REF-1230 · Meyler et al., 2025 · Sports Medicine · Paper · PubMed

Should my zones come from a threshold or a % of max heart rate?

Probably a threshold: in 4 small trials of sedentary adults, threshold zones beat % of HR reserve for VO2max (+4.1 vs +1.8 mL/kg/min), but those arms also trained harder.

A-fairly strong evidenceMeasured on non-athletes and recreational people

Meyler and colleagues asked whether prescribing training intensity relative to physiological thresholds (THR: lactate or ventilatory thresholds, maximal lactate steady state, critical power) gives larger or more consistent VO2max gains than prescribing it as a percentage of a traditional anchor (TRAD: VO2max, VO2 reserve, HRmax, heart-rate reserve or maximal work rate).

They searched PubMed and Scopus to October 2023 for training studies of at least 3 weeks in healthy adults with VO2max measured directly before and after, asked authors for individual data, and obtained it for 42 of 236 eligible studies (1544 individuals).

Two Bayesian distributional models estimated the mean and the standard deviation of individual change: one on "controlled" studies that had volume-matched THR and TRAD arms and a non-exercising control group, one on all THR and TRAD arms.

In the controlled studies VO2max rose by 4.1 mL/kg/min with THR and 1.8 with TRAD (Bayes factor above 100), and 64% vs 16% of individuals were predicted to exceed the minimum important difference of one MET (3.5 mL/kg/min).

The spread of change did not differ between THR and TRAD (1.5 vs 1.7, Bayes factor 0.55), while exercisers varied more than controls (1.9 vs 1.3, Bayes factor 12.4). In the non-controlled set the means were 4.4 vs 3.4 (Bayes factor 35.1), with 60% vs 47% above the minimum important difference.

The authors conclude that threshold-based prescription gives larger average gains and more people with a meaningful gain, through a larger mean rather than a narrower spread.

Our reading: the first outcome-level evidence that the anchor matters, consistent with acute studies showing that a fixed percentage of a maximum lands people in different domains (REF-038); but the controlled result is 90 previously sedentary adults in four studies from one research group (a co-author's; our identification from the tables), and there the threshold arms trained from below VT1 to above VT2 while the traditional arms trained at 40-65% of heart-rate reserve, so the anchor and the intensity are not separated.

Protocol

Registration and reporting: PROSPERO CRD42021226644; PRISMA-IPD.
Eligibility: randomized controlled and non-controlled training studies in English published before October 2023; healthy men and women 18 or older, BMI 30 or less, without acute or chronic disease; at least 3 weeks of continuous or interval training (or both) by walking, running or cycling; VO2max measured directly by indirect calorimetry before and after; intensity anchored to a traditional value (VO2max, VO2 reserve, HRmax, heart-rate reserve, maximal work rate) or to a threshold (lactate threshold, gas exchange threshold, VT1, MLSS, respiratory compensation point, VT2, critical power; delta method included).Arms with nutritional or environmental add-ons excluded.
Search and selection: PubMed and Scopus (2021, updated 2023), reference lists and contact with research teams; two independent reviewers in Covidence, a third for disagreements. Authors were asked for anonymised individual data (age, sex, height, body mass before and after, VO2max).
Datasets: controlled, studies with volume-matched THR and TRAD arms and a non-exercising control (4 studies; THR n = 46, TRAD 44, control 49; 29 ± 5 sessions over 13 ± 1 weeks; 3 continuous, 1 combined); non-controlled, any THR or TRAD arm (THR arms from 18 studies, 354 individuals; TRAD arms from 25 studies, 1190 individuals; 7 ± 4 vs 14 ± 14 weeks; 23 ± 11 vs 43 ± 39 sessions).
Which controlled studies (our identification): the manuscript does not name them; the four studies whose arms sum exactly to Table 5's counts are Byrd 2019, Dalleck 2016, Weatherwax 2019 and Wolpern 2015 (THR 11 + 10 + 16 + 9 = 46, TRAD 11 + 9 + 13 + 11 = 44; sessions 30, 21, 33, 31).In all four the traditional arm's continuous training was at 40-65% of heart-rate reserve and the threshold arm's from below VT1 to above VT2 (Table 4; Byrd 2019 also had both arms do intervals, listed for the threshold arm at 100% VO2max); three list Dalleck, a co-author of this meta-analysis, as an author, and the fourth shares an author (Weatherwax) with two of them.
Analysis: one-stage Bayesian distributional models with random study intercepts (brms and Stan), group as a predictor of the mean and the log standard deviation of change; medians with 95% credible intervals for means and 75% for standard deviations; Bayes factors for the group terms (above 1 favours a difference, below 1 no difference; 100 or more "extreme"); 10,000 posterior predictions for the share exceeding the minimum important difference of 3.5 mL/kg/min.HERITAGE (n = 562, a traditional arm) analysed with and without; the results did not change.
Risk of bias: RoB 2 for randomized trials and ROBINS-I for others; overall low 10%, some concerns 80%, high 10%; one domain (domain 5) was rated high risk in 90%.
Funding and interests: no funding; the authors declare no conflicts of interest relevant to the review.

Key Results

DatasetVO2max change, THR (mL/kg/min)VO2max change, TRADEvidence for a difference in meansPredicted share above 1 MET, THR vs TRADSD of change, THR vs TRADEvidence for a difference in SD
Controlled (4 studies; THR 46, TRAD 44)4.1 (95% CrI 3.1 to 5.0)1.8 (0.9 to 2.8)BF > 100 ("extreme")64% vs 16%1.5 (75% CrI 1.2 to 1.8) vs 1.7 (1.4 to 2.1)BF 0.55 (anecdotal, no difference)
Non-controlled (THR 354, TRAD 1190)4.4 (3.7 to 5.2)3.4 (2.8 to 4.1)BF 35.1 ("very strong")60% vs 47%3.0 (2.7 to 3.3) vs 3.2 (2.9 to 3.5)BF 0.41 (anecdotal, no difference)

Exercise (THR and TRAD pooled, 90) vs non-exercising controls (49) in the controlled studies: SD of change 1.9 (75% CrI 1.7 to 2.2) vs 1.3 (1.1 to 1.6), Bayes factor 12.4 ("strong"), the authors' evidence that trainability truly differs between individuals.

The four controlled studies as reported in Table 4 (our identification; mean change in mL/kg/min, THR vs TRAD): Byrd 2019, 4.8 vs 2.5; Dalleck 2016, 3.8 vs 2.2; Weatherwax 2019, 3.5 vs 1.4; Wolpern 2015, 4.1 vs 1.4. Every study points the same way.

Response rate: the larger share of meaningful responders came from a larger mean change, not from a smaller spread; the hypothesis that THR reduces response variability was not supported.
Individual data obtained: 42 of 236 eligible studies (18%); THR data (354) were far fewer than TRAD (1190).
Numbers that do not add up: the manuscript compares HERITAGE (n = 562) with "that of the other eligible studies combined (n=953)", which totals 1515, not 1544; Table 5 lists the non-controlled TRAD sexes as 565 and 622 (1187) for 1190 individuals. The Limitations refer to "Table 3" for study characteristics, but Table 3 is the Bayes-factor scale.Table 4 lists THR arms in 21 studies and TRAD arms in 29 (8 with both, 42 rows), against 18 and 25 in Table 5 (our count); 354 + 1190 = 1544, so the headline 1544 excludes the 49 non-exercising controls; references 83 and 90 are the same Wolpern 2015 paper. These are from the accepted manuscript and may be corrected in the typeset version.

Mechanisms

Measured: a larger mean VO2max gain with threshold-anchored prescription in both datasets; no difference in the spread of individual change between anchors; a wider spread in exercisers than in controls.
Authors' interpretation: a percentage of a maximum places individuals in different intensity domains, so some get a much weaker stimulus; a threshold anchor normalises the stimulus, which raises the mean; the advantage should be largest around the heavy-severe boundary, where small differences in work rate change the response most.They meet the objection that whichever approach gives the higher relative intensity wins by arguing that only a threshold anchor makes "high" intensity high for the individual (citing REF-993, where gains tracked intensity relative to critical power rather than to VO2max).
Inference: in the four controlled studies the traditional arms were prescribed by heart-rate reserve (40-65%) and the threshold arms from below VT1 to above VT2 (Table 4 states heart rate at the ventilatory thresholds for two of the four); the threshold arms very likely spent more time at higher relative intensity, so this result is partly an intensity effect, and the meta-analysis cannot separate the two.
Inference: the non-controlled comparison sets threshold arms (shorter, 7 ± 4 weeks, fitter at baseline, 42 vs 34 mL/kg/min) against traditional arms from other studies; it agrees in direction but cannot isolate the anchor.
Inference: because every included study measured VO2max directly in a maximal test, the traditional anchors were probably mostly measured maxima rather than age formulas; if so, even a measured maximum did worse here than a threshold.

Practical Application

Should my zones come from a threshold or a % of max heart rate? Probably a threshold: in 4 small trials of sedentary adults, threshold zones beat % of HR reserve for VO2max (+4.1 vs +1.8 mL/kg/min), but those arms also trained harder.
Anchor zones to a measured boundary: heart rate, power or pace at a measured threshold (VT1, VT2, lactate threshold, critical power) beats a percentage of HRmax or heart-rate reserve for building zones (A-, mostly untrained adults; transfer to athletes).
What counts as a threshold here: laboratory measures; a 20-30 min field LTHR, an HRV-derived threshold (REF-1231) or a watch's estimate (REF-1232) were not tested and keep their own, weaker evidence.
It raises the average, not the floor: there was no evidence that threshold prescription narrowed the spread of responses (anecdotal Bayes factors 0.55 and 0.41); some athletes will still gain little from a block, and the anchor alone does not explain that.
Not a licence to ignore intensity: in the controlled studies the threshold arms also trained harder (inference); pick the anchor and the intensity.

Limitations

Bearing on [SYNTH-38]: the threshold anchors were laboratory measures in mostly sedentary adults; the guide's field-effort anchor and its use in trained athletes stay practice or transfer.
The decisive comparison is small and from one group: 4 studies, 90 exercising participants and 49 controls, all previously sedentary, from studies authored by or linked to a co-author (our identification from the tables).
Anchor confounded with intensity: threshold arms ranged up to above VT2 while traditional arms trained at 40-65% of heart-rate reserve (controlled studies), and the non-controlled set compares different studies with different durations, doses and baseline fitness.
Selection of data: individual data from 18% of eligible studies; whether the missing 82% would change the result is unknown; risk of bias "some concerns" in 80%.
Outcome: VO2max only; no performance, threshold or durability outcome; adherence was not collected.
Source: read from the author accepted manuscript; the typeset version was not seen; the numbers checked here match the published abstract. Minor count inconsistencies (1515 vs 1544; 1187 vs 1190).
Funding and interests: no funding; no conflicts declared, though a co-author is an author of three of the four controlled studies (our identification).
How this entry was checked and changed
Why this level

meta-analysis of training studies with individual data (n = 1544), read from the author accepted manuscript (University of Hertfordshire and London Metropolitan University repositories, found via OpenAlex; the typeset article is closed). A- rather than A: the decisive controlled comparison rests on 4 small studies (90 exercising participants, 49 controls) that share authors with the review (our identification from the tables) and confound anchor with intensity, the larger non-controlled comparison sets threshold arms against traditional arms from different studies, risk of bias had "some concerns" in 80% of studies, and individual data came from 42 of 236 eligible studies (18%)

Full text (checked 2026-10-27)
Electronic databases were searched, yielding data for 1544 individuals from 42 studies (full text)
Controlled studies: 4.1 [95%CrI: 3.1 to 5.0] for THR vs 1.8 [95%CrI: 0.9 to 2.8] for TRAD, BF>100; THR (64%) compared to TRAD (16%) above the MID (full text)
Controlled studies, SD of change: THR 1.5 [75%CrI: 1.2 to 1.8] vs TRAD 1.7 [75%CrI: 1.4 to 2.1], BF=0.55; training groups 1.9 [75%CrI: 1.7 to 2.2] vs CON 1.3 [75%CrI: 1.1 to 1.6], BF=12.4 (full text)
Non-controlled studies: THR 4.4 [95%CrI: 3.7 to 5.2] vs TRAD 3.4 [95%CrI: 2.8 to 4.1], BF=35.1; 60% of individuals in THR and 47% expected in TRAD beyond the MID (full text)
Table 5: controlled studies 4 / 4 / 4, individuals 46 / 44 / 49, baseline VO2max 31±7 / 29±5 / 29±6, training 13±1 weeks, 29±5 sessions (full text)
Whilst 236 studies met the predefined criteria, the response rate of IPD obtained was 18% (N=42) (full text)
HERITAGE IPD analysed (n=562) compared to that of the other eligible studies combined (n=953) (full text)
Wolpern 2015: THR 20-30 min @ HR <VT1 to >VT2, change 4.1 ± 0.9; TRAD 20-30 min @ 40-65% HRR, change 1.4 ± 1.8 (full text)
Risk of bias, Overall: low risk 10%, some concerns 80%, high risk 10% (full text)
No funding and financial assistance was used for this review (full text)