Participants

Sixteen trained male cyclists (age: 30 ± 5 years; body height: 180.5 ± 8.0 cm; body mass: 75.2 ± 9.0 kg; body fat content: 10.2 ± 2.0%; VO_2max: 57.0 ± 6.0 ml/kg/min) took part in the study. An a priori power analysis was conducted using G*Power 3.1 (Faul et al., 2007). Assuming a repeated-measures ANOVA, α = 0.05, power (1–β) = 0.80, and effect size f = 0.40, the required sample size was n = 12. Considering an anticipated dropout rate of approximately 30%, the planned recruitment target was set to n = 16. All participants were familiar with the laboratory testing procedures and had up-to-date medical examinations confirming no contraindications to exhaustive exercise in a hypoxic environment. Written informed consent was obtained from each participant prior to enrollment. All participants successfully completed every stage of the study. The inclusion criteria were as follows: at least 4 years of training experience, VO_2max > 50 ml/kg/min, a minimum 6-month break from the last high-altitude training camp, and age between 25 and 40 years. The exclusion criterion was premature termination of the exercise test or failure to complete any stage of the study. The research project was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee for Scientific Research at the Opole Medical School, Opole, Poland (approval code: KB/270/FI/2020; approval date: 17 July 2020).

Measures

The experiment consisted of three test stages conducted under different conditions: S1 (normoxia, N), S2 (normobaric hypoxia: 2000 m, H2000), and S3 (normobaric hypoxia: 3000 m, H3000) (Figure 1). The fraction of inspired oxygen (FiO₂) differed across the stages: S1: FiO₂ = 20.9%, S2: FiO₂ = 16.5%, and S3: FiO₂ = 14.5%. The order of test stages was randomized for each participant using an Excel random number generator. All tests were performed in a normobaric hypoxic chamber (AirZone, AirSport, Poland). Participants were not informed of the actual conditions inside the chamber (normoxia or hypoxia), and they were also not informed of their heart rate or the power output during the tests. Throughout all tests, atmospheric conditions were controlled and kept constant: temperature 19–20°C, relative humidity 45–50%, carbon dioxide concentration 700–800 ppm, and oxygen fraction according to the experimental condition. In each stage, participants first performed an incremental cardiopulmonary exercise test (CPET). After a 3-day recovery period, they completed a set of constant-intensity efforts to determine the MLSS. The interval between particular stages (S1, S2, and S3) was 10 days. The methodology was consistent across all stages and all testing sessions were conducted at the same time of the day to minimize circadian variability. Participants were instructed to refrain from intense physical activity between sessions, to rest adequately, and to maintain a mixed diet. Their last meal was to be consumed no later than 3 hours before testing.

Figure 1

Scheme of the research protocol.

Design and Procedures

In the first test session of each stage (S1, S2, or S3), body mass and body composition were assessed in the laboratory between 7:00 and 7:30 a.m., before breakfast. Body height was measured with an anthropometer to the nearest 0.5 cm, and body mass and body fat were estimated using bioelectrical impedance analysis (BC-418, Tanita, Japan).

At each stage, 3 h after consuming a light mixed meal, participants performed a graded exercise test to volitional exhaustion on a Cyclus2 cycle ergometer (RBM Elektronik-Automation GmbH, Germany). The protocol began at 80 W, with increments of 40 W every 3 min until exhaustion. The heart rate (HR), pulmonary ventilation (V̇E), breathing frequency (BF), oxygen uptake (VO₂), carbon dioxide output (VCO₂), and the respiratory exchange ratio (RER) were continuously measured at rest (3 min before the test) and throughout exercise using a fast-response gas analyzer (MetaLyzer 3B, Cortex, Germany) with the breath-by-breath method. Before each test, the device was calibrated according to the manufacturer’s recommendations for the flow and gas sensors, using ambient air (O₂ = 20.9%) and a reference gas, followed by control measurements of room conditions to verify accuracy under the experimental settings. The MetaSoft® Studio (Cortex, Germany) software was used to analyze the results and determine the measured variables. At the end of each workload stage (last 15 s), fingertip capillary blood samples were collected to determine blood lactate (La) concentration (SUPER GL2, Dr. Müller Gerätebau GmbH, Germany). These data were used to analyze blood lactate kinetics and to determine the individual lactate threshold using the D_max method (LTD-max, Cheng et al., 1992). Previous studies (Czuba et al., 2009; Płoszczyca et al., 2020) demonstrated that LT determined by the D_max method corresponded to the MLSS.

After a 72-h rest period, participants performed a series of constant-load efforts on the Cyclus2 cycle ergometer to determine the MLSS under the conditions of each testing stage (S1: normoxia, S2: hypoxia at 2000 m, S3: hypoxia at 3000 m). Each test began with a 10-min warm-up at an individually determined intensity corresponding to 70% of the lactate threshold obtained with the D-max method (LTD-max) for the given environmental condition, performed at a cadence of 80–90 revolutions per minute. Following the warm-up, participants completed a 30-min constant-load effort at 100% LTD-max. During the constant-load tests, the HR, V̇E, BF, VO₂, VCO₂, and the RER were measured continuously using a fast-response gas analyzer (MetaLyzer 3B, Cortex, Germany), while hemoglobin saturation (SpO₂) was assessed with a 3150 WristOx2 oximeter (Nonin Medical Inc., USA). SpO₂ was measured during the blood-sampling intervals, with a one-minute finger recording at each time point—after drying the skin and stabilizing the wrist to minimize motion artifacts—corresponding to capillary blood samples taken at rest, after the warm-up, and every 5 min during the test. When La concentration during the final 20 min of exercise did not increase by more than 1 mmol/L, the test was repeated after a 3-day rest with the load increased by 20 W. Tests were continued until the final 20 min of the workload caused ΔLa to exceed 1 mmol/L; in such cases, the previous workload was considered the MLSS for that condition. The MLSS workload was typically determined after 2 (up to 3) tests performed under each condition, which was made possible by the previously conducted incremental test used to determine the workload corresponding to LTD-max, from which the MLSS test was initiated (Czuba et al., 2009, Płoszczyca et al., 2020).

Statistical Analysis

The primary endpoint of the study was the WRMLSS. At the time of study planning, no published data were available describing the expected magnitude of change in the WRMLSS under hypoxic conditions. Therefore, the a priori power analysis was informed by previous literature reporting hypoxia-related changes in aerobic performance variables in athletes (e.g., power output, VO_2max) and by our pilot observations (Mollard et al., 2007; Płoszczyca et al., 2021). Based on these sources, we assumed a large effect size (Cohen’s f = 0.40) for the effect of hypoxia on the WRMLSS.

Data were collected and processed using Statistica software (StatSoft, Tulsa, OK, USA). The Shapiro-Wilk test was first used to test the normal distribution of variables and the Levene’s test was used to check the homogeneity of variance. When the assumptions for parametric testing were met, a one-way repeated-measures ANOVA was performed. Before conducting the repeated-measures ANOVA, Mauchly’s tests were performed to assess the assumption of sphericity. All variables met the sphericity assumption (p > 0.05), allowing the use of standard repeated-measures ANOVA without applying any corrections. Effect sizes (partial η²) were calculated for all significant ANOVA results, and 95% confidence intervals (CIs) for the partial η² were reported. When significant differences were found by ANOVA, the Bonferroni post hoc test was used. Statistical significance was set at p < 0.05 for all comparisons.

Effects of Hypoxia on the Workload (Power Output) and Blood Lactate Concentration at the MLSS

In the present study, both H2000 and H3000 significantly reduced the WRMLSS by 9.3% and 18.5%, respectively. These findings are consistent with results reported by Beever et al. (2024) who observed an approximately 9% reduction in the WRMLSS already at an altitude of 2222 m a.s.l., which is consistent with the outcomes of our research. However, it is important to highlight certain methodological differences between the studies, particularly the combination of hypobaric and normobaric hypoxia used by Beever et al. (2024). Furthermore, that study did not specify whether the unusually high peak RER values (1.27) were recorded during exercise or recovery, which complicates the assessment of their accuracy.

Among the primary causes of the reduced WRMLSS under hypoxic conditions is the decrease in SpO₂. In hypoxia, when oxygen availability is limited, reductions in PaO₂ and SpO₂ occur, leading to impaired aerobic capacity and exercise performance (Mollard et al., 2007; Płoszczyca et al., 2021; Wehrlin and Hallen, 2006). The decline in aerobic capacity becomes more pronounced with greater hypoxic severity, although the extent of this decline shows considerable interindividual variability (Chapman, 2013). It has been suggested that reductions in SpO₂ trigger a central nervous system (CNS) response which limits exercise intensity to maintain adequate arterial oxygen saturation and protects against myocardial ischemia and dysfunction of vital organs (Noakes, 2001). In the present study, SpO₂ during MLSS testing decreased significantly to 84.8% at FiO₂ = 16.5% (H2000) and to 79.2% at FiO₂ = 14.5% (H3000). According to Amann and Calbet (2008), hypoxemia and hypocapnia may reduce the cerebral blood flow when SpO₂ falls below 82%. In our study, under H2000 and H3000 conditions, the reduction in SpO₂ was accompanied by a decrease in the WRMLSS. Our results indicate, however, that this impairment occurred already at SpO₂ of approximately 84%, suggesting that a CNS-mediated response may be triggered at higher saturation levels than previously proposed, thereby limiting the athletes’ capacity to generate power and contributing to faster fatigue accumulation. These findings underscore the importance of monitoring SpO₂ during hypoxic training, as even moderate desaturation may compromise power output at intensities corresponding to the MLSS

Another potential mechanism underlying the reduction of the WRMLSS under hypoxic conditions is the acceleration of anaerobic glycolysis caused by reduced O₂ availability. This process leads to an increase in La concentration, as well as greater accumulation of H⁺ ions and inorganic phosphate compared with normoxia. Consequently, excitation-contraction coupling within muscle fibers is impaired, which has been identified as a factor contributing to loss of tension during high-intensity exercise (Amann and Calbet, 2008). It is also noteworthy that during exercise under hypoxic conditions, catecholamines (particularly adrenaline and noradrenaline) further stimulate anaerobic glycolysis, resulting in elevated La production. Thus, sympathetic nervous system activation plays a key role in the increased La and H⁺ concentrations (Piotrowicz et al., 2023).

In the present study, a significant increase in blood La concentration was observed at H2000 (+36.52% compared with N) and H3000 conditions (+34.49% compared with N), accompanied by a significant reduction in the WRMLSS. In a recent study, Park et al. (2022) reported that La concentration increased by ~36% during submaximal exercise (60% VO_2max) at 3000 m compared with normoxia. The authors attributed this response to reductions in SpO₂, sympathetic nervous system activation, and increased adrenaline release, which together enhanced glycolytic activity. Similar conclusions were presented by Sumi et al. (2018) who observed a significantly greater rise in La concentration under hypoxia (FiO₂ = 14.5%) than in normoxia during interval endurance exercise (10 × 3 min at 95% V̇O_2max) followed by 30 min of continuous running at 85% V̇O_2max.

Effects of Hypoxia on Cardiorespiratory Variables

In the present study, hypoxia at H2000 and H3000 resulted in a significant reduction in both relative and absolute VO₂MLSS values (7.2% and 17%) compared with normoxia. We attribute this primarily to the lower power output generated at the MLSS. These changes are mainly associated with the markedly reduced O₂ availability during exercise (as reflected by the significant decline in SpO₂), which is consistent with observations reported by other authors (Lawler et al., 1988), as well as with peripheral muscle fatigue. Increased acidosis, a high rate of phosphocreatine hydrolysis, and the accumulation of inorganic phosphate stimulate afferent feedback from muscle fibers, leading the nervous system to reduce the recruitment of motor units and, consequently, the number of active muscle fibers (Amann and Dempsey, 2016).

In a previous study, Beever et al. (2024) reported a significant reduction in VO₂MLSS of approximately 12% at an altitude of 2222 m a.s.l. In our study, the percentage change in VO₂MLSS was smaller compared with the findings of Beever et al. (2024), amounting to ~7% at 2000 m, whereas at 3000 m VO₂MLSS decreased by 17%, which closely corresponds to the ~18.5% decline in the WRMLSS. It is worth emphasizing again that the study by Beever et al. (2024) was conducted at a natural altitude of 1111 m a.s.l. (hypobaric hypoxia), combined with a hypoxic breathing mixture, which in our view had a substantial impact on the results. The differences between physiological responses to normobaric and hypobaric hypoxia continue to be debated in the literature. Their impact on the body remains inconclusive. Initially, it was assumed that the only difference between the two conditions was the partial pressure of oxygen (PO₂), responsible for inducing equivalent physiological changes. However, Coppel et al. (2015) demonstrated that differences exist, for example, in cardiorespiratory variables, which may influence outcomes. Similarly, Vinetti et al. (2024) found that hypobaric hypoxia was associated with higher ventilation, a lower respiratory exchange ratio, and higher VO_2max compared with normobaric hypoxia—results that are consistent with those of Angeli et al. (2019).

In our study, no significant changes in V̇E were observed during the MLSS under hypoxic conditions, despite the lower mechanical work performed in hypoxia. V̇EMLSS remained at a similar level to normoxia, even though WRMLSS was significantly reduced. Bouissou et al. (1987), in a study conducted at 3000 m a.s.l. at 60% VO_2max, reported lower VO₂ without changes in V̇E or the HR, suggesting an increased ventilatory cost under hypoxic conditions. In general, hypoxia induces an increase in V̇E as an adaptive response to oxygen deficiency in the tissues (Czuba et al., 2018). Similarly, as with V̇EMLSS, hypoxia did not affect the HRMLSS. Despite significant reductions in the WRMLSS, VO₂MLSS, and SpO₂ at H3000, the HRMLSS remained unchanged. These findings are consistent with previous studies assessing the effects of hypoxia on the cardiorespiratory system (Beever et al., 2024). For example, Burtscher et al. (2006), in two hypobaric hypoxia exercise tests lasting 5 and 50 min at 3200 m a.s.l., found no significant changes in the HR, although the WR decreased significantly by ~11% in both tests.

Practical Recommendations and Training Implications

When setting out to achieve specific training goals, it is crucial to exercise at a precise intensity level. At the MLSS or below, La production and clearance remain balanced, while higher-intensity exercise causes La accumulation due to production exceeding removal, as indicated in the introduction. Training at MLSS intensity induces meaningful physiological adaptations, as demonstrated by Ferreira et al. (2007) who observed resting bradycardia (21%) and a 28% improvement in running performance in mice following an eight-week MLSS program. In long-distance runners, MLSS-based training leads to small increases in MLSS speed and VO_2max, yet importantly extends time to exhaustion by approximately 50%. This type of training enhances endurance by improving lactate clearance capacity, while lactate concentration at the MLSS itself may remain unchanged (Billat et al., 2003; Czuba et al., 2009).

Unplanned exceeding of MLSS intensity may lead to different physiological adaptations to exercise and, in prolonged effort, to profound fatigue or overtraining (Sitko et al., 2025). The present study demonstrated that under hypoxic conditions the MLSS occurs at lower exercise intensities than in normoxia and depends on changes in the altitude/level of hypoxia, which is crucial for effective training. Maintaining the pace or power (WR) at the MLSS under hypoxic conditions without referencing normoxic values will result in rapidly increasing fatigue already at an altitude of 2000 m. Moreover, understanding how the MLSS shifts with the altitude is crucial for determining an optimal warm-up, pacing, and intensity strategy during competition, especially in cycling, where precise control of exercise intensity based on power output is essential. Precisely adjusting warm-up and race intensities to MLSS shifts relative to the altitude at which the athlete is competing helps prevent premature increases in anaerobic glycolysis, reduces the risk of early fatigue, and supports achieving the best possible performance under hypoxic conditions.

Training under hypoxic conditions with the accurately determined MLSS may enhance sea-level performance. This notion is supported by our previous findings (Czuba et al., 2011, 2018) which demonstrated that cyclists’ training in hypoxia at intensities close to the lactate threshold determined using the D_max method—physiologically corresponding to the MLSS domain—achieved improvements in aerobic capacity and endurance performance. Therefore, although direct evidence on MLSS-specific hypoxic training remains limited, the available findings suggest that maintaining MLSS-related workloads in hypoxia may be an effective strategy for enhancing sea-level performance.

Based on our findings and available literature, we recommend adjusting training intensity during hypoxic exercise by reducing the WRMLSS relative to normoxia by ~10% at 2000 m a.s.l. and by ~20% at 3000 m a.s.l. These adjustments are consistent with the observed reduction in VO_2max and allow for comparable metabolic loads to those at the sea level, thereby minimizing the risk of excessive physiological stress. Excessive exercise intensity exceeding the WRMLSS (e.g., WRMLSS + 20W) frequently resulted in test termination by participants, confirming the need for precise adjustment of training loads. Similar observations regarding reduced exercise tolerance in hypoxia have been reported by Beever et al. (2024). In situations where direct monitoring of power output is not possible, the HR is recommended as an alternative indicator of training intensity. The absence of significant changes in the HRMLSS under moderate hypoxia (2000–3000 m a.s.l.) suggests that HR zones determined under normoxic conditions can also be applied during training in hypoxic environments.

Limitations of the Study

One of the main limitations of this study is that MLSS changes were analyzed exclusively under normobaric hypoxia. In the literature, there is an ongoing debate about potential differences in the physiological responses of the body to normobaric versus hypobaric hypoxia, which may affect the interpretation of the present findings (Coppel et al., 2015). Another limitation was the inability to provide constant accommodation for participants during the experiment, which restricted full control over their diet, recovery, and the actual training loads. Additionally, assessing intra-individual variability could provide further insight into the consistency of physiological responses across participants, thereby strengthening the interpretation of the results. Furthermore, our study included just one sports discipline and only male participants, which limits the generalizability of the results. Future research should therefore be extended to women and to other endurance disciplines, such as long-distance running, cross-country skiing, and rowing, while accounting for their specific exercise characteristics. It would also be valuable to conduct studies in female athletes and across different training levels, which would enable a more precise characterization of adaptive responses in diverse populations.