Does polarized training raise VO2max more than other mixes?
A little: SMD 0.24 over 11 trials, only in blocks under 12 weeks and highly trained athletes; no evidence of a gain in time trials, threshold power or TTE.
Silva Oliveira, Boppre and Fonseca asked whether polarized training (POL: about 75-80% of time below the first threshold, 15-20% above the second, under 10% between) improves endurance performance more than other training intensity distributions (TIDs).
They searched PubMed, Scopus and Web of Science to 20 October 2022 for controlled trials of at least 4 weeks and three sessions a week that compared POL with threshold (THR), pyramidal (PYR), HIIT, high-volume or other programmes and reported VO2peak, a time trial (TT), time to exhaustion (TTE) or speed or power at the second ventilatory or lactate threshold.
Seventeen studies (437 participants) were included and 14 pooled.
POL raised VO2peak more than other TIDs, by a small amount (SMD 0.24, 95% CI 0.01 to 0.48; 11 studies, 284 participants, I² 0%), and only in interventions under 12 weeks (SMD 0.40) and in highly trained/national-level athletes (SMD 0.46); in longer interventions (0.04) and in trained/developmental athletes (0.08) there was no evidence of a difference.
TT (−0.01), TTE (0.30, not significant) and threshold speed or power (0.04) did not differ. Taken one comparator at a time (THR, PYR, HIIT, control), no difference was detectable for any outcome. The authors note that the highly trained studies were also the short ones, so level and duration cannot be separated.
Our reading: a small, short-lived VO2peak advantage of a polarized block that does not reach performance; it fits REF-991 (no POL-PYR difference overall, possibly more VO2max on POL in competitive athletes) and argues for using a polarized phase as a bounded VO2peak tool rather than a season-long default.
Protocol
Registration and reporting: PROSPERO CRD42022365117, registered before data collection; PRISMA 2020.
Search: PubMed, Scopus and Web of Science, 10-20 October 2022, no filters, plus reference lists; English and peer-reviewed only; authors contacted for missing data; WebPlotDigitizer used to read values from figures.
Eligibility (PICOS): humans of both sexes aged 15 to 65 without limiting comorbidity; a POL intervention with three or more sessions a week for at least 4 weeks; any other TID as comparator (THR, PYR, HIIT or SIT, high-volume training) at three or more sessions a week; randomized or non-randomized controlled trials with baseline and post-test; at least one of VO2peak, TT, TTE or speed/power at VT2/LT2 (lactate threshold, lactate turn point and respiratory compensation point treated as synonyms).
Selection: 275 records, 205 after duplicates, 42 full texts, 17 included; two reviewers independently, a third for disagreements; no disagreements in the final selection.
Included studies (Table 1): 17 trials, 4 to 16 weeks (median 10); 437 participants (317 male, 89 female, 31 not reported), 183 in POL and 223 in other TIDs, one study without group sizes; sample size per study 11 to 52 (median 22); group-mean ages 17 to 44 years; 15 of 17 studies only or mostly men.Comparators: THR, PYR, HIIT, high-volume low-intensity, block training, CrossFit Endurance, moderate continuous training, low-intensity-only and training-load-guided programmes, and control groups (one described as continuing endurance training; the control group of the sedentary trial is not described in the text). Three studies did not report the TID.The polarized arms ranged from 68/6/26% (zone 1/2/3) to 93/2/6%.
Effect measures: SMD (Hedges' g), random-effects inverse-variance model, R "meta" package, p < 0.05; I² for heterogeneity; leave-one-out sensitivity; funnel plot and Egger's test for VO2peak and threshold power; GRADE for certainty.
Subgroups: intervention duration (< 12 vs ≥ 12 weeks), starting level (highly trained/national vs trained/developmental), TT duration (< 12 vs ≥ 12 min) and comparator type; a pre-registered sex analysis could not be run because sex was too poorly reported.
Funding and interests: Portuguese Foundation for Science and Technology (FCT) grants to the research centres and authors; the authors declare no competing interests.
Key Results
POL vs all other TIDs (Table 3 and the Results text):
| Outcome and subgroup | Participants | SMD (95% CI) | p | I² |
|---|---|---|---|---|
| VO2peak, all (11 studies) | 284 | 0.24 (0.01 to 0.48) | 0.040 | 0% |
| VO2peak, interventions < 12 weeks | 163 | 0.40 (0.08 to 0.71) | 0.01 | 0% |
| VO2peak, interventions ≥ 12 weeks | 121 | 0.04 (−0.32 to 0.40) | 0.83 | 0% |
| VO2peak, highly trained/national level | 125 (abstract) or 118 (Table 3) | 0.46 (0.10 to 0.82) | 0.01 | 0% |
| VO2peak, trained/developmental | 159 | 0.08 (−0.23 to 0.39) | 0.62 (Table 3) | 0% |
| Time trial, all | 221 | −0.01 (−0.28 to 0.25) | 0.92 | 0% |
| Time to exhaustion, all | 66 | 0.30 (−0.20 to 0.79) | 0.24 | 0% |
| Speed or power at VT2/LT2, all | 253 | 0.04 (−0.21 to 0.29) | 0.75 | 0% |
One comparator at a time (Table 2), VO2peak: POL vs THR 0.24 (−0.46 to 0.94; 78 participants; I² 53%, the only notable heterogeneity, driven mostly by Stöggl 2014 — removing it did not change the result); POL vs PYR 0.08 (−0.39 to 0.55; 70); POL vs HIIT 0.34 (−0.27 to 0.95; 43); POL vs control groups 0.34 (−0.07 to 0.76; 93). TT, TTE and threshold power showed no difference in any pairing (for example POL vs PYR threshold power −0.14, −0.57 to 0.30).
TT by test length: under 12 min −0.00 (−0.37 to 0.37), 12 min or longer −0.03 (−0.41 to 0.36); TTs ranged from a 100-m swim to a 40-km cycling time trial.
Publication bias: Egger's test not significant for VO2peak (p = 0.632) or threshold power (p = 0.092).
Certainty: GRADE "high" for all outcomes; RoB-2 "some concerns" (randomization and selection of reported results) for every randomized trial except two rated low risk in the randomization domain.
Numbers that do not add up: the highly trained VO2peak subgroup is n = 125 in the abstract and N 118 in Table 3; 125 + 159 = 284 matches the pooled VO2peak total and 118 + 159 = 277 does not (our arithmetic).The Results text gives p = 0.82 for trained athletes' VO2peak and p = 0.84 for TT in interventions of 12 weeks or more, while Table 3 gives 0.62 and 0.53, and the table's z values (0.50 and −0.63) give the table's p values (our arithmetic). The Methods say mean differences were used for VO2peak and TTE, but every VO2peak and TTE result is reported as an SMD.The Methods apply GRADE to the primary outcome (VO2peak) while the Results rate "all outcomes" high certainty, and they define the duration subgroups as ≤ 12 vs ≥ 12 weeks while Table 3 uses < 12 vs ≥ 12.The abstract's 125 is the figure that reconciles with Table 1 (Stöggl 2014's polarized arm counted once per comparator, plus Hebisz, Treff and Hebisz: 125; plus Festa 2020's 38 gives the < 12-week N of 163; our arithmetic). One ROBINS-I sentence cites a non-included paper (the polarization-index article) as one of two low-risk studies.
Mechanisms
Measured: VO2peak rose more with POL in short interventions; threshold speed or power, TT and TTE did not differ. Nothing physiological (blood volume, muscle) was pooled.
Authors' interpretation: high-intensity work expands blood volume and activates PGC-1α-driven mitochondrial biogenesis faster than low or moderate intensity, so a POL block raises VO2peak sooner; other TIDs reach the same VO2peak more slowly, which is why the advantage disappears beyond 12 weeks. Threshold adaptations (lactate transport and clearance) follow from both low- and high-intensity training, so POL has no edge there.
Authors' interpretation: the level effect (only highly trained athletes gained more) cannot be separated from intervention length, because the highly trained studies were also the short ones.
Inference: from Table 1, the four highly trained studies that measured VO2peak (Stöggl 2014, both Hebisz 2021 mountain-bike studies, Treff 2017) all lasted under 12 weeks and make up most of the under-12-week subgroup, so the two significant subgroups are largely the same trials.
Inference: in several included trials the polarized arm trained for more time than its comparator (Stöggl 2014: 104 vs 66 h for HIIT; Carnes 2018: 283 vs 117 min a week for CrossFit Endurance), in at least one for less (Neal 2013: 381 vs 458 min a week), so part of a VO2peak edge can be more training rather than a better distribution.
Practical Application
Does polarized training raise VO2max more than other mixes? A little: SMD 0.24 over 11 trials, only in blocks under 12 weeks and highly trained athletes; no evidence of a gain in time trials, threshold power or TTE.
Use it as a bounded block, not a season mode: a polarized phase of a few weeks is a reasonable way to push VO2peak in a highly trained athlete before a block that depends on aerobic power; in trials of 12 weeks or more there was no evidence of a difference (SMD 0.04), but those were different, mostly trained-level trials, so no catch-up was measured; "other distributions catch up more slowly" is the authors' interpretation (the "few weeks" is our reading).The authors' advice to put polarized blocks into pre-competitive phases is their extrapolation; no included trial compared phases.
For a trained (Tier 2) athlete the choice is open: there was no evidence of a difference in VO2peak (SMD 0.08, −0.23 to 0.39) or any other outcome; choose the distribution the week's hours and the goal allow (SYNTH-26). REF-991 (A-, individual-data network meta-analysis) found that recreational athletes may even gain more VO2max on pyramidal.
Do not promise a faster race from the label: the time-trial pooled effect was −0.01 (−0.28 to 0.25), with tests from 100 m to 40 km and no difference by test length.
Against a single named alternative the evidence is thin: threshold, pyramidal and HIIT comparisons each had 42 to 132 participants per outcome and none differed; REF-303's large polarized advantage over threshold training was the main source of heterogeneity in that comparison.
What "polarized" meant here: the polarized arms ranged from 68/6/26% to 93/2/6% of time in zones 1/2/3, and three studies gave no distribution at all; a programme labelled polarized is not one prescription (REF-467 on how counting method changes the label).
Limitations
Small trials. Median 22 participants per study (11 to 52); subgroup and head-to-head comparisons rest on 42 to 163 participants.
Two significant subgroups that overlap. The authors state that highly trained studies were also the short ones; the level and duration effects cannot be told apart.
Comparators and load. "Other TIDs" mixes threshold, pyramidal, HIIT, high-volume, block, CrossFit Endurance and control groups who trained; several arms were not matched for training time, and three studies did not report the distribution. The authors list under-reporting of TID, TRIMP, volume and frequency as a main limitation.
Not only athletes. The pooled VO2peak result includes a trial in sedentary young women with overweight or obesity (the largest study, 52 participants), in a review titled on athletes.
Certainty. GRADE "high" for every outcome, although RoB-2 rated every randomized trial "some concerns" (randomization and selection of reported results; two were low risk for randomization).
Men. 15 of 17 studies only or mostly men; the planned sex analysis was impossible.
Reporting. Subgroup n and two p values differ between abstract, text and Table 3 (see Key Results); VO2peak effect measure described as a mean difference in the Methods and reported as an SMD.
Scope. English only, three databases, search to October 2022; later trials are not included.
Funding: public (FCT); no competing interests declared.
How this entry was checked and changed
Why this level
a systematic review and meta-analysis of controlled trials, read from the full text (open access, CC BY). The rubric gives A- rather than A because the trials are mostly small (median 22 participants, range 11 to 52) and the main effect sits just above zero; the authors' "high certainty" GRADE rating stands beside RoB-2 "some concerns" for every randomized trial (two were low risk only in the randomization domain), several comparisons were not matched for training time, and the pooled VO2peak result includes one trial in sedentary adults
Full text (checked 2026-10-26)
Seventeen studies met the inclusion criteria (n = 437 subjects): 317 males, 89 females, 31 undisclosed (full text)
VO2peak: SMD = 0.24 [95% CI 0.01, 0.48]; z = 2.02 (p = 0.040); 11 studies, n = 284; I2 = 0% (full text)
Superiority only in interventions under 12 weeks, SMD = 0.40 [95% CI 0.08, 0.71], n = 163; at 12 weeks or more 0.04 (-0.32; 0.40) (full text)
Highly trained athletes: SMD = 0.46 [95% CI 0.10, 0.82], n = 125 in the abstract and N 118 in Table 3; trained/developmental 0.08 (-0.23; 0.39) (full text)
TT SMD -0.01 [95% CI -0.28, 0.25], n = 221; TTE 0.30 [-0.20, 0.79], n = 66; V/P VT2/LT2 0.04 [-0.21, 0.29], n = 253 (full text)
POL vs THR VO2max/peak 0.24 (-0.46; 0.94), I2 53%; POL vs PYR 0.08 (-0.39; 0.55) (full text)
Studies including highly trained athletes were also those that had a shorter duration (< 12 weeks) (full text)
The sample size in each study varied between n = 11 and n = 52 participants, with a median of n = 22 per study (full text)