Zone 2 or intervals? What actually raises VO₂max

The gradient that started the argument

Between 1991 and 2014, the Cleveland Clinic put 122,007 patients through a symptom-limited treadmill test and followed them. Mandsager and colleagues reported the results in 2018: median follow-up 8.4 years, 13,637 deaths across 1.1 million person-years, and a mortality gradient steep enough to be startling.¹

Two things are worth holding onto. The first is that this is a retrospective cohort of people referred for a cardiac stress test, not a random sample and not a randomised trial. Fitness here is measured, not assigned, and people who perform badly on a treadmill are disproportionately people who are already unwell. Statistical adjustment narrows that problem without dissolving it. The right reading of a hazard ratio of 0.20 is that low fitness marks a group at very high risk, not that a training programme will multiply anyone's survival by five.

The second is that the pattern replicates across independent designs. Kodama and colleagues pooled 33 cohort studies of healthy men and women — 102,980 participants, 6,910 deaths — and found all-cause mortality risk about 13% lower per 1-MET increment in fitness (RR 0.87, 95% CI 0.84–0.90).² So the question should I care about my aerobic fitness? has a defensible answer before we reach protocol.

What the training trials actually show

Here the evidence changes character, and improves. Milanović, Sporiš and Weston pooled 28 controlled trials covering 723 participants of mean age 25.1 ± 5 years and baseline VO₂max 40.8 ± 7.9 mL/kg/min — an unremarkable starting point.³

Against no-exercise controls, interval training raised VO₂max by 5.5 mL/kg/min and continuous endurance training by 4.9. Head to head, intervals gained a further 1.2. The authors report these as mean differences with confidence limits of ±1.2, ±1.4 and ±0.9, using magnitude-based inference rather than null-hypothesis testing, so the head-to-head interval runs roughly 0.3 to 2.1 — excluding zero, but not by much. Their verdict was a possibly small beneficial effect. Intervals win, narrowly.

Why the interval advantage is smaller than it looks

Crowley and colleagues reviewed the reviews: 11 systematic reviews covering 179 primary studies.⁴ Three of the six meta-analyses comparing low- against high-intensity protocols favoured the higher intensity, with small to moderate effects. But the overview's verdict was that the apparent differences between LIT and HIT protocols on VO2max were either small, trivial, or inconclusive, and that any advantage appear[s] to be moderated by training variables other than intensity (e.g., training impulse, interval length, training volume, and duration) and participants' baseline characteristics (e.g., age and fitness levels).

That last clause explains most of the apparent disagreement in this literature. Interval and continuous arms are rarely matched on total work — the authors argue they must be before any comparison can be trusted — so part of what looks like an intensity effect may be a difference in dose. Baseline fitness cuts the same way. In the pooled trials both modes gave larger gains to those who started less fit: a group beginning near 40 mL/kg/min improves on almost any stimulus, a trained one far less. Two competent meta-analyses can therefore disagree, because they average different mixtures of participants and doses.

The case for going slowly, and where it is weakest

The strongest argument for low-intensity work is not a mechanism, it is an observation about what fast people do. Seiler and Kjerland tracked 11 well-trained junior cross-country skiers, VO₂max 73 ± 4 mL/kg/min, across 347 endurance sessions in 32 consecutive days.⁵ Classified by session goal, 75 ± 3% of sessions sat below the first ventilatory threshold, 8 ± 3% between the thresholds, and 17 ± 4% above the second. Perceived-exertion ratings and blood lactate agreed: across 60 consecutive sessions, 71% ran under 2.0 mmol/L. Counted as raw time, 91% of training minutes fell below the first threshold. Almost nothing happened at threshold itself.

This is descriptive, though: eleven teenagers on one training block tell you what fast athletes do, not what made them fast. The stronger test is randomised. Stöggl and Sperlich assigned 48 trained runners, cyclists, triathletes and skiers to nine weeks of four intensity distributions; 41 completed.⁶

The polarised group improved most, from 60.6 ± 8.3 to 67.4 ± 7.7 mL/kg/min. Note what that group was doing: 26% of its sessions were high intensity. Polarised training is not low-intensity training — it is a large base of easy work with hard intervals on top, and the arm closest to pure easy volume (83% low, 1% high) did not improve VO₂peak significantly. Easy work accumulates oxidative stimulus at low fatigue cost and makes the hard work survivable. It does not appear to substitute for it.

One popular argument for going easy does not survive contact with the data. Adherence is routinely asserted to favour gentler training; Santos and colleagues pooled 188 studies covering 8,928 participants and found otherwise.⁷ Attendance at supervised sessions averaged 89.4% (SD 11.8) for interval training and 92.5% (SD 10.6) for continuous training, not a significant difference across 65 studies (Hedges' g 0.015, 95% CI −0.088 to 0.118, p = 0.78). For unsupervised programmes, adherence was 63% (SD 21.1) against 68.2% (SD 16.2), again not significant across the 10 studies available (g −0.313, 95% CI −0.681 to 0.056, p = 0.096), on evidence the authors themselves graded very low certainty. There is no demonstrated adherence advantage either way.

Zone 2 is a filing system, not a metabolic switch

Seiler and Kjerland, whose data underpin much of the polarised-training case, were blunt about the zones themselves. Governing bodies had adopted scales of up to five aerobic zones; the authors warned that these numerous intensity zones suggest a degree of physiological specificity that is not really present, as the intensity zone boundaries are not clearly anchored in underlying physiological events.⁵ Their own three-zone model is bounded by two measured ventilatory thresholds, which in these skiers sat at 74 ± 2% and 89 ± 2% of VO₂max — 81 ± 2% and 91 ± 2% of maximum heart rate. Those are individually determined values, and they sit higher than the heart rates most people associate with the phrase.

The mitochondrial story attached to Zone 2 rests on thinner evidence than the VO₂max data does. Its most-cited support is San-Millán and Brooks, who measured fat and carbohydrate oxidation alongside blood lactate in 22 professional cyclists, 20 moderately active men and 10 men with metabolic syndrome.⁸ Peak fat oxidation was 0.66 g/min in the cyclists against 0.38 and 0.12 in the other two groups, and lactate correlated inversely with fat oxidation throughout (r = −0.97 in the cyclists). It is elegant, and cross-sectional: 52 men, tested once, no training intervention. It establishes that highly trained people oxidise more fat at a given workload; it does not establish that any particular training intensity produced that difference.

Whatever the distribution, it means nothing unrecorded. Kettle logs runs and rides, including GPS routes, and pairs with heart-rate monitors; the data stays on your iPhone and, with permission, can be written to Apple Health.