respective tests (WAnT 1 and WAnT 2 ) or time factors were grouped to include all the different observations for each variable, regardless of the factor.In the test of normality of residuals, the evidence proved adherence to normality.Heterogeneous variance and discrepant data in the analysis of the graphics of the residuals versus adjusted values in the boxplot of the residuals confirmed the adequacy of the statistical model of Mixed Models.In the variables where interactions were not observed, comparisons were made to present only the effect of the recovery method (Table 1).These data were analyzed by considering only the test factors, regardless of the recovery method and time factor (Supplemental Table S2).In the variables where interactions were found, the effects of the recovery method and test were presented in Table 2.The analysis of covariance verified the influence of age for  VO 2 (P = 0.04) and of transition time for EPOC (P = 0.006), VCO 2 AUC (P = 0.03) and VE AUC (P < 0.001) and were added to the statistical model only for these variables.Blood [La − ] analyses were adjusted to their baseline levels (blood [La − ] PRE ), even though no

Cardiorespiratory measurements
The cardiorespiratory variables collected at pre WAnT 1 did not show significant differences between the recovery methods (P > 0.05).The averages found were as follows: HR supine (58.0 ± 8.4 beats⋅min −1 ), VO 2 PRE (336.4 ± 45.2 mL⋅min −1 and 4.7 ± 0.7 mL⋅kg −1 ⋅min −1 ), VCO 2 PRE (294.4 ± 50.2 mL⋅min −1 ), RER PRE (0.9 ± 0.08), VE PRE (9.5 ± 1.5 L⋅min −1 ), and HR PRE (69.4 ± 7.5 beats⋅min −1 ).The  VO 2 was higher in the CWI than in the CON (P = 0.02), and  VCO 2 and  VE did not show differences between the recovery methods (Table 1).HR was lower during water immersion recovery compared to CON (P < 0.001), with test effects (interactions) observed only for CON and HWI (Table 2).During the slow phase, all variables showed values asymp and 10 min higher than the pre-exercise situation, regardless of the recovery method (P < 0.0001).VCO 2 asymp and VE asymp were higher during water immersion recovery compared with CON (P = 0.02; P = 0.01, respectively).RER 10min was not different between the two recovery methods (Table 1).Interactions between recovery method and test were found only for VO 2 asymp and HR asymp .VO 2 asymp was greater during water immersion recov- ery compared with CON under test dependence (P < 0.01), and HR asymp was greater during HWI compared to CWI and CON (P < 0.001) (Table 2).HR was the only variable that showed all three parameters ( HR , HR asymp and HR AUC ) under the influence of the test factor, with a cumulative effect only for HR and HR AUC .
The EPOC and VCO 2 AUC were higher during water immer- sion recovery when compared with CON; however, significant differences were verified only between CWI and CON (P < 0.001).RER AUC , did not show differences between the recovery methods (Table 1).VE AUC and HR AUC showed higher values during HWI compared with CWI and CON (P < 0.001) (Tables 1 and 2).

Physical performance variables
The mechanical power values showed a reduction along the 5-second intervals of WAnT, regardless of the recovery method (F = 121.0;P < 0.0001).Disregarding the individual values of the recovery methods, the mean value obtained at the 5-s interval was lower during WAnT 2 compared with WAnT 1 , and the values at the 20, 25, and 30 s were higher (test vs time interaction P < 0.001).No threeway interaction was observed between recovery method vs tests vs time (Fig. 5).Absolute and relative mean power output during WAnT 2 were influenced by the recovery method and were higher after HWI compared to CON (P < 0.05) and CWI (P < 0.0001).The values produced after CWI were lower compared to CON (P < 0.05).The values produced during WAnT 2 were 3.2% higher compared to WAnT 1 after HWI (P < 0.0001) and showed a recovery method vs test interaction (Fig. 6 and Table 2).Absolute and relative peak power output and mechanical work were greater during testing after HWI when compared with CWI (P = 0.03) and showed no differences with CON.Although CWI had the lowest values, it showed no difference with CON (Table 1).Total and mean pedaling rate and distance traveled presented results similar to mean power output (Fig. 7 and Table 2).Fatigue index did not present differences between the recovery methods, nor between the tests.

Subjective sensation recovery and transition time
Among athletes, 57% answered that they felt better during CWI, 33.5% during CON and 9.5% during HWI.Regarding the worst feeling, 62% considered HWI to be the worse, 24% responded that CWI was the worse, and 14% found CON to be the worse.Thus, CWI was considered as the recovery method with best subjective sensation during recovery, while HWI induced the worst sensation.HWI recovery also presented results indicating the lowest preference.Three cases of nausea and malaise were observed after completion of the collections with HWI.In four individuals, physiological tremor was observed during CWI.

Results of correlations
All correlations were performed to eliminate the effect of the test and recovery method factors.Initially, the analyses were performed to verify the relationship between the fast and slow response of cardiorespiratory outcomes during exercise recovery and their respective AUC.Regarding fast response, the highest values found were between  VO 2 and EPOC (r = 0.425, P < 0.001), and between  VCO 2 and VCO 2 AUC (r = 0.357, P = 0.004).Regarding slow response, the highest correlations were between VO 2 asymp relative and EPOC (r = 0.629, P < 0.001), VO 2 asymp absolute and EPOC (r = 0.618, P < 0.001), VCO 2 asymp and VCO 2 AUC (r = 0.587, P < 0.001), VE asymp and VE AUC (r = 0.478, P < 0.001), and HR asymp and HR AUC (r = 0.624, P < 0.001).Thus, slow response was the parameter that correlated most with AUC in all cardiorespiratory variables.Weak and nonsignificant correlations were found between fast and slow response when analyzing the same cardiorespiratory outcome.Among different cardiorespiratory endpoints, the highest results were between EPOC and VE AUC (r = 0.625, P < 0.001).
The main correlation of blood [La − ] overall with cardiorespiratory variables was with VE AUC (r = 0.443, P < 0.001).Between cardiorespiratory and performance variables, the highest values found were between HR and absolute (r = 0.331, P = 0.009) and relative mean power output (r = 0.340, P = 0.007), between HR asymp and absolute (r = 0.482, P = 0.0001) and relative mean power output (r = 0.501, P = 0.0001), between HR AUC and absolute (r = 0.413, P < 0.001) and relative mean power output (r = 0.439, P < 0.001) and between VE AUC and fatigue index (r = 0.349, P = 0.005).Blood [La − ] overall values showed no significant correlations with performance variables.In general, the VE AUC was the variable that best correlated with physiological variables (EPOC and blood [La − ]) and with the fatigue index, while HR ( HR, HR asymp and HR AUC ), although low-mod- erate, was the one that best correlated with performance.Transition time showed weak and non-significant correlations with physiological and performance variables.

Discussion
The responses to post-exercise water immersion and its use as a recovery tool following physical exercises have received special attention in the past few years.Based on previous studies, we hypothesized that post-exercise shortterm water immersion following anaerobic exercise would result in increased aerobic energy expenditure, delaying the VO 2 kinetics and increasing EPOC especially during CWI.Furthermore, we expected that the other respiratory parameters would increase during immersion and that HR recovery would be accelerated during CWI and delayed during HWI.It was further anticipated that the recovery of blood [La − ] would be accelerated during CWI and that the anaerobic mechanical power output would be decreased after CWI and increased after HWI.This randomized and controlled crossover experimental design demonstrated that recovery from anaerobic power exercise using water immersion at different temperatures compared to CON as follows: (1) it provided increased aerobic energy expenditure and other respiratory parameters (including VO 2 kinet- ics and EPOC), a higher VE response in HWI compared to CWI; (2) higher HR values during HWI and a similarity between CWI and CON; (3) the slow recovery phase influenced the overall cardiorespiratory responses (AUC) more than the fast phase; (4) a similarity in the behavior and magnitude of blood [La − ] between the three recovery methods; (5) increased performance during WAnT 2 carried out after HWI and decreased performance after CWI; (6) while the temperature in the CWI protocol in the present study falls outside of the range, this may still constitute CWI, as evidenced by the divergent power output parameters between CON and HWI conditions, and (7) performance results which showed weak correlations with the blood [La − ] and with other cardiorespiratory variable.

Kinetics of respiratory variables
From a respiratory standpoint, the hypothesis of this study was confirmed through the observed increase in EPOC, VCO 2 AUC and VE AUC .The effect of immersion temperature was significantly higher only in the VE AUC during HWI.In general, the variables with higher area presented higher values of τ and asymp .A higher τ means a delay in the fast recovery phase, while a higher asymp means a delay in the slow recovery phase.Both conditions can impact the AUC, as it represents an integrative response between the behavior and magnitude of the respiratory variable and an oxidative metabolism more intense during recovery.The asymp was the parameter that exerted the greatest influence on the AUC in all cardiorespiratory variables.In relation to VO 2 the CWI produced the greatest distur- bance in EPOC accompanied by the highest values of  VO 2 and VO 2 asymp (absolute and relative).These differences were significant only in relation to CON.The EPOC presented a weak correlation coefficient with  VO 2 and a good coef- ficient with VO 2 asymp .This means that the slow phase of the recovery curve of VO 2 was the parameter that most influ- enced the magnitude of EPOC.A low correlation was found between and τ and asymp .
The delay observed in the recovery of VO 2 during CWI and HWI might be linked with a possible O 2 delivery limita- tion or an inadequate O 2 utilization.From the point of view of a O 2 delivery limitation, the blood redistribution triggered by immersion would have delayed the fast recovery phase of the VO 2 recovery phase by slowing down the restoration of muscle reserves of O 2 , adenosine triphosphate (ATP) and creatine phosphate (CP) (Linnarsson 1974;Hughson 1990).The greatest  VO 2 during CWI would have been produced as a consequence of reduced peripheral blood flow resulting from convergent redistribution triggered by the hydrostatic pressure gradient and peripheral vasoconstriction (Arborelius et al. 1972;Echt et al. 1974;Risch et al. 1978b;Srámek et al. 2000).A reduction in HR could also trigger a reduc- tion in  VO 2 , since an increase in HR is associated with an increase in  VO 2 (Hughson and Morrisey 1983); however, this was not confirmed (Table 2).During HWI, the  VO 2 presented a delay of smaller magnitude with no differences from CON.In this case, possible blood redistribution in divergent directions triggered by a hydrostatic pressure gradient and mainly peripheral vasodilation probably explain its lower values compared to CWI.The lower HR also did not influence the  VO 2 during HWI.A VE did not influence the  VO 2 in both situations, because its  VE did not show dif- ferences between the retrieval methods.These results seem to confirm the data of Srámek et al. (2000) that attribute the early responses of the VO 2 during immersion in a liquid medium to vascular adjustments and the late responses to other thermoregulatory adjustments such as cardiorespiratory changes and increase in cellular metabolism.
In this sense, different from the fast recovery phase, the greater VO 2 asymp would result from the increase triggered by the increase in cellular metabolism especially during CWI and cardiorespiratory changes during HWI.Considering that a greater heat gradient was produced during CWI recovery after intense exercise (Cannon and Keatinge 1960), it is possible to attribute the increased response of the VO 2 asymp during CWI to the need for maintenance of cell temperature (Mekjavic and Bligh 1989), since, with the exception of VE AUC , the cardiorespiratory responses of CWI were not different from those of the CON.Conversely, greater cardiorespiratory work was observed during HWI.Both the VE asymp and VE AUC , as well as the HR asymp andHR AUC , were higher during HWI in comparison to the other conditions.The incredible ventilatory increase observed during HWI would result from the need to provide heat energy loss, since the energy absorbed during warm immersion would hinder the loss of cellular heat energy produced during physical exertion (Gaudio and Abramson 1968).Despite the minor importance of heat exchange through respiration in thermoregulation in humans (Tsuji et al. 2012(Tsuji et al. , 2016)), it is likely that the elevation in body temperature contributed to the increase in ventilatory response.In addition, the increase seen in HR could result from a need to counterbalance the decrease in systolic volume caused by the vasodilation triggered by HWI (Keatinge and Evans 1961).In any case, these adjustments had repercussions in the increase observed in HR AUC and VE AUC and, consequently, in EPOC in relation to CON.
In contrast, other studies have downplayed the importance of O 2 delivery limitation in determining the VO 2 kinetics responses and highlighting the influence of inadequate O 2 utilization at a peripheral level.This means that the increase or decrease in O 2 uptake, changes in blood flow or cardiorespiratory adjustments would not necessarily imply changes in the VO 2 kinetics (Grassi et al. 1998;  Grassi 2000).In this sense, the delay of  VO 2 during immersion recovery could result from changes in tissue temperature on the enzymatic velocity and cellular metabolism, since similar responses were found in the on-transient pulmonary VO 2 kinetics after cooling during immer- sion (Ferretti et al. 1995;Stanley et al. 2014).In line with this reasoning, during CWI the decrease in tissue temperature could reduce hydrolysis (Ferretti et al. 1992) and ATP resynthesis (Blomstrand et al. 1984;Binzoni et al. 1990;Ishii et al. 1992) and the need to increase heat production suggests a decrease in cellular energy efficiency (Nedergaard et al. 2001;Cannon and Nedergaard 2011), producing a delay in the recovery curve of the VO 2 and an increase in EPOC.However, due to the short period of time, it is possible that the  VO 2 was not influenced by this mecha- nism, and it is more likely that only the slow phase of the recovery curve of VO 2 may be attributed (Mekjavic and  Bligh 1989).As such, Choukroun and Varene (1990) verified a higher arteriovenous difference during immersion at rest at 25 ℃, in comparison to 34 ℃, without difference in cardiac output, reinforcing the effect of temperature on the extraction of O 2 by the tissue.It is also possible that the physiological shivering triggered by thermoregulation increases the cellular energy requirement (Golozoubova et al. 2001) contributing to the increase in VO 2 asymp and EPOC; however, this phenomenon was observed in low intensity in only four athletes.In relation to HWI, it is likely that heat energy directly increased tissue temperature, elevated cellular enzymatic activity, and decreased cellular energy efficiency, because the VO 2 increased with the individual remaining at rest (Cannon and Keatinge 1960;Kingma et al. 2012).However, similarly to CWI, the  VO 2 was not influenced by this mechanism due to the short time frame (Mekjavic and Bligh 1989).
In this sense, although it was not the objective of the research to determine which parameters influenced the VO 2 kinetics, the factors related to the O 2 delivery limita- tion seem to explain the behavior of VO 2 at the beginning of immersion recovery, and the combination of both ( O 2 deliv- ery and O 2 utilization) seem to explain the late responses.Although the highest values of VO 2 were found during CWI, these responses were not significantly different from HWI.This result makes the issue complex, since the immersion temperature triggered different physiological responses, especially in relation to HR and VE .Research models associ- ating these outcomes with different immersion temperatures and duration deserve to be developed.
Similar to EPOC, the response of VCO 2 AUC revealed higher values during immersion, with no difference between temperatures.Therefore, immersion did not cause important changes in RER, keeping values higher than 1.0 throughout the period for the three recovery methods.This result can be attributed to the high blood [La − ] oxidation that exhibiteed similar behavior and magnitude during the recovery period for the three experimental conditions (Scott 1998(Scott , 2000)).As the EPOC and a VCO 2 AUC were higher during immersion, and blood [La − ] levels were similar, it is possible to speculate that a longer recovery period might reveal differences in blood [La − ] levels.A longer recovery time might also reveal differences in aerobic energy recovery highlighting the effect of non-shivering and/or shivering thermogenesis, reduced energy efficiency, or possible modifications in RER between the methods (Srámek et al. 2000).

Heart rate kinetics
From the cardiac point of view, the hypothesis of this study was that the recovery temperature would trigger antagonistic responses in HR.HR recovery during CWI would be accelerated, presenting reduced levels during recovery, while during HWI, there would be a delay and elevation in its values.CWI accelerated and maintained HR recovery only after WAnT 2 , while its values of HR asymp and HR AUC values did not present the expected decrease, being little different from that observed during CON.The lower values of HR during CWI may be attributed to a faster withdrawal of sympathetic tone and a faster reactivation of parasympathetic tone (Al Haddad et al. 2010) as an effect of immediate immersion.We know that the baroreflex stimulation generated by immersion triggers a reduction in HR and a decrease in muscle sympathetic nerve activity (MSNA) (Mano et al. 1991).These effects are potentiated in immersions below the thermoneutral temperature, being interrupted at very low immersion temperatures (Srámek et al. 2000), and proportional to the HR values measured before the individual entered the water (Kruel et al. 2014).However, this mechanism does not seem to have influenced the slow recovery phase of HR.An HR asymp was similar between CWI and CON, which may be attributed to late thermoregulatory responses due to the need to maintain body temperature observed during prolonged immersions at or below 20 ℃ related to increased sympathetic activity (Srámek et al. 2000).Unlike other studies that verified such responses at immersion temperatures below 20 ℃ (Srámek et al. 2000), the occurrence in this study could be justified because the design consisted of recovery from a high-intensity physical test.Thus, we suggest that thermoregulatory responses could occur at immersion temperatures above 20 ℃ due to physiological stress prior to immersion.
Conversely, the lower values of HR during HWI do not seem to represent a more accelerated recovery, but a mathematical consequence of the rise of its asymptote (HR asymp ) and the reduction in the amplitude of the variable recovery response ( A 1 ).A HR asymp during HWI was between 10 to 17 bpm higher than the other recovery methods and this contributed to the elevation of its values of HR AUC .These results seem to be related to hemodynamic responses resulting from vasodilation triggered by thermoregulation to heat and stimulation of cardioaccelerator centers due to likely elevation of core body temperature (Becker et al. 2009).In this sense, it is likely that prior exercise intensity, type of environment, temperature, and immersion time may modify HR behavior during recovery.In this study, the environment and the effect of immersion temperature were tested.These factors may influence the monitoring of the recovery period if HR is used as an evaluation parameter.

Lactate kinetics
Contrary to the initial hypothesis, the increased aerobic energy expenditure following CWI had no influence on blood [La − ] responses.We could verify that the sum of all blood [La − ] values throughout the recovery was not different between various recovery methods.Individual blood [La − ] values showed significant differences between time measurements without dependence on the recovery method.There was a cumulative effect of blood [La − ] between WAnT 1 and WAnT 2 so that the recovery values of WAnT 2 were higher than those of WAnT 1 .With the exception of VE AUC , weak correlations were found between blood [La − ] overall values and the other physiological and performance variables.
Contrary to our results, Connelly et al. (1990) found that both blood [La − ] and blood catecholamine levels returned more rapidly to pre-exercise values after recovery in water.This effect was attributed to a reduction in sympathetic activity triggered by baroreflex stimulation during immersion.However, lower levels of these two variables had already been observed in the last two stages of the physical test, since the test was performed in the liquid medium and this would have influenced its recovery.Nakamura et al. (1996) also verified lower levels of blood [La − ] during recovery in immersion, and the lowest values were found at lower immersion temperatures (30 ℃ vs 38 ℃ vs CON).The authors attribute these results to redistribution of blood flow as a response to thermoregulation.This phenomenon would cause the diversion of blood located in the skeletal muscles to the splanchnic region, providing increased blood supply to organs that perform blood [La − ] oxidation, such as the liver, kidneys and heart (Rowell et al. 1966;Belcastro and Bonen 1975).This hypothesis, together with the expectation that immersion would provide a significant decrease in MSNA (Mano et al. 1991), decreasing noradrenaline levels (Grossman et al. 1992), and accelerating post-exercise parasympathetic reactivation (Al Haddad et al. 2010), would justify the use of immersion for blood [La − ] removal after exercise.Furthermore, the elevation of EPOC and VE during HWI and CWI would strengthen this justification.Increased activity of respiratory muscles, especially the diaphragm (Cooper et al. 1976), is related to higher blood [La − ] oxidation (Romer et al. 2002).
However, the results did not confirm these expectations.The increase in O 2 delivery , VO 2 , VE and HR reduction during CWI or HR elevation during HWI did not modify the removal of blood [La − ] during the evaluated period.The alteration of these factors did not accelerate blood [La − ] removal under the conditions of this study or they are not as important as the consumption of limbs skeletal muscles (Gaesser and Brooks 1984;Scott 1998;Thomas et al. 2005).The immersion temperature used might not have been low enough to provide changes.However, very low temperatures may hinder blood [La − ] recovery as both blood catecholamines and HR increase during immersions below 20 °C (Srámek et al. 2000).Another possibility is that the athletes performed two supramaximal tests with an insufficient recovery period to generate changes in the blood [La − ].However, favorable results were found during the same recovery period: Bastos and collaborators (2012) verified a reduction of blood [La − ] from the 4th to the 8th minute in supine positions out of water, after 6 min of immersion at 11 ℃, after treadmill running until exhaustion in maximal aerobic speed.Nakamura and collaborators (1996) verified a decrease of blood [La − ] out of water after 10 min in immersion at 30 ℃ after pedaling at 80% of VO 2 max .A common aspect in these studies is that the reduction of blood [La − ] was verified only out of water, after the immersion period.In any case, if the recovery conditions were insufficient to generate changes in the blood [La − ], they were sufficient to modify subsequent physical performance.Although blood [La − ] is still used today as a parameter in determining physiological exercise intensity (Lundby and Robach 2015), metabolic pathway, (Gladden 2001) and recovery from exercise (Thomas et al. 2005), its importance as a triggering mechanism of fatigue has been widely questioned (Gladden 2004;Robergs et al. 2004), and the results of this study contribute to this discussion.

Anaerobic mechanical performance after cold and hot water imersion recovery
Confirming the hypothesis of this study, the anaerobic mechanical power output decreased after CWI and increased after HWI.With the exception of the fatigue index, HWI increased all the variables of physical performance, being higher than those obtained after CWI.Absolute and relative mean power output were higher after HWI compared to CWI and CON, and lower after CWI compared to CON.
The deterioration of physical performance due to adverse environmental conditions, such as when the athlete feels cold or wet (pre-immersion), has been observed in other studies, especially during exercises executed at high intensity (Bergh and Ekblom 1979;Howard et al. 1994;Schniepp et al. 2002).Contrary to these results, Marsh and Sleivert (1999) found an increase in power after performing pre-immersion for 30 min at a temperature between 12 and 18 ℃.However, this study performed pre-immersion only of the trunk, with the legs out of the water, and the individuals performed a 10-min warm-up before starting the 70-s anaerobic power test in a cycloergometer.
The main factor explaining the decrease in power after pre-immersion in CWI has been the decrease in muscle temperature that may be triggered by the conduction of the heat and consequent decrease in muscle blood flow (Suzuki et al. 1980;Blomstrand and Essen-Gustavsson 1987;Bonde-Petersen et al. 1992).For these reasons, changes in energy metabolism may occur (Blomstrand et al. 1984;Ferretti et al. 1992;Perini et al. 1998) or in the balance of agonist-antagonist electromyographic activity of the muscle (Oksa et al. 1995).The reduction of blood flow added to the O 2 demand during exercise may intensify the dependence on anaerobic metabolism for ATP resynthesis and increase acidosis (Schniepp et al. 2002).For example, Beelen and Sargeant (1991) found higher peak blood [La − ] when muscles exercised at 70% of VO 2 max had been previously submitted for water cooling at 12 ℃ for 45 min.However, in the present study, even though peripheral vasoconstriction occurred, it seems that it was not enough to modify lactic metabolism, and physical performance was independent of blood [La − ] response in the three recovery methods.
The reduction in muscle temperature could also be related to the decrease in ATP resynthesis due to the decrease in enzymatic activity in glycolytic and oxidative reactions (Blomstrand et al. 1984;Binzoni et al. 1990;Ishii et al. 1992) and the decrease in ATP hydrolysis rate (Ferretti et al. 1992).These factors, besides reducing the energy supply to the muscle, decrease the rate of cross-bridge disconnection (Ferretti et al. 1992), leading to greater muscle stiffness and, consequently, greater mechanical resistance in cooled muscles (Faulkner et al. 1990).The reduction in muscle temperature may also generate deterioration in the activation of motor units caused by the decrease in nerve impulse frequency (Vangaard 1975).Oksa et al. (1995) verified that, under conditions of muscle cooling, there was a decrease in the electromyographic signal integral (iEMG) of the agonist muscle and an increase in the iEMG of the antagonist muscle.The authors highlighted that in high-speed situations, a slight alteration in the agonist/antagonist balance can impair the coordination of movements and hinder an adequate production of force.Bergh and Ekblom (1979), on the other hand, did not find alterations in the EMG signal during exercise in cycloergometer after muscular cooling from 38 to 30 ℃; however, they verified a decrease of 4-6% in the power produced for each 1 ℃ of reduction in the muscular temperature.In this sense, muscle cooling seems to impair performance, especially at high velocities of muscle contraction (Ferretti et al. 1992) by reducing pedaling frequency.
In contrast, the performance improvement seen after HWI could be attributed to the increase in muscle temperature through heat conduction and by peripheral vasodilation, contributing to faster restoration of energy stores and acceleration of the rate of cross-bridge disconnection (Stienen et al. 1996;He et al. 2000).Although confirmed only when associated with water movement at thermoneutral temperatures (Sato et al. 2014(Sato et al. , 2015)), stimulation of peripheral thermoreceptors during immersions at 40 °C could contribute to increased performance by virtue of increased corticospinal excitability, decreased intracortical inhibition and increased intracortical facilitation (Sato et al. 2015).All these factors could explain the increase in muscle contraction speed.As the braking force was the same between the recovery methods, the increase in mechanical work was a result of the increase in pedaling frequency and, consequently, of the total distance traveled.The WAnT 2 performed after HWI recovery obtained an increase in two pedal rotations, approximately 15 m of the travel distance, at an average rhythm of 2.13 Hz, while after CWI, there was a reduction of close to one rotation, approximately 5 m, at an average rhythm of 2.03 Hz.The difference between the methods was slightly higher than three rotations or 20 m in 30 s of testing (Table 2).This difference was observed in both the total and average values assessed during the test (Fig. 7 and Table 2).
Thus, it is possible that both HWI and CWI modified the force-velocity (F-V) relationship of the muscles, which can be observed through the alteration of total pedal rotations during test time (Fig. 7).During WAnT 2 , the previous elevation of temperature increased the contraction velocity and pedaling frequency, while the decrease of temperature reduced the contraction velocity and pedaling frequency for the same braking force.In study with skinned fibers, conditions of low pH and high inorganic phosphate levels, the force and velocity of contraction are reduced for both fast and slow muscle fibers, independent of blood [La − ].However, muscle temperature can alter these effects: fast fibers are more sensitive at low temperatures, whereas at temperatures close to thermoneutral, slow fibers are more reduced (Fitts 2008).On the other hand, high temperatures increase maximum velocity (V 0 ) and maximum force of contraction (P 0 ) in both fiber types and reduce their curvature (a/P 0 ratio) of F-V relationship (see Fig. 5 in Fenn and Marsh 1935;Driss and Vandewalle 2013;Alcazar et al. 2019).Added to the previously mentioned factors, these effects can be explained by the altered sensitivity to Ca ++ during muscle contraction: decreased at low temperatures and increased at high temperatures (Fitts 2008).In vivo studies have confirmed similar effects of the temperature dependence of the F-V relationship (De Ruiter and De Haan 2000), even at temperatures higher (39 ℃) than the thermoneutral (Binkhorst et al. 1977).Despite not characterizing a fatigue situation (see CON in Fig. 6), the present study found an interaction effect (recovery method vs test) on pedaling speed and mechanical power for the same braking force after HWI, even after previous physiological stress (WAnT 1 ), contrary to other studies that found a greater reduction of performance in fatigue condition after immersion in thermoneutral temperatures (De Ruiter and De Haan 2000;2001).Unlike the in vivo studies cited above, the present research was conducted during a multi-joint exercise.
The immersion temperature was responsible for the increase of 2.2% in mean power output after HWI and the reduction of 2.4% after CWI when compared to CON, and for the decrease of about 3% in peak power output after CWI when compared to HWI (Fig. 8).Analyzing the power-velocity (P-V) relationship, the elevation of muscle temperature seems to increase both the speed of maximum power (P max ) and its value, the opposite happening with the reduction of muscle temperature (De Ruiter and De Haan 2000).In this study a similar behavior was observed in mean power output, while in peak power output this effect was significant only after cooling.It is important to remember that these values represent only one point of the P-V relationship (Fig. 8).Although this is not in consensus due to different methods of determination, the braking force of 7.5% BW seems to be adequate and close to the optimal intensity of P max during WAnT for expert endurance cyclists (Dotan and Bar-Or 1983;Driss and Vandewalle 2013).Similar repercussions of temperature in the P-V relationship have been seen both in skinned fibers (Fitts 2008) and in vivo studies, especially at high muscle contraction velocities (Sargeant 1987).The athletes in this study were instructed to pedal as powerfully as possible.
It is interesting that despite the increase in performance, the HWI was the recovery method considered by most athletes as having produced the worst subjective sensation, including two cases of nausea and malaise after performing the protocol.Conversely, CWI was considered by most to provide the best subjective sensation, despite being associated with decreased performance.In general, both conditions produced greater physiological disturbance in comparison to CON; however, they differed antagonistically in relation to performance and subjective sensation.As such, according to the model used in this study, a better physiological recovery (considering a faster return and lower values) was not associated with the best physical performance nor with the best subjective sensation, and the best physical performance was not influenced by the physiological and subjective parameters analyzed.

Practical implications
The use of immersion in a liquid medium for short periods of recovery from anaerobic exercises is a method that needs to be used with caution.Except for some parameters, immersion delays the recovery of cardiorespiratory variables and increases aerobic energy expenditure without generating advantages in relation to blood removal [La − ].The anaerobic performance increases after recovery in warm water and may decrease after recovery in cold water.While small, the changes in performance may be relevant from a competitive point of view, because a similar intervention before a training session, evaluation or competition, for example, could have an impact on your final result.Although recovery in cold water provides improvement in a subjective sensation during recovery, its use for short recovery intervals between series or competitive trials increases aerobic energy expenditure and reduces later anaerobic performance.On the other hand, recovery in warm water increases posterior anaerobic performance, despite subjective discomfort and increased energy expenditure.Another issue is that most previous studies implementing CWI as a recovery modality used temperatures ranging from 5 to 15 °C (Higgins et al. 2017;Stephens et al. 2017).While the temperature in the CWI protocol in the present study falls outside of the range, this may still constitute CWI, as evidenced by the divergent power output parameters between CON and HWI conditions.In contrast to other studies that describe their water immersion recovery without providing a detailed description of temperature control and maintenance, in this study, we developed an equipment capable of controlling and maintaining a constant temperature in a 250-L tank during the recovery period.If reaching and maintaining a constant immersion temperature close to 10 °C is difficult, maintaining a temperature of 20 °C is easier and approaches the limited conditions that a coach or athlete has in their area of expertise.In this regard, reaching and maintaining water at 20 °C in training or competition places is easier and less expensive.Moreover, in addition to causing physiological shivering, very low (as well as high) temperatures impair human comfort.
The choice of WAnT and a short recovery period is justified in order to provide a high anaerobic demand in a short time to perform two tests in sequence, significantly increasing the response of lactate and cardiorespiratory variables and thus potentiating the effect of the recovery method in both physiological responses and subsequent performance.Despite the existence of other effective means to generate thermal interference during the recovery period and performance improvement after the recovery period (Faulkner et al. 2013), this research model is appropriate for competition, evaluation and training situations where the recovery time between series is very short (not only cycling but other situations where anaerobic demand is predominant or very important), as well as providing experimental evidence on the effect of environmental conditions on the EPOC and VO 2 kinetics response and their relationship with anaerobic performance.

Study limitations
This study analyzed only men.We did not select women for the sample to avoid possible variation in physiological and performance responses triggered by the different thermal effects of water due to the sexual dimorphism (Gonzalez and Blanchard 1998) when subject responses are not corrected for body fat and size (Tikuisis et al. 2000).Furthermore, hormonal variations could trigger variations in recovery energy expenditure (Henderson 2014).Therefore, specific studies deserve to be carried out to test these hypotheses.Also, we did not analyze the maximal oxygen consumption.Despite being an important outcome for sample characterization, this study was concerned with evaluating responses to anaerobic performance.Nevertheless, we used secondary criteria based on physical training data to characterize the sample (De Pauw et al. 2013).A limitation of the calculation of the recovery kinetics of physiological variables was the existence of transition times between the end of the WAnT and the beginning of the recovery evaluation.The athlete's displacement and the procedures (removal of sneakers) were standardized in order to maintain similar transition times between the three recovery methods.and WAnT 2 , respectively).Pedaling rates for peak mechanical power was determined during the five-second time interval and for the mean mechanical power was determined as an average of 30 s. Open and filled symbols represent WAnT 1 and WAnT 2 , respectively.Square, circle, and triangle represent CON: Outside Water Immersion, CWI: Cold Water Immersion at 20 ℃, and HWI: Hot Water Immersion at 40 ℃, respectively.Braking forces (N) represents mean values for all conditions.Blue and red arrows represent significant temperature effects for CWI and HWI, respectively.For mean mechanical power, CWI is minor than CON and HWI and HWI is major than CON and HWI during WAnT 1 .For peak mechanical power, CWI is minor than HWI Although small, differences in transition time were found between immersed and CON recovery after WAnT 1 .The first seconds of rapid recovery phase are important for the calculation of τ.To mitigate this limitation, the transition time was inserted as a covariate in the statistical model in the variables that presented significance.A positive aspect is that, with the exception of HR, the variables of the rapid phase were not influenced by the test factor, since after the WAnT 2 , the differences in transition time between the methods were suppressed (Table 2).Another question was the 10 min recovery period's.This fact may have influenced the effect of immersion on the behavior of blood [La − ].It is possible that a longer period of recovery might have revealed different blood [La − ] actions between the recovery methods and consequently influenced the physical performance.However, variations of performance were observed independently of the blood [La − ] results and other physiological variables.Moreover, the recovery period did not allow the physiological variables to return to pre-exercise conditions.Despite this, the athletes were able to repeat the performance after CON.For this reason, the increase in performance verified after HWI, and the deterioration after CWI, justifies the protocol of this study and proves the effect of immersion temperature during recovery on anaerobic performance.Recovery during CON served as the control condition in this crossover design, and performance after CON was not significantly modified.This result helps to refute any suggestion that the thermal effect of HWI could have enhanced performance by virtue of not performing warm-up prior to WAnT.Although it is not a consensus (Inbar et al. 1996), there is evidence that warming up performed before WAnT does not improve its performance (Hawley et al. 1989).This result refutes any possible learning effect and reinforces the robustness of the method (Figs. 5 and 6).
Other aspect to be highlighted is the infeasibility of blinding participants in relation to recovery methods.Differences in the sensations of CWI vs HWI would be interpreted as a source of bias regarding subjective feelings of recovery and motivation for subsequent performance.However, participants could not predict the subjective sensation of the following experimental session.Furthermore, it is impossible to avoid the subjective sensation's influence on the subsequent test.This fact is precisely one of the effects of the recovery method on subsequent performance.Despite the subjective sensation responses showing high agreement among the participants, it was not in complete accord.Finally, although the physiological variables analyzed in this study do not include all the relevant adaptive responses (for example muscle oxygenation using NIRS and EMG to assess muscle activation), they are widely used in the evaluation of physical tests, control of training and monitoring of recovery periods during both applied and laboratory conditions.

Conclusion
In this crossover study design, the head-out, short-term water immersion during anaerobic post-exercise recovery showed distinct out of water and immersion temperature-dependent results.The kinetics of VO 2 during recovery was delayed, and EPOC presented higher magnitude during immersed recovery; moreover, HWI slowed the recovery of HR and VE .We hypothesized that the elevation of EPOC during immersion recovery could accelerate the recovery of blood [La − ] and consequently modify physical performance.However, the blood [La − ] during immersed recovery remained unchanged compared to out of water recovery conditions; even so, the recovery interventions affected the physical performance.While their performances after HWI increased and was superior to the other recovery methods, CWI deteriorated their performances, compared to CON.Moreover, in contrast to HWI, the improved subjective recovery sensation following CWI was accompanied by a worsened performance in subsequent anaerobic exercise.In summary, the effectiveness of water immersion as a resource for anaerobic performance recovery depends on the water temperature and physiological variables evaluated.And these parameters were not related to subsequent performance.Therefore, future studies in this area should try to establish an optimal temperature capable of promoting physiological recovery while minimizing harm or maintaining subsequent anaerobic performance.
World Medical Association's Declaration of Helsinki-Ethical Principles for Medical Research Involving Human Subjects 2008.
Consent to participate Written informed consent was obtained from all individual participants that performed this research.

Consent for publication
No personal data are published.

Fig. 3
Fig.3Monoexponential decay curve plotted from individual VO 2 data obtained during the recovery period after the Wingate Anaerobic Test: y 0 is the asymptotic value of the curve, A 1 is the amplitude of the response between the highest value of the curve and its asymptote, and τ (tau) is the time constant (TC) that represents the decay for the variable to reach 63% of its asymptote

Fig. 4
Fig. 4 Blood lactate evaluated over time at each two-second interval after the first and second Wingate Anaerobic Tests (post-WAnT 1 and post-WAnT 2 , respectively): CON (Control: non-immersed condition), CWI (Cold Water Immersion: 20 ℃), and HWI (Hot Water Immersion: 40 ℃).Mean blood [La − ] PRE on the y-axis for all experimental sessions.There was no difference between recovery methods.Differ-

Fig. 5 Fig. 6 Fig. 7
Fig. 5 Mechanical power output evaluated over time at each fivesecond interval during the first and second Wingate Anaerobic Tests (WAnT 1 and WAnT 2 , respectively): CON (Control: non-immersed condition), CWI (Cold Water Immersion: 20 ℃), and HWI (Hot Water Immersion: 40 ℃).Different letters indicate significant differences between time intervals (P < .001).*Significant differences between tests and the average values of the recovery methods for the same time interval (values not shown) (WAnT 1 and WAnT 2 , P < .001).Interaction between test and time intervals (P < .001)

Fig. 8
Fig.8Peak and mean mechanical power output at differents pedaling rates during the first and second Wingate Anaerobic Tests (WAnT 1 and WAnT 2 , respectively).Pedaling rates for peak mechanical power was determined during the five-second time interval and for the mean mechanical power was determined as an average of 30 s. Open and filled symbols represent WAnT 1 and WAnT 2 , respectively.Square, circle, and triangle represent CON: Outside Water Immersion, CWI:

Table 1
Different letters indicate significant differences between recovery methods (P < 0.05) CON control: non-immersed condition, CWI cold water immersion: 20 ℃, HWI hot water immersion: 40 ℃ significant covariance effect was found.The resistive braking force mean during WAnT was 53.4 ± 5.8 N.

Table 2
Mean and standard error of physiological and performance variables organized by recovery methods (CON, CWI and HWI) and test factor (WAnT Maximum mean values were observed at the 6th (post-WAnT 1 ) and 4th minute (post-WAnT 2 ) of recovery.The values verified at the 10th minute of recovery showed significant differences with the values of the 2nd minute only in post-WAnT 1 .Blood [La − ] PRE showed no differences between conditions (TableS1; Supplementary data).
2 and HR pre-exercise values (supine and sitting conditions) were added to test factor and resulted differences and interactions with VO 2 asymp and HR asymp values (not presented here).Mean Pedal Rotations and Distance of six 5-second intervals