REF-1276 · Milanović et al., 2015 · Sports Medicine · Paper · PubMed

How much will my VO2max rise if I start training?

Untrained young adults gained ~5 mL/kg/min over controls in 28 trials (HIT 5.5, continuous 4.9), more if less fit; publication bias may inflate this. Trained athletes gain less (A-).

Main caveat: Tables and figures not read. The text was read in full, but Tables 1-4 and Figures 1-5 are separate repository files.
fairly strong evidenceMeasured on non-athletes and recreational people

Milanović, Sporiš and Weston meta-analysed the effects of continuous endurance training and HIT on the VO2max of healthy, young to middle-aged adults, against no-exercise controls and against each other.

They searched six databases to 28 February 2014 for randomised and non-randomised controlled trials in English published after 1995, in healthy, untrained, sedentary, recreational and non-athletic men and women aged 18-45 (overweight and obese participants excluded), with at least 2 weeks of training and VO2max measured before and after.

Endurance training was classed as moderate intensity (60-85% HRmax) — most such studies ran three 40-60-min sessions a week — and HIT as "all-out", "supramaximal", "maximal" or "high (90-95% HRmax)"; programmes combined with strength training were excluded.

Of 84 full texts, 28 studies with 723 participants (25.1 ± 5 y; initial VO2max 40.8 ± 7.9 mL/kg/min) were pooled in random-effects models and judged by magnitude-based inference.

Against no-exercise controls, endurance training raised VO2max by 4.9 mL/kg/min (95% confidence limits ±1.4; "possibly large") and HIT by 5.5 (±1.2; "likely large"); compared directly, HIT gave a possibly small extra 1.2 (±0.9).

Gains were larger at lower baseline fitness and over longer interventions in both forms of training, and larger in typically younger participants after endurance training. The study-to-study SD was 1.3, 1.3 and 0.8 mL/kg/min, so the authors put typical study means at 3.6-6.2 (endurance), 4.2-6.8 (HIT) and −0.4 to 2.0 (HIT vs endurance).

Egger's test indicated publication bias for all analyses. No funding was used; the authors declare no directly relevant conflicts.

Our reading: about 5 mL/kg/min — roughly 12% from a start near 41 (our arithmetic) — is the mean to expect over a programme for an untrained young adult, whichever form is chosen, and publication bias means the true means are probably somewhat lower; HIT's edge over continuous training is about a fifth of the gain itself.

Protocol

Search: MEDLINE, PubMed, SPORTDiscus, Web of Science, CINAHL and Google Scholar, all records to 28 February 2014; search, screening and quality assessment by two authors (ZM, GS) with a third reviewer for disputes (the Methods also say one author extracted the data and the other checked it); reported to PRISMA.
Study type: randomised and non-randomised controlled trials in English, published after 1995; uncontrolled and cross-sectional studies excluded.
Participants: healthy, untrained, sedentary, recreational and non-athletic men and women aged 18-45 with no acute or chronic disease; no exclusion on baseline fitness; overweight and obese participants excluded.
Interventions: at least 2 weeks; endurance training classed as moderate intensity (60-85% HRmax); HIT classed as "all-out", "supramaximal", "maximal" or "high (90-95% HRmax)"; a no-exercise control group; programmes combined with strength training excluded; nutritional interventions only if applied to all participants.
Selection: 804 records plus 17 from reference lists → 84 full texts → 56 excluded → 28 studies; n = 723; age 25.1 ± 5 y; initial VO2max 40.8 ± 7.9 mL/kg/min.
Analysis: random-effects meta-analysis (Comprehensive Meta-Analysis v2), separately for endurance training vs no exercise, HIT vs no exercise and HIT vs endurance training; 95% confidence limits and probabilistic clinical inferences (unclear if the odds ratio of benefit to harm was below 66); magnitude thresholds 0.2, 0.6 and 1.2 between-subject SDs of baseline VO2max, i.e. 0.8, 2.4 and 4.7 (endurance vs no exercise), 0.8, 2.3 and 4.7 (HIT vs no exercise) and 0.9, 2.6 and 5.3 mL/kg/min (HIT vs endurance); study-to-study variation as an SD; publication bias by Egger's test (P < 0.10).Meta-regression on five moderators (age, baseline fitness, intervention duration, work:rest ratio, HIT repetition duration), each expressed as the effect of two SDs of the moderator, i.e. a typically low against a typically high value.
Not read: Tables 1-4 (study characteristics and per-study effects) and Figures 1-5 are separate repository files; the sex split, modes, session formats and programme lengths sit there.
Funding and interests: "No sources of funding were used to assist in the preparation of this review"; the authors have no conflicts of interest directly relevant to its content.

Key Results

VO2max change in mL/kg/min, ± 95% confidence limits, with the authors' magnitude labels (abstract and Results):

ComparisonMain effectStudy-to-study SD; typical study meansModerators (additional effect of two SDs of the moderator)
Endurance training vs no-exercise control+4.9 (±1.4), possibly large1.3; 3.6 to 6.2typically younger participants +2.4 (±2.1) and longer interventions +2.2 (±3.0), possibly moderate; lower baseline fitness +1.4 (±2.0), small
HIT vs no-exercise control+5.5 (±1.2), likely large1.3; 4.2 to 6.8lower baseline fitness +3.2 (±1.9) and longer interventions +3.0 (±1.9), likely moderate; typically longer HIT repetitions −1.8 (±2.7), small lesser effect ("likely" in the Results); age +0.8 (±2.1) and work:rest ratio +0.5 (±1.6), unclear
HIT vs endurance training+1.2 (±0.9), possibly small, favouring HIT0.8; printed as −0.4 to 2.0small additional improvements with typically longer HIT repetitions +2.2 (±2.1), older participants +1.8 (±1.7), longer interventions +1.7 (±1.7), greater work:rest ratio +1.6 (±1.5) and lower baseline fitness +0.8 (±1.3)
Publication bias: Egger's test indicated funnel-plot asymmetry for all analyses (P < 0.10).
A number that does not add up: the first two typical ranges are the mean ± one study-to-study SD (4.9 ± 1.3 = 3.6 to 6.2; 5.5 ± 1.3 = 4.2 to 6.8), but 1.2 ± 0.8 gives 0.4 to 2.0, not the printed −0.4 to 2.0 (our arithmetic) — most likely a sign slip.
Authors' conclusion: endurance training and HIT both elicit large VO2max improvements in healthy young to middle-aged adults, with larger effects in the less fit and greater gains after HIT than after endurance training.

Mechanisms

Measured: VO2max only; no performance outcome.
Authors' interpretation: aerobic training has an adaptive effect that favours the less fit; HIT's added benefit over endurance training was larger in older and less fit subjects, which they read as an appeal of HIT for older adults and patients. Physiological mechanisms are outside the review's scope.
Inference: the endurance arms (60-85% HRmax) span the moderate and the heavy domain. In a six-group RCT moderate-domain continuous cycling did not raise VO2max beyond control while heavy-domain continuous work did (REF-036), so +4.9 is an average over continuous programmes of mixed intensity, not the yield of truly easy riding.
Inference: publication bias in every analysis usually means small trials with small or null effects are missing, so the pooled means are more likely too high than too low; the size of any overstatement is not reported.
Inference: the moderator effects compare a typically low with a typically high value of the moderator (two SDs apart), so "+3.2 for lower baseline fitness" is not a gain per mL/kg/min, and "longer interventions" cannot be turned into a rate per week without Table 1.
Inference: the longer-HIT-repetition moderator points two ways — smaller against controls (−1.8, ±2.7; this interval includes zero) but larger against endurance training (+2.2, ±2.1; 0.1 to 4.3, our arithmetic) — and the Discussion still speaks of "enhanced beneficial effects for HIT with longer repetitions, greater work:rest ratios", although the work:rest effect against controls was unclear.

Practical Application

How much will my VO2max rise if I start training? Untrained young adults gained ~5 mL/kg/min over controls in 28 trials (HIT 5.5, continuous 4.9), more if less fit; publication bias may inflate this. Trained athletes gain less (A-).
HIT or steady training for a beginner's VO2max: both work; HIT added 1.2 (±0.9) mL/kg/min over continuous training, a "possibly small" difference. Not from this paper; the base's own practice: let time available and preference decide.The continuous programmes behind the +4.9 were mostly three 40-60-min sessions a week at 60-85% HRmax (full text); for interval dose in inactive adults, HIIT reached 80% of its VO2max gain at about 11 min a session and flattened by about 23 (REF-749, A).
Quote the mean as a range, not a promise. Typical study means ran from about 3.6 to 6.8 mL/kg/min across both forms of training (the authors' ranges), individual results scatter further around them, mostly from measurement noise (REF-1274), and publication bias argues for the lower half; give the date of the retest that will replace the group number (inference).
For trained athletes the base has no equivalent gain over no exercise. It holds an added effect — 2.5 mL/kg/min more VO2peak when work above the maximal metabolic steady state is added to training kept below it (REF-989, A-) — and a before-after time-trial gain of about 2% (REF-1275, A-).
Session shape: not from this paper; the base's own practice for VO2max intervals is SYNTH-33.

Limitations

Tables and figures not read. The text was read in full, but Tables 1-4 and Figures 1-5 are separate repository files; per-study characteristics (modes, programme lengths, sex split) and the funnel plots were not checked. The text read is the author manuscript, not the typeset article; its abstract numbers match the PubMed abstract.
Publication bias in all three analyses (Egger's test, P < 0.10); no adjusted estimate is given.
Design mix and size: randomised and non-randomised controlled trials, English only, published after 1995; 723 participants across 28 trials (about 26 per trial, our arithmetic).
Magnitude-based inference. Effects are labelled "possibly", "likely" or "unclear" by a method whose error rates have been disputed in the statistical literature (a general methodological note, not from this paper); the 95% confidence limits are reported and are used here.
Wide moderator limits. Several moderators have limits as wide as or wider than their estimates (for example longer interventions +2.2, ±3.0 for endurance training), and the HIT repetition-length moderator changes sign between comparisons.
Reporting slips. The HIT-vs-endurance typical range is printed as −0.4 to 2.0 where the mean and SD give 0.4 to 2.0 (our arithmetic). The Discussion describes an increased work:rest ratio as "greater recovery in-between HIT repetitions", the opposite of what a higher work:rest ratio means (our reading), and cites reference 21 both for an unclear HIT-vs-endurance effect and for a significant 3.03 mL/kg/min difference.
Who. Healthy, untrained or recreational, non-athletic, normal-weight adults aged 18-45, mean 25, starting at 40.8 mL/kg/min.
VO2max only, with no performance outcome.
Funding and interests: no sources of funding; the authors declare no conflicts of interest directly relevant to the review. No device maker, federation or data platform involved.
How this entry was checked and changed
Why this level

a meta-analysis of controlled trials, held below A because the trials are small (723 participants in 28 trials, about 26 per trial on average — our arithmetic), non-randomised trials were eligible, Egger's test indicated publication bias for all three analyses, and effects are judged by magnitude-based inference rather than conventional tests. Read from the full text: the author manuscript in the Teesside University repository, listed by Europe PMC and OpenAlex (closed access at the publisher); Tables 1-4 and Figures 1-5 are separate files and were not read

Full text (checked 2026-10-28)
Participants were healthy, untrained, sedentary, recreational and non-athletic men and women aged between 18 and 45 y; studies with overweight and obese participants were excluded (full text)
Endurance training intensity was classified as moderate-intensity (60-85% HRmax), HIT as "all-out", "supramaximal", "maximal" or "high (90-95% HRmax)"; studies combining training with strength training were excluded (full text)
Searches covered all available records up to 28 February 2014; 804 potential manuscripts plus 17 from reference lists, 84 studies after screening, 56 excluded, 28 included (full text)
A random effects meta-analysis; magnitude thresholds 0.8, 2.4 and 4.7 (endurance vs no exercise), 0.8, 2.3 and 4.7 (HIT vs no exercise) and 0.9, 2.6 and 5.3 mL•kg-1•min-1 (HIT vs endurance training); unclear if the odds ratio of benefit to harm was <66 (full text)
Endurance training vs controls 4.9 mL•kg-1•min-1 (95% confidence limits ±1.4), study-to-study SD 1.3; HIT vs controls 5.5 (±1.2), SD 1.3; HIT vs endurance training 1.2 (±0.9), SD 0.8 (full text)
The mean effect lies typically between 3.6 and 6.2 mL•kg-1•min-1 for endurance training, between 4.2 and 6.8 for HIT, and between -0.4 and 2.0 for HIT compared to endurance training (full text)
The Egger's test results indicated publication bias for all analyses (P < 0.10) (full text)
Most of the endurance studies undertook three moderate-intensity sessions per week lasting 40 to 60 minutes (full text)
No sources of funding were used to assist in the preparation of this review; the authors have no conflicts of interest directly relevant to its content (full text)