Exercise training in normobaric hypoxia in endurance runners. I. Improvement in aerobic performance capacity

Stéphane P. Dufour, Elodie Ponsot, Joffrey Zoll, Stéphane Doutreleau, Evelyne Lonsdorfer-Wolf, Bernard Geny, Eliane Lampert, Martin Flück, Hans Hoppeler, Véronique Billat, Bertrand Mettauer, Ruddy Richard, Jean Lonsdorfer


This study investigates whether a 6-wk intermittent hypoxia training (IHT), designed to avoid reductions in training loads and intensities, improves the endurance performance capacity of competitive distance runners. Eighteen athletes were randomly assigned to train in normoxia [Nor group; n = 9; maximal oxygen uptake (V̇o2 max) = 61.5 ± 1.1 ml·kg−1·min−1] or intermittently in hypoxia (Hyp group; n = 9; V̇o2 max = 64.2 ± 1.2 ml·kg−1·min−1). Into their usual normoxic training schedule, athletes included two weekly high-intensity (second ventilatory threshold) and moderate-duration (24–40 min) training sessions, performed either in normoxia [inspired O2 fraction (FiO2) = 20.9%] or in normobaric hypoxia (FiO2 = 14.5%). Before and after training, all athletes realized 1) a normoxic and hypoxic incremental test to determine V̇o2 max and ventilatory thresholds (first and second ventilatory threshold), and 2) an all-out test at the pretraining minimal velocity eliciting V̇o2 max to determine their time to exhaustion (Tlim) and the parameters of O2 uptake (V̇o2) kinetics. Only the Hyp group significantly improved V̇o2 max (+5% at both FiO2, P < 0.05), without changes in blood O2-carrying capacity. Moreover, Tlim lengthened in the Hyp group only (+35%, P < 0.001), without significant modifications of V̇o2 kinetics. Despite similar training load, the Nor group displayed no such improvements, with unchanged V̇o2 max (+1%, nonsignificant), Tlim (+10%, nonsignificant), and V̇o2 kinetics. In addition, Tlim improvements in the Hyp group were not correlated with concomitant modifications of other parameters, including V̇o2 max or V̇o2 kinetics. The present IHT model, involving specific high-intensity and moderate-duration hypoxic sessions, may potentialize the metabolic stimuli of training in already trained athletes and elicit peripheral muscle adaptations, resulting in increased endurance performance capacity.

  • maximal oxygen uptake
  • time to exhaustion
  • competitive endurance runners

at sea level, it is well known that the training-induced improvements in endurance performance progressively level off as the aerobic fitness progresses. Therefore, the use of the metabolic stimulus provided by living (i.e., living high-training low) or training at altitude (i.e., living low-training high) has gained popularity in athletes to further enhance endurance performance. In this context, methods that impose short-term altitude exposure while exercising have progressively emerged to cope with the growing evidence that long-term altitude exposure possesses several detrimental effects, including a limited aerobic power, reducing both the metabolic and mechanical components of the total training load (24). To conciliate altitude training with a maintained training load, it has recently been proposed to use altitude in several but not all training sessions, included into a training program otherwise performed in normoxia [intermittent hypoxia training (IHT)] (38, 4446).

To date, these training programs have provided conflicting results in endurance athletes (38, 44, 46), which could be due to the various combinations of duration and intensity of the hypoxic training sessions employed (32). Accordingly, improvement of performance in competitive swimmers has not been observed after an IHT program, including very short high-intensity (30–60 s) hypoxic sessions (44), whereas longer periods of high-intensity hypoxic exercise (70 s to 3 min) improved maximal power output at sea level in professional cyclists (39). In addition, significant maximum oxygen uptake (V̇o2 max) improvement at sea level has been reported in trained subjects after hypoxic exercise bouts of 2- to 12-min duration (38). Recent evidence also demonstrated no beneficial effects of IHT programs, when the hypoxic exercise intensity is set below 80% of normoxic V̇o2 max (44, 46). Collectively, these findings point to a pivotal role for a minimal hypoxic exercise duration and intensity in IHT models, especially in trained athletes. Based on these observations, we assumed that two successive hypoxic training bouts, of 12–20 min, performed at the second ventilatory threshold (VT2) (∼80% of normoxic V̇o2 max) are likely to comply with the above information. Moreover, integrated within the usual normoxic training of competitive runners, the intermittent nature of such specific hypoxic sessions would allow maintaining high levels of total training load and may elicit significant improvement of endurance performance capacity.

Characterization of the endurance performance capacity in athletes involves incremental exercise testing, allowing for the determination of the ventilatory thresholds [first ventilatory threshold (VT1) and VT2], V̇o2 max, as well as their associated minimal running velocities (vVT1, vVT2, and vo2 max). Additionally, since vo2 max falls among the significant predictors of endurance performance (4, 5), the time to exhaustion at vo2 max (Tlim) is thought to constitute an important determinant of the endurance performance capacity. Despite its athletic relevance, the effect of IHT program on Tlim in endurance athletes remains unknown. Since the maximal rate (i.e., V̇o2 max or vo2 max) (6, 23) and/or kinetic changes in the O2 flux adjustment (13) are expected to contribute to Tlim performance, the possible influence of IHT on both of these respective properties of aerobic metabolism is also not elucidated.

Therefore, the purpose of this study was to test the hypotheses that an original IHT model, including two weekly moderate-duration (24–40 min) and high-intensity (VT2) hypoxic sessions within the usual normoxic training of already trained athletes, 1) improves running velocities at sea level due to amelioration of aerobic energy provision, including V̇o2 max; and 2) lengthens Tlim at sea level with concomittant adaptations of aerobic metabolism properties, mainly V̇o2 max and/or oxygen uptake (V̇o2) kinetics.



Eighteen highly trained male distance runners were recruited from local athletic teams and completed the study before the beginning of their competitive season. Their main physical and physiological characteristics are shown in Table 1. After all the potential risks were explained, the athletes gave a voluntary written consent to participate to the protocol, approved by our hospital and national review boards. In the weeks before and during the study, the subjects lived under the altitude of 300 m and were engaged in a regular training schedule comprising five training sessions per week, including two weekly training sessions performed specifically at VT2 (49). Their respective individual training schedule remained unaltered during the experimental period. All were highly motivated to participate in the study, familiar with treadmill running, and with current 10,000 m or equivalent personal-best times of <35:00 (min:s).

View this table:
Table 1.

Anthropometric data and performance capacity of the subjects

Experimental Design

As shown in Fig. 1, the study was organized in four successive phases: a basal medical examination, the pretraining treadmill performance evaluation, the training process, and the posttraining treadmill performance evaluation.

Fig. 1.

Study design with the 4 phases of the experimental protocol. Graphic, Incremental running tests;Graphic, all-out running tests in normoxia. Normoxia: inspired O2 fraction (FiO2) = 20.9%. Hypoxia: FiO2 = 14.5%. Phases 1, 2, 3, and 4 are the respective experimental phases (see text for details). vVT2, running velocity corresponding to the second ventilatory threshold.

Basal medical examination.

Two weeks before the beginning of the training period, each subject came to the laboratory for anthropometric measurements, physical examination, resting electrocardiography, and echocardiography recordings. To verify their exercise and hypoxic tolerance under careful cardiac monitoring, all athletes also performed maximal graded cycle tests in normoxia and hypoxia. These tests did not reveal any abnormality that could prevent the subjects from being included in the experimental protocol.

Pre- and posttraining treadmill performance evaluation.

In the week before and after the training intervention, all of the subjects performed three exercise tests on a motorized treadmill (Gymrol 2500 SP, Tecmachine), which were separated by at least 24 h of rest: 1) a treadmill incremental exercise test (IET) to exhaustion in Nor [IETN; inspired O2 fraction (FiO2) = 20.9%]; 2) a treadmill IET to exhaustion in Hyp (IETH; FiO2 = 14.5%, equivalent to an altitude of 3,000 m); and 3) a normoxic all-out test at pretraining vo2 max. For a given subject, all tests were performed at the same time of day in a climate-controlled environment (21–23°C).

Training program.

During the 6 wk of the study, both groups continued their usual training program (5 sessions/week), including their two weekly sessions at VT2 that were performed in the laboratory. All of the laboratory training sessions were performed under careful supervision of an experimented physician. For the group who trained in normoxia (Nor group), VT2 was determined during the IETN, and for the group who trained in hypoxia (Hyp group), VT2 was determined during the IETH. Each VT2 session began with a 10-min warm-up at 60% V̇o2 max (<VT1), followed by two periods at VT2 (time run at VT2 specified in Fig. 1), separated by 5-min recovery at 60% V̇o2 max. For the Hyp group, the subjects trained under hypoxic conditions only during the running periods at VT2 by breathing through face masks connected to a mixing chamber via appropriate tubing. Warm-up and recoveries were performed under normoxia. The training load during the laboratory sessions was organized into two 3-wk periods in which the exercise duration at VT2 increased progressively (Fig. 1). At the 4th wk, the training velocity was readjusted to maintain an exercise heart rate (HR) corresponding to the one achieved at the first training session. Throughout the study, each athlete underwent a total of 12 controlled laboratory training sessions. No athletes withdrew from the study before the achievement of the posttraining treadmill performance evaluation, and none complained of health complications throughout the study.


Altitude simulation.

Normobaric hypoxic conditions corresponding to an altitude of 3,000 m (FiO2 = 14.5%) were simulated by diluting ambient air with nitrogen via a mixing chamber, with the dilution being constantly controlled by a Po2 probe (Alti-Trainer200, Sport and Medical Technology). This device allows the inspired Po2 to be set at a predetermined value to simulate altitude. The precision of the Po2 is of ±0.82 Torr. The respiratory effort induced by the device at 6 l/s was negligible (<0.01 W).

Treadmill tests.

The IETN or the IETH were performed in random order on a motorized treadmill with 0% slope, to determine VT1, VT2, V̇o2 max, the associated velocities, and the running economy (RE) in both conditions of oxygen availability. During each IET, the initial running speed was set at 10 km/h and increased by 1 km/h every 2-min until volitional exhaustion. Each subject was encouraged to give a maximum effort. Arterialized blood samples were obtained from the earlobe at rest, at exhaustion, as well as at the first and third minute of recovery to determine total blood lactate concentration ([La]).

The all-out running test was performed in normoxia at pretraining vo2 max, i.e., the same absolute running speed before and after training. The test began by 10-min warm-up at 60% of the subject's vo2 max (lower than vVT1 in all subjects). The subjects were then connected to the test equipments during a 5-min period of rest and immediately asked to run at their individual vo2 max for as long as possible. The transition from rest to vo2 max occurs within a 20-s delay (range 17–23 s), necessary for the treadmill to reach the desired speed. No information about the time elapsed was provided to the athletes. During this test, arterialized blood samples were obtained from the earlobe at rest, at exhaustion, as well as at the 1st and 3rd min of recovery to determine total blood [La].

HR monitoring.

During all of the running tests, as well as during the controlled training sessions, HR was continuously monitored by telemetry (Polar Vantage, Kempeley, Finland).

Gas exchange measurements.

During all tests, inspiratory (V̇i) and expiratory minute ventilation (V̇e), V̇o2, and carbon dioxide output (V̇co2) were measured breath by breath with an open-circuit metabolic cart with rapid O2 and CO2 analyzers (Sensor Medics MSE, Yorba Linda, CA). Before each individual exercise test, the pneumotachograph was calibrated with several strokes given by a 3-liter calibration syringe. The gas analyzers were calibrated by using reference gases with known O2 and CO2 concentrations (12% O2, 5% CO2). FiO2 and fraction of O2 in the expired air (FeO2) were analyzed continuously for each breath. Therefore, V̇o2 was calculated in normoxia and hypoxia by the following formula, where all parameters are expressed in stpd conditions: V̇o2 = V̇i × FiO2 − V̇e × FeO2.

During the IET, each athlete was encouraged to give a maximal effort. Peak treadmill velocity was defined as the last achieved running speed sustained for at least 30 s. V̇o2 max was always defined as the highest 30-s averaged V̇o2 value. As previously described by Billat and Koralsztein (4), vo2 max was defined as the minimal velocity at which V̇o2 max occurred. In detail, if V̇o2 max was reached during the last stage, which was maintained >90 s, that particular velocity was taken as vo2 max. If that velocity eliciting V̇o2 max was sustained <60 s, then vo2 max was taken as the velocity at the previous stage. If that velocity eliciting V̇o2 max was maintained between 60 and 90 s, then vo2 max was considered to be equal to the velocity during the previous stage plus the half velocity increase between the last two stages, i.e., (1 km/h)/2 = 0.5 km/h (29). Ventilatory thresholds were assessed by using established criteria (3, 49). VT1 corresponds to the break point in the plot of V̇co2 as a function of V̇o2. At that point, the ventilatory equivalent for O2 (V̇e/V̇o2) increases without an increase in the ventilatory equivalent for CO2 (V̇e/V̇co2). VT2 was located between VT1 and V̇o2 max, when V̇e/V̇co2 starts to increase while V̇e/V̇o2 continues to increase. The oxygen pulse (O2p) was calculated as the ratio between V̇o2 and HR, also representing stroke volume times arteriovenous oxygen difference [Δ(a-v)O2] (30). RE was defined as the rate of V̇o2 for a given submaximal work rate (9). Therefore, RE corresponds to the 1-min average of the V̇o2 values recorded at the end of the 12 km/h stage during each IET. This speed was lower than VT1 for all of the subjects in both environmental conditions and allows an estimation of RE for an exercise intensity expected to be mainly aerobic. To provide additional insights in the effect of IHT on RE, we also determined RE at 18 km/h in IETN and at 15 km/h in IETH. These running speeds amounted to ∼92 and 90% of the respective normoxic and hypoxic vo2 max, corresponding to recommended speed for RE determination in athletes (10).

Blood O2-carrying capacity and lactate.

On the first day of the treadmill performance evaluation before and after training, blood was drawn from an antecubital vein in each group to immediately measure hematocrit (Hct) and hemoglobin concentration. Earlobe blood samples obtained during all running tests were also immediately analyzed for total blood [La] by an enzymatic method.

Oxygen saturation.

During each exercise test, hemoglobin saturation was monitored continuously by earlobe pulse oximetry (Oxypleth, Novametrix-Medical System).

o2 Kinetics

Data modelization.

To describe the V̇o2 kinetics [V̇o2(t)] during the all-out test, we used a mathematical model with two exponential functions (2): Math(1) where U1 = 0 for t < td1 and U1 = 1 for t ≥ td1; U2 = 0 for t < td2 and U2 = 1 for t ≥ td2; V̇o2b is the rate of V̇o2 at rest before the start of the all-out test; A1 and A2 are the asymptotic amplitudes for the first and second exponential terms, respectively; τ1 and τ2 are the time constants and represent the time to reach 63% of the total amplitude of the respective fast and slow V̇o2 components; td1 and td2 represent the time delays for the fast and the slow components, respectively. As the initial cardiodynamic phase of the V̇o2 adjustment to a rest-to-exercise transition does not influence the fast component of V̇o2 (36) and because we focused on the fast and slow components of the V̇o2 response, the cardiodynamic phase was excluded from analysis by removing the data from the first 20 s of the all-out test. The parameters of the model were determined with an iterative procedure that minimizes the sum of the mean squares of the differences between the model V̇o2 estimates and the corresponding V̇o2 measurements. To exclude aberrant breaths from analysis, breath-by-breath V̇o2 values that were greater than three standard deviations from the modeled V̇o2 were removed and assumed to represent events unrelated to the physiological response of interest (31, 39). These values represented <1% of the total data.

Slow component of V̇o2 kinetics.

Because the asymptotic value of the second exponential term is not necessarily reached at the subject's exhaustion, the amplitude of the slow component was computed as A2 (7): Math(2) where Tlim is the time at the end of the all-out exercise test. Moreover, to compare the amplitude of the V̇o2 slow component at consistent time before and after training, we also calculated the amplitude of the V̇o2 slow component achieved posttraining when the subjects attained their pretraining Tlim value (A2old).

O2 deficit calculation.

According to Whipp and Ozyener (51), the fast component of the V̇o2 kinetics represents an “expected V̇o2,” whereas the slow component is the manifestation of an “excess V̇o2” occurring later during exercise (i.e., after td2). Consequently, the oxygen deficit (O2def) is estimated from the area between the fast-component response curve and the fast-component asymptote (13): Math(3) where O2def is in milliliters, td1 and τ1 are in seconds, and A1 is in milliliters per second.

Computation of the time sustained at pretraining V̇o2 max.

Besides Tlim, which could be considered as a mechanical parameter of endurance performance (reflecting the total mechanical work performed at vo2 max), we also calculated a metabolic correlate (Eq. 4), from the time sustained while the athlete ran at >95% of pretraining V̇o2 max (Tlim @ V̇o2 max). This percentage was chosen to account for a 5% random error in the determination of V̇o2 max (33) and also because all athletes did not necessarily reach 100% V̇o2 max in Tlim testing (13). Math(4) where Tlim is the time to exhaustion while the athletes ran at the pretraining minimal velocity associated with V̇o2 max (s), and the time to attain V̇o2 max (TA V̇o2 max) corresponds to the time necessary to reach 95% of pretraining V̇o2 max (s). Depending on whether the V̇o2 kinetics were better described by a mono- or a double-exponential model, TA V̇o2 max was computed from the equations below.

1)For the monoexponential model (fast component in Eq. 1) Math 2) For the double-exponential model (fast + slow component in Eq. 1) Math

Evaluation of Training

All athletes were asked to report their individual training schedule into detailed training logs, including duration, distance, and intensity of each training sessions. Laboratory as well as field work bouts were taken into account to provide both quantitative and qualitative characterization of the overall training load. Duration and intensity of the training sessions performed out of the laboratory were assessed based on the running velocity spread out in four intensity zones: low (<vVT1), moderate (vVT1vVT2), heavy (vVT2vo2 max), and severe intensity (>vo2 max).


Whether a mono- or biexponential model better described the V̇o2 kinetics during the all-out tests was determined using a Fisher test. We used the bootstrap method to obtain an estimation of the accuracy of the parameters describing the V̇o2 kinetics (7, 8, 17). This method, creating 1,000 different samples of the same size than the original data set, allows the determination of a coefficient of variation for each mathematical parameter on an individual basis.

Data were first tested for distribution normality and variance homogeneity. Subsequently, the differences between groups before the training period were analyzed with the Mann-Whitney procedure. To test for both treatment (Hyp vs. Nor) and time (before vs. after) effects on each of the measurements during the training period, we used a two-way ANOVA for repeated measures. When significant modifications were found, the Student-Newman-Keuls post hoc procedure was performed to localize the difference. Pearson linear regression analysis was used to determine any potential linear relationship between variables. All statistical analyses were performed with the SigmaStat 3.0 software (SPSS, Chicago, IL), and the level of significance was chosen for P < 0.05. Values are means ± SE.


The anthropometric and treadmill performance characteristics of the athletes are shown in Table 1. No significant differences were reported between the two experimental groups before the training period. Moreover, in both groups, the training period did not modify anthropometric and blood parameters, including body mass [Hyp after: 70.5 ± 2.2 kg, nonsignificant (NS); Nor after: 71.3 ± 2.2 kg, NS], hemoglobin (Hyp after: 15.8 ± 0.5 g/dl, NS; Nor after: 15.7 ± 0.5 g/dl, NS), and Hct (Hyp after: 46.4 ± 1.5%, NS; Nor after: 46.9 ± 1.2%, NS).

Training Load

Laboratory training sessions.

At the beginning of the study and according to the training environment, the Hyp group trained at a significantly lower running speed (Table 2). These different running speeds corresponded to the same exercise HR, whether expressed in absolute (Hyp: 166 ± 3 vs. Nor: 172 ± 3 beats/min; NS) or in relative value (Hyp: 96 ± 1 vs. Nor: 94 ± 1%; NS). During the first VT2 training session in the Hyp group, blood oxygen saturation stabilized at a value of 80 ± 1%. At the 4th wk, the running speed of the VT2 bouts was increased by 0.4 ± 0.1 km/h for the Hyp group, but not modified for the Nor group. As a result, HR values remained unaltered throughout the 6-wk intervention in both groups as it was the case for blood oxygen saturation in the Hyp group. For the Hyp group, the total duration of hypoxic exposure amounted to 384 min (i.e., week 1 + week 2 +…+ week 6) and was well tolerated.

View this table:
Table 2.

Training load characteristics

Total training load.

The total training load (i.e., field and laboratory training sessions) was comparable in both groups. During the 6-wk training, Hyp and Nor groups performed, respectively, 33.0 ± 0.6 and 31.2 ± 1.7 training sessions, leading to no difference in total training time and total training distance (Hyp: 2,013 ± 114 min and 478 ± 27 km vs. Nor: 2,095 ± 158 min and 498 ± 40 km). Consequently, the averaged running speed over the 6-wk training intervention was similar in both groups (Hyp: 14.3 ± 0.2 vs. Nor: 14.2 ± 0.2 km/h). No significant differences appeared either in total time or in total distance run in the respective intensity zones (Table 2).

Exercise Capacity in Hypoxia


Table 3 and Fig. 2A report the effects of the 6-wk training on results of the IETH. Only the Hyp group significantly improved submaximal and maximal running velocities (Table 3) under hypoxia. Indeed, vVT1, vVT2, and vo2 max increased, respectively, by +7, +8, and +5% after IHT. The V̇o2 associated with these velocities improved in the same proportions by +7, +7, and +5%, respectively (Fig. 2A), but RE did not change. The maximum O2p (O2pmax) improved (+5%) only in the Hyp group after IHT (Table 3). Conversely, the Nor group demonstrated no improvement of all of these parameters under hypoxic conditions.

Fig. 2.

Percent change in maximal oxygen uptake (V̇o2 max) in hypoxia (A) and in normoxia (B) for each individual subject of the hypoxia group (Hyp; left) and normoxia group (Nor; middle), before and after the training program. Horizontal solid lines with vertical bars represent group changes. Horizontal dashed lines are the zero level. Right: absolute mean changes for all subjects from the Hyp and Nor groups. Hyp and Nor represent the groups of subjects that performed the laboratory-controlled training sessions under hypoxic or normoxic conditions, respectively. All values are presented as means ± SE. Significant differences before vs. after training, *P < 0.05.

View this table:
Table 3.

Running velocities, running economy, and selected maximal physiological parameters measured in normoxic and hypoxic incremental tests before and after the 6-wk training period

Exercise Capacity in Normoxia


vo2 max improved significantly by +4 and +3% and vVT2 increased significantly by +5 and +3% in the Hyp and Nor groups, respectively (P < 0.05), under normoxic conditions (Table 3). However, only the Hyp group significantly enhanced V̇o2 max as well as V̇o2 at VT2 by +5 and +7%, respectively (P < 0.05), with no modification of the RE (Fig. 2B). Again, O2pmax increased (+6%) only in the Hyp group after IHT (Table 3). The Nor group disclosed no significant changes, neither for exercise V̇o2 nor for RE.

All-out exercise test.

The all-out exercise tests were performed in normoxia at the same absolute running velocity before and after training, i.e., pretraining vo2 max. After training, this speed amounted to 96 and 97% of the posttraining vo2 max for the Hyp and Nor group, respectively, therefore corresponding to the same relative running speed in both groups. As shown in Fig. 3, training significantly enhanced Tlim in the Hyp but not in the Nor group (+35 vs. +10%, P < 0.05). Similar changes in the time sustained at pretraining vo2 max were obtained when the transition period required for treadmill speed stabilization was subtracted from Tlim (+35 vs. +10%, P < 0.05). Concomitantly, the end-exercise V̇o2 achieved during the all-out test increased in the Hyp group only (+6%, P < 0.05), whereas the maximal [La] values remained unchanged after training (Table 4).

Fig. 3.

Percent change in time to exhaustion for each individual subject before and after the training program in the Hyp (A) and Nor (B) groups. Horizontal solid lines with vertical bars represent group changes. Horizontal dashed lines are the zero level. C: absolute mean changes for all subjects from the Hyp and Nor groups. Values are presented as means ± SE. Hyp and Nor represent the group of subjects who performed the laboratory-controlled training sessions under hypoxic or normoxic conditions, respectively. Tlim, running time to exhaustion at the pretraining minimal velocity associated with V̇o2 max. Significant differences before vs. after training, *P < 0.05.

View this table:
Table 4.

Training effects on the time until exhaustion and the parameters of the V̇O2 kinetics

The kinetics of V̇o2 response of a typical subject from the Hyp and Nor group are shown in Fig. 4. Training did not modify parameters of the fast component of V̇o2 kinetics (Table 4) and O2def remained unchanged (Hyp group: before 3,319 ± 266 vs. after 3,372 ± 469 ml O2, NS; Nor group: before 2,793 ± 239 vs. after 2,563 ± 169 ml O2, NS). A slow component of V̇o2 kinetics was consistently observed in only five subjects in the Hyp and six subjects in the Nor group, respectively. Its computed amplitude (A2) did not change after training, even when expressed at similar exercise time after vs. before training (A2old). Neither V̇o2 kinetics alterations nor IETN-derived factors significantly correlated with the observed modifications in Tlim, either in absolute or in delta (after vs. before) values. There was no difference between groups in TA V̇o2 max or in Tlim at V̇o2 max before training (Table 4). However, Tlim at V̇o2 max significantly improved by 2.5 times only in the Hyp group after IHT, without modification of TA V̇o2 max.

Fig. 4.

Kinetics of the oxygen uptake (V̇o2) response of one representative individual from the Hyp group (A) and Nor group (B) during the all-out run at the pretraining minimal velocity associated with V̇o2 max, before (▪) and after (□) the 6 wk of intermittent hypoxia training program. Note that Tlim changes appeared, despite no modification in the kinetics of the V̇o2 response in both subjects.


Major Findings

This study demonstrates that, when the hypoxic sessions of an IHT program features moderate duration (24–40 min) and high intensity (VT2), significant improvements of V̇o2 max are obtained in already trained athletes, not only at altitude but also at sea level. Despite similar total training load (i.e., absolute and relative values), no such amelioration in the maximal rate of O2 fluxes was observed in a control group exercising under permanent normoxia. The second finding of this work is that the present IHT program significantly lengthened Tlim, specifically in the Hyp group, without significant changes in V̇o2 kinetics. These results suggest that IHT did not change the control of O2 flux adjustment to high-intensity exercise in competitive runners. Moreover, Tlim improvement in the Hyp group was correlated neither with V̇o2 max nor with ventilatory thresholds changes.

Maximal Aerobic Capacity and Ventilatory Thresholds

In hypoxia.

This study demonstrates that the present IHT program elicits significant improvements of maximal and submaximal running velocities under hypoxia (vo2 max, vVT2, and vVT1). Accordingly, all of the athletes of the Hyp group required an increase of the training velocity under hypoxia (+0.4 km/h) to maintain the initial HR values throughout the 6-wk IHT program. Since no RE changes resulted from the training period, the improvements observed in running speeds are mainly due to significant increases in the associated O2 flux rates in the Hyp group only (V̇o2 max and V̇o2 at the ventilatory thresholds). These findings expand the observations reported by Terrados et al. (43) in professional cyclists, demonstrating a specific increase of exercise capacity under hypoxia after altitude training only. Moreover, the present data also extend to already trained athletes the results obtained in untrained subjects, in which some consensus has been reach about the beneficial effect of hypoxic training on V̇o2 max at altitude (21, 47).

At sea level.

The effects of IHT on the aerobic performance capacity at sea level remains highly debated, especially in trained subjects (32). Despite both groups improving their running velocities at sea level (vo2 max and vVT2) in quite near proportions, the underlying physiological adaptations may well have been different. vo2 max and vVT2 increased in the Nor group, through concomitant changes of V̇o2 and RE values (although not statistically significant). Conversely, one important result of this study is that the running speed improvements of the Hyp group were associated with increases in V̇o2 max and V̇o2 at VT2, with no RE alterations. These findings suggest that a normoxic training effect was present in the Nor group over the 6-wk period and that this effect was further potentialized by IHT in the Hyp group, through an additional effect of IHT vs. normoxic training on aerobic power. This amelioration of aerobic power in the Hyp group is further exemplified by the increased V̇o2 at exhaustion during the all-out test. According to the specific intensity and duration of the present hypoxic training sessions, our results are in agreement with previous observations (38, 44, 46). Studies reporting no improvement in V̇o2 max after IHT either used lower hypoxic exercise intensity (at VT1) (46) or shorter hypoxic exercise bouts (0.5–1 min) (44). On the other hand, similar increase in V̇o2 max has been recently reported with an IHT model, including longer periods of hypoxic exercise (2–12 min) (38). A specific oxygen-sensing transcription factor, the hypoxia-inducible factor-1α (HIF-1α), is expected to play a pivotal role for the functional adaptations to hypoxic training (1, 11, 47). Of note, the duration and intensity of the hypoxic exercise bouts included in the present IHT model are in good agreement with the properties of HIF-1α expression at the cellular level in humans. Not only does the half-time of the HIF-1α response to hypoxia fall in the range of 12–13 min (25), but also the magnitude of this response varies exponentially with the degree of Hyp in the physiological range (26). These observations further reinforce the necessity to combine a minimum duration and intensity of hypoxic exercise in IHT programs, to reduce oxygen pressure in the active muscle (37) and achieve a substantial HIF-1α response, resulting in peripheral muscle adaptations. Consequently, present and previous results suggest that the combination of sufficient hypoxic exercise intensity and duration within IHT programs is of paramount importance to obtain significant performance ameliorations in already trained athletes. An additional advantage of the hypoxic sessions in the present IHT design (i.e., 19% of the total training time in the present study) is the possibility to maintain the usual training load, which could also participate in the V̇o2 max improvement that we observed.

Some of our findings let us consider that peripheral adaptations might have been involved. We observed that O2pmax improved in the Hyp group only after IHT. Because O2p represents the product of stroke volume with Δ(a-v)O2, and because invasive experiments have shown that O2pmax is largely determined by Δ(a-v)O2 (35, 42), O2pmax is likely to have increased via an Δ(a-v)O2-mediated mechanism after IHT, suggesting an enhanced tissue O2 extraction. Because our study was not designed to investigate O2 extraction, further studies are needed to verify this hypothesis. Nevertheless, several muscle changes have already been observed after hypoxic training programs in endurance-trained subjects, such as larger deoxygenation in active muscles (28) and, although not reaching significance, a 36% increase in capillary density (43), supporting the concept of an improved O2 extraction after IHT. Moreover, modelization studies have suggested that exercising in Hyp may increase the relative contribution of peripheral factors (i.e., muscle perfusion, peripheral diffusion, and mitochondrial capacity) to O2 delivery and utilization (14, 15, 19, 48). We believe that the intensity and duration of the hypoxic exercise bouts included in the present IHT program are sufficient to induce the signaling cascade initiated by HIF-1α, leading to molecular and tissue changes within the exercising skeletal muscles of our Hyp subjects (34). The results disclosed in the two companion papers of our study, appearing in the present issue, also support this concept, at least in part. Conversely, as far as O2 transport is concerned, we observed that hemoglobin and Hct were similar in both groups, before vs. after training, in agreement with previous reports (22, 28, 34, 38, 43). Together with the unchanged maximum HR (HRmax), these results suggest that O2 delivery capacity is unlikely to represent a major cause of the V̇o2 max improvement of the Hyp group after IHT.

Tlim at vV̇o2 max and Oxygen Kinetics

A major finding of the present study is that Tlim is specifically improved after IHT (+35%) but unchanged after normoxic training. Due to the exponential shape of the running velocity/time-to-fatigue relationship, our observed 3.7-min lengthening of Tlim suggests that larger improvements of endurance time at lower velocities may have occurred. Thus this observation can be considered as a hallmark of an enhanced performance capacity in middle and long-distance running events. Consequently, Tlim lengthening in the present study extends previous findings, demonstrating that 3 wk of IHT dramatically delayed fatigue during a submaximal constant-load test in elite triathletes (45).

To date, the mechanisms leading to Tlim improvement remain poorly understood. It has been proposed that normoxic training may lengthen the endurance time at a given absolute running velocity, due to increases of vo2 max and/or submaximal running velocity (velocity at the lactate threshold), reducing the relative running speed the subjects have to sustain (i.e., expressed in percentage of the posttraining vo2 max) (12, 23). In the present study, we did not find any correlation between Tlim changes and alterations of maximal (V̇o2 max) and submaximal (ventilatory thresholds) V̇o2 nor with their associated velocities (vo2 max, vVT2, and vVT1). Therefore, it is unlikely that changes in O2 fluxes (i.e., V̇o2 max) and/or running velocities (i.e., vo2 max) are the major causes of the Tlim improvement that we observed. However, as V̇o2 max and ventilatory thresholds improved concomitantly with Tlim in the Hyp group, we cannot rule out the possible relevance of these changes, and this point warrants further investigations.

Alternatively, V̇o2 kinetics have also been proposed as a determinant of Tlim that may be improved after normoxic training. To the best of our knowledge, the effect of hypoxic training on V̇o2 kinetics has never been reported, especially in already trained athletes. A speeding of V̇o2 adjustment has been proposed as a potential contributor of the delayed fatigue after high-intensity training at sea level (13). These changes are expected to reduce the reliance toward anaerobic metabolisms for energy provision, which have been reported to amount to ∼15% of energy expenditure during such Tlim testing (18). Nevertheless, we failed to observe such a mechanism, as illustrated by an unchanged fast component of V̇o2 kinetics, leading to unaltered O2def in both experimental groups. Additionally, sea level training was often demonstrated to reduce the amplitude of the V̇o2 slow component, thereby contributing to improve exercise tolerance and delay fatigue (20). Again, we recorded no alterations in the V̇o2 slow component, even when expressed at consistent exercise time before vs. after IHT (A2old). Taken together, the unchanged fast and slow components of V̇o2 kinetics suggest that the dynamic control of O2 fluxes is not a likely contributor to Tlim changes after IHT in already trained athletes. Therefore, neither the rates of O2 fluxes nor V̇o2 kinetics significantly account for the Tlim lengthening that we observed, suggesting that IHT may improve Tlim by specific, hypoxic-related adaptations.

A 2.5 times longer Tlim at V̇o2 max was observed in the Hyp group after IHT, indicating an improved capacity to sustain high levels of O2 fluxes close to or above pretraining V̇o2 max, before exhaustion occurs. This observation appears, despite unchanged [La] values recorded at exhaustion during the all-out test after vs. before IHT. Collectively, these findings suggest either a slower rate of blood lactate accumulation and/or a better tolerance of high levels of blood lactate after IHT. This might be associated with a concomitant amelioration of metabolite exchange and/or removal, contributing to enhance cellular homeostasis, thereby delaying the time at which fatigue occurs. This idea has already been suggested by a previous study, demonstrating that Tlim is related to the capacity of lactate exchange and removal. Due to its coupled transport with H+ (27), an improved lactate exchange and removal could have contributed to slow down the progressive lowering of muscle pH while running at pretraining vo2 max. Although purely speculative in the present study, additional supports for the peripheral hypothesis underlying the improvement of endurance performance capacity after IHT are presented in the two following papers appearing in this issue. The second companion paper of the present study suggests that IHT induces qualitative mitochondrial changes leading to an enhanced channeling of energy within the muscle cell, whereas the third companion paper shows that IHT training induces transcriptional changes, potentially mediated by HIF-1α, leading to enhanced metabolite exchanges and improved aerobic metabolism within the skeletal muscle cell.

Limitations of the Study

A limitation of the present study is related to the IHT design and management of training intensities. First, we speculated that VT2 might be more effective in IHT designs than lower (i.e., VT1) or higher (i.e., vo2 max) training intensities, because of the achievement of a unique combination of intensity and duration of the hypoxic training stimulus. Moreover, this protocol was chosen as it allowed the usual training load of athletes to be unaltered (Table 2). Nevertheless, we did not test this hypothesis in the present study by including additional experimental groups training at either lower or higher intensity during the hypoxic sessions. Therefore, it remains to be determined whether different hypoxic training intensities and durations elicit similar beneficial effects on endurance performance capacity in already trained athletes. Especially including a group trained at, or close to, VT1, would have been helpful and remains to be done.

Second, only the Hyp group required its laboratory running speed to be increased at the end of week 3 to maintain the initial HR level, raising the question as to whether the Hyp group may have trained harder than the Nor group. We believed that this possibility is not supported by the unchanged VT2 at the end vs. the beginning of training, when expressed in percentage of posttraining vo2 max, indicating that both groups trained at the same relative intensity during the laboratory sessions (Table 3). Nevertheless, a different time course of training speed and vo2 max improvements may have led to a transient increase (i.e., weeks 4 and 5) in relative training intensity, thereby potentially acting as a confounding factor in our results. We believe that this possibility should have been counterbalanced by the transient lower relative intensity that could be expected in the Hyp group just before training speed adjustments (i.e., weeks 2 and 3). Therefore, differences in relative training intensity, if present, may have probably played a minor role in the present study. Nevertheless, future studies need to incorporate serial V̇o2 max testing to completely eliminate this possibility.

On the same token, an additional Tlim test performed at the new vo2 max after training (same relative intensity before vs. after training) could have been helpful to clarify the role of V̇o2 max and vo2 max in the improvement of Tlim that we observed in the Hyp group (+35%). However, since both groups improved vo2 max in quite near proportions, the posttraining Tlim was performed at a similar relative intensity in both groups (96 vs. 97% of the posttraining vo2 max in the Hyp and Nor group, respectively). Therefore, although not performed at 100% of posttraining vo2 max, the changes in relative testing intensity are unlikely to account for the Tlim improvements that we observed.

On a methodological standpoint, it can be argued from our relatively low maximum respiratory exchange ratio and maximum [La] values (Table 3) that V̇o2 max might have been underestimated. Nevertheless, 67–89% of the subjects reached a true V̇o2 plateau (i.e., always at least 6 of 9 subjects in each test), and the HRmax were close to (97%) the theoretical HRmax. Moreover, V̇o2 max as well as HRmax were significantly higher on the treadmill than the ones previously obtained on the cycle ergometer at the time of subjects' basal medical examination. Conversely, these parameters were similar between IET and Tlim testing. Therefore, we believe that true V̇o2 max has been at least closely approached. On the other hand, our RE values were estimated at moderate mainly aerobic (12 km/h) and high (18 and 15 km/h in normoxia and hypoxia, respectively) running speeds, yielding results consistent with previous reports (40, 41, 50). Nevertheless, since they were not measured at steady state during constant load exercise, these RE values must be interpreted with caution, until appropriate RE testing is done by further investigations.

In conclusion, the present study investigates the effects of a carefully calibrated IHT program, designed to avoid reduction in training load, by including high-intensity (VT2) and moderate-duration (24–40 min) hypoxic sessions, into the usual normoxic training of already trained athletes. Such an IHT model provides an original framework, in which the metabolic stimulus is enhanced through hypoxic sessions, without altering the mechanical component of the usual training load. Significant improvements of several indexes of aerobic performance capacity were observed not only at altitude but also at sea level, including V̇o2 max and Tlim. Additionally, IHT did not significantly modify V̇o2 kinetics such that Tlim lengthening was correlated neither with changes in the rate of V̇o2 adjustment nor with V̇o2 max and ventilatory thresholds. Collectively, these findings suggest that the enhanced endurance performance capacity obtained with IHT might be due to specific muscle adaptations to hypoxic training. This hypothesis is further explored in the two following companion papers of our study appearing in the present issue.


This project was supported by grants from the International Olympic Committee and the Ministère Français de la Jeunesse et des Sports. The scientific and sport coordination were, respectively, assumed by Prof. Jean-Paul Richalet and M. Laurent Schmitt, to whom we express our sincere gratitude.


The authors thank all of the athletes for enthusiastic participation; the whole laboratory staff from the Département de Physiologie et des Explorations Fonctionnelles and the Equipe d'Accueil 3072 for daily technical support; M. François Piquard for statistical advices; and Valérie Bougault and Frédéric Daussin for contribution during the training sessions. The help of M. Fabio Borrani was greatly appreciated for the application of the bootstrap method to the statistical treatment of the V̇o2 kinetics.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract