Is my critical power the same as my lactate steady state?
Close on average — CP ran 12 W higher in 10 small lab studies — but for one rider the 95% limits ran from about 19 W below to 44 W above. Treat them as different numbers.
Main caveat: Small, homogeneous evidence. 10 studies of 6 to 13 young participants; 19 of 111 were women; no masters, elite or clinical samples.
Micheli and colleagues (Urbino, Verona, Western Ontario, Toronto, Groningen) asked how far apart critical power (CP) and the maximal lactate steady state (MLSS) sit, how well they agree for one person, and whether the answer depends on the variable used to express intensity.
They searched PubMed, Web of Science and Scopus to March 2023 and kept only within-subject studies in which participants rode (or ran) a constant-intensity trial at CP and at MLSS, so that oxygen uptake, heart rate and lactate at each boundary were measured rather than extrapolated from a ramp. Ten of 782 records qualified (11 samples, 111 participants, 19 women).
In power output, CP was higher than MLSS by 12.42 W on average (95% CI 4.69 to 20.16; prediction interval −11.62 to 36.47 W), with Bland-Altman limits of agreement of −19.23 to +44.08 W.
Expressed as oxygen uptake (0.09 L/min), heart rate (0.61 bpm) or blood lactate (1.63 mM), the differences were not significant, but those analyses held five, three and three studies, and three of the four studies with the largest power gaps reported no internal-load data.
In meta-regression the power gap grew with the mean CP (slope 0.21 W per W) and the mean MLSS (0.243 W per W) of the study samples. The authors conclude that CP exceeds MLSS in power output, more so in fitter riders, and that the wide individual spread means the two should not be used interchangeably; they attribute much of the gap to how each is measured.
Our reading: this is the first pooled estimate the base holds for the CP-MLSS gap, and it fits the base's two existing positions — MLSS sits somewhat below CP (REF-168) and threshold markers agree on the mean but not for the individual (REF-732).
Protocol
Design: systematic review and meta-analyses, PRISMA, PROSPERO CRD42021261155. Searches of PubMed, Web of Science and Scopus with weekly alerts to March 2023; two reviewers screened independently, the senior author resolved disagreements. English and Portuguese papers.
Eligibility (PICOS): apparently healthy adults aged 18-65; constant-intensity exercise at CP and at MLSS in the same participants; outcomes power output, oxygen uptake, blood lactate or heart rate; crossover or within-subject design. Studies that only predicted the intensity at CP or MLSS from an incremental test were excluded.
Included: 10 studies, 11 samples (one study reported women and men separately), n = 111 participants (19 women); sample sizes 6 to 13; mean ages 21.5 to 28 years. Six researchers supplied additional and unpublished data; Table 2 marks four studies (Ozkaya 2022, Okuno 2011, Puga 2009, Caen 2022) as using author-provided data.
CP determination: 2-parameter hyperbolic (3 studies), 3-parameter hyperbolic (2), "best fit" model (4), not stated (1); exhaustive trials lasting from 1 to 20 min across studies. MLSS determination: lactate steady state between minutes 10 and 30 in nine studies, 15 and 30 in one.
Analysis: random-effects meta-analysis of paired mean differences (CP minus MLSS) separately for power output, oxygen uptake, heart rate and blood lactate, with 95% confidence and prediction intervals; missing SDs of the difference imputed from the median correlation of the studies with the same unit; meta-regression of the power-output difference on mean CP and mean MLSS power (moderators with at least 10 effect sizes only); Bland-Altman meta-analysis of limits of agreement with outer 95% CIs (Tipton and Shuster method); standardised mean differences for the four samples reporting both power and oxygen uptake; sensitivity analysis on those samples.R (metafor, dmetar, clubSandwich).
Risk of bias: a revised Cochrane tool for crossover trials; randomised trial order high or unclear in all 10 studies, allocation concealment high in all, carry-over unclear in all.
Funding and interests: funded by the European Union - NextGenerationEU under the Italian Ministry of University and Research (MUR) National Innovation Ecosystem grant ECS00000041 - VITALITY - CUP H33C22000430006; the funders had no role in design, analysis or publication. Competing interests: Carlo Ferri Marini is an Academic Editor for PeerJ.Per the paper's reference list, review author D.A. Keir is a co-author of 4 of the 10 included studies (Keir 2015, Maturana 2016, Iannetta 2019, Iannetta 2022) and S. Pogliaghi of one of them (Keir 2015); the competing-interests statement names only the editorial role, and screening and inclusion were done by L. Micheli and C. Ferri Marini.
Key Results
| Intensity expressed as | Studies | Mean difference, CP minus MLSS (95% CI) | p | 95% limits of agreement | Heterogeneity |
|---|---|---|---|---|---|
| Power output | 10 samples | +12.42 W (4.69 to 20.16); prediction interval −11.62 to 36.47 W | 0.005 | −19.23 to +44.08 W (outer CI −30.91 to 55.76) | Q(9) = 96.490, τ = 10.065 W, p < 0.001 |
| Oxygen uptake | 5 studies (6 samples) | +0.09 L/min (−0.02 to 0.21) | 0.097 | −0.34 to +0.52 L/min | Q(5) = 15.238, τ = 0.090, p = 0.009 |
| Heart rate | 3 | +0.61 bpm (−7.70 to 8.91) | 0.784 | −15.84 to +17.05 bpm | Q(2) = 3.767, p = 0.152 |
| Blood lactate | 3 | +1.63 mM (−2.61 to 5.86) | 0.240 | −2.85 to +6.11 mM | Q(2) = 24.352, τ = 1.612, p < 0.001 |
Meta-regression (power output): the gap rose with mean CP (slope 0.21 W per W, 95% CI 0.107 to 0.314, intercept −39.378; F(1,8) = 21.091, p = 0.002) and with mean MLSS (slope 0.243, 95% CI 0.078 to 0.408, intercept −44.634; F(1,8) = 11.114, p = 0.010). Our arithmetic from the CP model: about 3 W at a mean CP of 200 W and about 24 W at 300 W — a between-study relationship across 10 samples, not a prediction for one rider.
Study-level gaps in power: from 1.0 ± 16.6 W (Iannetta 2019) and 2 ± 12 W (Keir 2015) to 26.5 ± 7.4 W (Caritá 2009) and 30.0 ± 10.9 W (Ozkaya 2022, well-trained cyclists); within-study correlations between CP and MLSS were 0.88-0.99.
Sensitivity (the four studies, five samples, reporting both power and oxygen uptake): 10.59 W (95% CI −4.34 to 25.51) and 0.11 L/min (−0.04 to 0.25), similar to the full analyses. Standardised mean differences 0.648 (power, p = 0.098) and 0.482 (oxygen uptake, p = 0.064).
Our arithmetic: the mean CP across the 10 power samples is about 249 W, so the mean gap is about 5% and the limits of agreement run from about 8% below to about 18% above MLSS.
Mechanisms
Measured: CP exceeded MLSS in power output on average, with no significant difference in oxygen uptake, heart rate or lactate at the two intensities in the subset of studies that measured them; the power gap was larger in samples with a higher CP or MLSS.
Authors' interpretation: the gap is mostly methodological. CP depends on the number and duration of exhaustive trials and the model fitted; MLSS on the time window used to judge a lactate steady state and the power steps between trials.Fitter, more experienced riders may reach a truer exhaustion in the CP trials while the submaximal MLSS test is unaffected, widening the gap.Citing others' work, they write that stricter protocols (three to five exhaustive trials with CP standard error under 5%, the shortest 2-3 min and the longest over 10 but not over 15 min; a narrower MLSS window such as minutes 15 or 20 to 30) can reconcile the two.
Authors' interpretation: the lack of a difference in oxygen uptake, heart rate and lactate may partly reflect which studies reported internal load: three of the four studies with the largest power gaps did not.
Inference: the moderator is the sample's mean power, which mixes fitness, testing experience and the lab's protocol; the paper cannot separate them.
Practical Application
Is my critical power the same as my lactate steady state? Close on average — CP ran 12 W higher in 10 small lab studies — but for one rider the 95% limits ran from about 19 W below to 44 W above. Treat them as different numbers.
Pick one anchor and keep it (A-, 111 mostly young men on cycle ergometers, lab protocols): a CP and a lab MLSS for the same rider may sit a few percent apart or close to a training zone apart; switching from one to the other mid-season can look like a change in fitness that is not there. This agrees with REF-732 (CP vs lactate markers, ±22 to ±52 W) and the base's rule in SYNTH-53.
Long steady "threshold" reps a little below CP (inference from the mean gap; the base's practice in [SYNTH-27]): the pooled gap, about 5% of CP, is in line with prescribing sub-threshold reps a few percent below CP when a steady state is intended (REF-750). The meta-analysis tested agreement, not training.
Fitter riders, bigger gap (between-study meta-regression): a well-trained rider with a high CP should expect CP to sit further above a lab MLSS than a recreational one does; this is a group pattern across 10 samples, not a correction factor.
Heart rate or lactate at CP (three studies each; weak): mean heart rate at CP and at MLSS differed by under 1 bpm, but for one person by up to about ±16 bpm, so a heart-rate number taken at one boundary does not mark the other (SYNTH-38).
Limitations
Small, homogeneous evidence. 10 studies of 6 to 13 young participants; 19 of 111 were women; no masters, elite or clinical samples. The authors call for studies across sex, age and fitness.
Few studies for internal load. The oxygen-uptake analysis rests on five studies and the heart-rate and lactate analyses on three each; "no difference" there is low-powered, and the studies with the largest power gaps did not contribute.
Mixed methods. CP models (2- or 3-parameter hyperbolic, "best fit"), trial durations (1-20 min), one intermittent-test study and MLSS windows varied; moderators for these could not be tested (fewer than three effect sizes per subgroup). Several SDs of the difference were imputed from a median correlation.
Risk of bias. Trial order was not shown to be randomised in any study, and carry-over was unclear in all.
Overlap of authorship. Per the reference list, two review authors (Keir, Pogliaghi) co-authored 4 of the 10 included studies, which are 4 of the 10 power samples; the competing-interests statement names only an editorial role at PeerJ; screening and inclusion were done by two other authors (LM, CFM). Funding was public (EU NextGenerationEU via the Italian MUR, VITALITY).
Reporting. The abstract and the Conclusions say the gap was larger in "fitter and younger" individuals and was affected by age; the Introduction's aims name age as a moderator, but the Methods restrict meta-regression to CP and MLSS power and the Results report only those, and every sample's mean age was 21.5-28 years.The age claim is not supported by an analysis shown in the paper. The Discussion's "CP is consistently higher than MLSS" overstates the study-level data (Iannetta 2019: mean CP 214 W vs MLSS 215 W, although its difference is tabulated as +1.0 W; prediction interval for a new study −11.62 to 36.47 W).The lactate forest-plot caption names Keir 2015, whose Table 3 row has no lactate difference, while Ozkaya 2022's has one (our reading of Table 3).
No training outcome. Agreement between two lab boundaries; nothing here says which anchor gives better adaptations (REF-173).
How this entry was checked and changed
Why this level
read from the full text (Europe PMC, CC BY 4.0). The rubric's A- line is an SR/MA "with serious heterogeneity or mostly small trials": 10 studies of 6 to 13 participants (n = 111), significant heterogeneity in the power-output analysis (Q(9) = 96.490, τ = 10.065 W), randomisation of trial order rated high or unclear risk in every study, and the heart-rate and lactate comparisons rest on three studies each. A measurement-agreement question, not a training outcome
Full text (checked 2026-10-27)
Ten of 782 studies were included; overall, 111 participants were included in the data analysis, and only 19 females were enrolled out of 111 participants (full text)
The pooled MD in PO between CP and MLSS was 12.42 W (95% CI [4.69–20.16], PI = −11.62 to 36.47), p = 0.005 (full text)
Agreement: mean bias of 12.42 W with 95% LoA = −19.23 to 44.08, outer 95% CI of LoA −30.91 to 55.76 (full text)
Oxygen uptake MD 0.09 L/min (p = 0.097), heart rate MD 0.61 bpm (p = 0.784), blood lactate MD 1.63 mM (p = 0.240); LoA −15.84 to 17.05 bpm for heart rate and −2.85 to 6.11 mM for lactate (full text)
Heterogeneity in W: Q(9) = 96.490, τ = 10.065 W, p < 0.001 (full text)
Meta-regression slopes: CP 0.21 (p = 0.002), MLSS 0.243 (p = 0.009); test of moderators F(1,8) = 21.091 and F(1,8) = 11.114 (full text)
Funding: European Union - NextGenerationEU, MUR grant ECS00000041 - VITALITY - CUP H33C22000430006; Carlo Ferri Marini is an Academic Editor for PeerJ (full text)