Of three intermittent hypoxia methods, only live high-train low raised VO2max in athletes beyond control, by 2.17 mL/kg/min (95% CI 0.77 to 3.56; 9 studies, none blinded); training in hypoxia (live low-train high, +0.89, −0.27 to 2.05) and passive hypoxia at rest (−1.07, −3.51 to 1.37) did not, though both helped non-athletes (+1.70 and +2.26). A meta-analysis of 62 controlled studies (1332 people), most at some concern or high risk of bias; no dose feature of live high-train low was linked to the gain. Read in full
Main caveat: The authors': most studies had methodological concerns that may bias the estimates; few applied blinding (none in LHTL), and placebo effects are possible.
Kohlbrenner, Ghaith and colleagues (Grenoble) asked which way of adding intermittent hypoxia to normal life raises VO2max in healthy people, and which features of the exposure go with a larger change.
They compared three methods against identical normoxic control groups: living (mostly sleeping) in hypoxia while training low (LHTL), training in hypoxia while living low (LLTH), and sitting in hypoxia at rest during waking hours (PHC). From 5,244 records they included 62 controlled studies with 1,332 participants and analysed athletes and non-athletes separately.
In athletes, LHTL raised VO2max by 2.17 mL/kg/min more than control; LLTH (+0.89) and PHC (−1.07) did not differ from control. In non-athletes, LLTH (+1.70) and PHC (+2.26) raised VO2max more than control, and LHTL rested on a single study.
Meta-regression found a lower inspired oxygen fraction associated with the LLTH change and hours per day and frequency of exposure with the PHC change, but no feature of LHTL dose (FiO2 14.5-17%, 11-14 h a day, 15-29 days) was associated with the change.
Most studies were at some concern or high risk of bias, chiefly because few were blinded, and the authors call the gains relatively small and ask for placebo-controlled trials. Note that the abstract speaks of standardised mean differences while the analyses report mean differences in mL/kg/min.
Participants
Pooled: 1,332 participants in 62 studies, 610 athletes and 722 non-athletes; three studies ran two methods and appear in two groups.
By method (Table 1, medians with quartiles): LHTL 194 participants, age 23 (21-24), 45% male, 91% athletes, baseline VO2max 63 (60-67) mL/kg/min; LLTH 884 participants, age 23 (21-26), 81% male, 45% athletes, VO2max 52 (45-59); PHC 254 participants, age 23 (22-25), 56% male, 33% athletes, VO2max 42 (39-56).
Athletes: classified by each study's description and/or a baseline VO2max above 60 mL/kg/min. LHTL athlete studies were in runners, cyclists, cross-country skiers and triathletes; LLTH athlete studies included cyclists, runners, swimmers, biathletes, triathletes, boxers, basketball and rugby players; PHC athlete studies were in runners and swimmers.
Non-athletes: sedentary, active or undefined healthy adults, one study in postmenopausal participants.
Protocol
Search: PubMed, EMBASE, Web of Science Core Collection from inception to the end of October 2025; English only; two reviewers screened, a third settled disagreements.
Inclusion: chronic (more than one session) intermittent (less than 24 h a day) hypoxia in healthy humans; an identical active control in normoxia; VO2max or VO2peak by gas exchange, relative to body mass, before and after. Excluded: patients, animals, continuous hypoxia (live high-train high).
Doses (Table 1, medians): LHTL FiO2 17.5% (Q1-Q3 15.0-19.0; the paper's discussion and Table 2 give 14.5-17% for the LHTL studies, see Limitations), 12.5 h a day, 19 sessions, 21 days, 0% of studies blinded; LLTH FiO2 18.6%, 1.0 h a day, 12 sessions, 28 days, 32% blinded; PHC FiO2 18.0%, 1.1 h a day, 20 sessions, 28 days, 25% blinded.
Analysis: mean differences between hypoxic and normoxic groups in mL/kg/min; three-level hierarchical random-effects model (study groups within studies), shared control groups duplicated where a study had several hypoxic arms; univariate and adjusted meta-regressions on FiO2, exposure duration, exercise type and intensity, age, sex, blinding, sample size, athlete status and baseline VO2max; R 4.5.1.
Risk of bias: RoB 2 by two reviewers; publication bias by funnel plot and Egger's test.
Funding and interests: no specific funding; the first author held a European Respiratory Society fellowship; no conflicts of interest reported.
Key Results
Change in VO2max, hypoxic minus normoxic control (mL/kg/min, 95% CI):
| Method | Athletes | Non-athletes |
|---|---|---|
| Live high-train low | +2.17 (0.77 to 3.56); 10 comparisons, 9 studies | +3.10 (−0.23 to 6.43), not significant; 1 study |
| Live low-train high (training in hypoxia) | +0.89 (−0.27 to 2.05), not significant; 21 comparisons, 21 studies | +1.70 (0.85 to 2.54); 26 comparisons, 24 studies |
| Passive hypoxic conditioning | −1.07 (−3.51 to 1.37), not significant; 4 comparisons, 4 studies | +2.26 (1.00 to 3.52); 8 comparisons, 7 studies |
Meta-regression (coefficients B, 95% CI):
| Method | Univariate | Multivariate |
|---|---|---|
| LHTL | no significant predictor | no significant predictor |
| LLTH | FiO2 0.99 (0.24 to 1.75) | FiO2 0.98 (0.20 to 1.77) |
| PHC | athlete status 3.40 (0.87 to 5.92); baseline VO2max −0.10 (−0.19 to −0.01); exercise type −4.66 (−7.47 to −1.85); sample size 0.23 (0.01 to 0.46) | hypoxic hours per day 2.46 (0.07 to 4.85); frequency of exposure −0.28 (−0.54 to −0.02) |
Risk of bias: overall some concerns or high risk in most studies; most were at low risk for randomisation, missing data and selective reporting, and at elevated risk for deviations from the intended intervention and outcome measurement.
Publication bias: no relevant funnel asymmetry; Egger p = 0.67.
Reading the coefficients (our reading): the text does not give the coding or units of the predictors.A positive coefficient on FiO2 would mean milder hypoxia went with a larger LLTH gain, while the authors read the result as favouring more severe hypoxia; for PHC the text says longer and more frequent exposure was more effective, but the printed frequency coefficient is negative.The abstract's "severity, duration and frequency" claim therefore rests on coefficients whose direction we could not verify, and it does not apply to LHTL, where nothing was associated.
Practical Application
For an athlete, of the three ways tested only a live high-train low block raised VO2max beyond control, and by about 2 mL/kg/min (2.17, 95% CI 0.77-3.56; median block 21 days at about 12.5 h a day in hypoxia). No LHTL study was blinded, and the authors note these studies are prone to placebo effects.
Training sessions in hypoxia and sitting in hypoxia at rest did not raise athletes' VO2max more than the same training in normal air (+0.89 and −1.07, both not significant); both helped non-athletes, whose starting VO2max was lower.This agrees with SYNTH-06 steering athletes away from passive hypoxia, and it disagrees with a network meta-analysis that found training in hypoxia improved aerobic capacity in trained athletes (REF-1675, which pooled VO2max with peak power and used standardised differences).
Dose within LHTL was not linked to the gain: across studies of 11-14 h a day, FiO2 14.5-17% and 15-29 days (the ranges the discussion gives), no feature predicted the VO2max change. SYNTH-06's dose lines (height, hours a day, block length) come from a narrative review; this analysis neither confirms nor refutes them.
Inference, not tested here: for an amateur below about 60 mL/kg/min, the non-athlete results suggest training or resting in hypoxia may add a little VO2max, but the classification was by VO2max or study label, not by training, and no performance outcome was measured.
Limitations
The authors': most studies had methodological concerns that may bias the estimates; few applied blinding (none in LHTL), and placebo effects are possible; methods were pooled across training types, with type handled by meta-regression; samples were mostly young and male.
Inclusion asked for controlled studies, and RoB 2 is a tool for randomised trials; the text does not say how many included studies were randomised.
Athlete status was assigned from study labels or a VO2max above 60 mL/kg/min, so the athlete and non-athlete groups overlap in training status.
LHTL in non-athletes rests on one study and PHC in athletes on four.
VO2max only: no race, time-trial or haemoglobin outcome. The abstract's "standardised mean differences" do not match the mean differences analysed, and the direction of two meta-regression coefficients does not match the authors' reading (Key Results); Table 1's median LHTL FiO2 of 17.5% (Q1-Q3 15.0-19.0) lies above the 14.5-17% range the discussion gives for the same studies, and above every FiO2 Table 2 prints for them.
We did not read the supplementary files (search strings, RoB 2 plots, forest plots, code).
Revision 2026-11-10: the two FiO2 figures for the live high-train low studies now name their sources (Table 1's median 17.5%, Q1-Q3 15.0-19.0; the discussion's range 14.5-17%), and the mismatch between them is listed in Limitations.
How this entry was checked and changed
Why this level
a meta-analysis of controlled studies whose risk of bias we read in the full text; capped at A- because that risk is serious (most studies at some concerns or high risk overall, mainly from deviations from the intended intervention and outcome measurement; no LHTL study blinded), with few studies in several cells (4 PHC studies and 1 LHTL study in their groups); VO2max is a capacity outcome, so no marker step