The decision. train in the fat-burning zone is a popular prescription: exercise at the intensity that maximises the rate of fat oxidation (FATmax, a.k.a. LIPOXmax or maximal fat oxidation / MFO intensity), found per-person from a graded exercise test. Does prescribing aerobic exercise at FATmax actually improve body composition and cardiometabolic markers in adults with overweight or obesity — and is it any better than just doing ordinary moderate aerobic exercise? Jiang 2026 is a gold-tier meta-analysis (19 RCTs) that pools FATmax training against non-exercise or diet-only controls. (inferred from Jiang et al., 2026)
FATmax is a way to set the intensity, not a distinct kind of exercise
The single most important thing to get right about this literature: FATmax is a prescription method — an individually-measured intensity target — applied within ordinary aerobic exercise (walking, jogging, cycling), not a separate modality. In this stratum it lands at a low intensity. The recommended target is «structured exercise at FATmax inten sity (typically 42%–49% ̇VO2 max)» (Jiang et al., 2026), and «the trials we included most often prescribed or elicited intensities between approximately 35% and 66% ̇VO2 max» (Jiang et al., 2026). (corrected 2026-10-06: 34.5-46% -> ~35-66%; self-critique)
That band sits mostly at the lower end of, and overlapping, the moderate-aerobic (AE-M, 45-65% VO2max) category ranked by Exercise Modality for Body Composition in Obesity. So FATmax is best understood as individualised low-to-moderate aerobic exercise — the same activity that framework already covers, prescribed by substrate-crossover intensity rather than a fixed %VO2max. The low band is not an obesity-specific downshift: «Croci et al. [81] reported no between-group difference in FATmax intensity expressed as a percentage of ̇VO2 max in physically active overweight versus lean men matched for cardiorespiratory fit ness, suggesting that adiposity per se does not necessarily imply a lower relative FATmax intensity» (Jiang et al., 2026). The low intensity «may more plausibly reflect the predominance of untrained/low-active participants and the frequent use of low-impact modalities (e.g., walking) [83], rather than a universal shift toward lower % ̇VO2 max» (Jiang’s hedge) (Jiang et al., 2026). (corrected 2026-10-06: flat reflects -> Jiang’s may more plausibly reflect; self-critique)
Is FATmax distinct from moderate aerobic, or a relabel? — a distinction, not a tension
Before comparing the FATmax effect to O’Donoghue’s moderate-aerobic ranking, the parameters have to be matched. They are not the same quantity on any axis:
| Parameter | Jiang (FATmax MA) | O’Donoghue (AE-M, NMA) | Same quantity? |
|---|---|---|---|
| Object compared | an intensity-prescription method | an exercise modality category | No |
| Intensity | typically ~42-49% VO2max (Jiang’s hedge — not a fixed band; ~35-66% observed) | AE-M = 45-65% VO2max | No — overlaps the lower/mid AE-M band |
| Comparator | non-exercise / diet-only control | network of 6 modalities + control | No |
| Population | overweight+obese BMI>=25, incl. adolescents, T2D/NAFLD | obesity BMI>=30, adults | No |
| Body-weight effect | MD -3.66 kg vs control | -0.05 to -1.01 kg across modalities | No — not comparable |
Because every cell differs, the apparent clash (FATmax’s -3.66 kg vs O’Donoghue’s ~1 kg ceiling) is not a joined tension — it is a distinction at different scope and comparator (not-joined check (ii)). The two sources answer different questions: O’Donoghue asks which modality; Jiang asks how to set the intensity. And critically, Jiang holds no active moderate-aerobic comparator arm — every trial is FATmax-vs-control or FATmax-vs-diet. So this meta-analysis cannot establish that FATmax beats ordinary moderate aerobic exercise; there are also «no head-to-head trials» of FATmax vs lactate-threshold prescription (Jiang et al., 2026). Whether individualising intensity to the fat-oxidation crossover adds anything over a fixed moderate dose is an open gap (G), not a demonstrated benefit. (inferred from Jiang et al., 2026)
What FATmax training moves — the pooled effect menu
Across 19 RCTs (N = 644), versus non-exercise or diet-only control, FATmax aerobic training (3-5x/week, ~20-90 min/session — most 45-60, the Tunisian paediatric trials 90 — over 8-20 weeks; Table 1 (Jiang et al., 2026); corrected 2026-10-06 from 45-60 min, 8-16 weeks; self-critique) produced (MD for anthropometry, SMD for cardiometabolic; all p<0.01):
- Body composition: body weight -3.66 kg (95% CI -4.98, -2.35); fat mass -2.86 kg (-3.89, -1.83); body fat -2.32% (-3.65, -1.00); waist -4.21 cm (-6.27, -2.16); BMI -1.66 (-2.23, -1.09).
- Cardiometabolic: VO2max SMD 1.45 (0.98, 1.93, large); fasting glucose SMD -1.28 (-1.84, -0.72, large); HOMA-IR -0.57 (-0.83, -0.31); fasting insulin -0.53 (-0.79, -0.27). The VO2max gain was larger in under-18s — «greater improvements were observed among participants aged <18 years, those with BMI ≥ 30 kg/m2, and among African cohorts» (Jiang et al., 2026) — who make up 8 of 19 trials (Table 1, chunk 02), and publication bias is flagged for VO2max and fasting glucose, so the adult magnitude is likely smaller.
- Weak or null: the lipid panel (TC, TG, HDL-C, LDL-C) and DBP were non-significant, with at most a SBP signal (exact CIs figure-bound).
The point estimates and p-values are the abstract’s (Jiang et al., 2026); the 95% CIs and prediction intervals are read from the typeset results section (§3.4-3.5, PDF p.6) because that dense two-column paragraph did not extract fully into the chunk text — several CI/PI bounds are absent from the chunk markdown, so they are a direct-read, not a chunk-locatable quote (Jiang et al., 2026).
Two honesty flags on these numbers. First, the 95% prediction intervals cross zero for BMI, body fat %, waist and fasting glucose — so while the group mean is a real reduction, the effect in a new population or individual may be null; the authors say as much, that «effects may not be consistent across all populations, likely reflecting methodological variability in FATmax assessment» (Jiang et al., 2026). Second, every outcome is a surrogate — no mortality or cardiovascular event is measured -> Surrogate Outcomes. Overall certainty is low: the abstract reports «overall moderate methodological quality» and that «Overall GRADE ratings were predominantly very low to» moderate (Jiang et al., 2026) (the per-outcome GRADE table is Supporting-Information-bound).
Part of the body-composition signal may be a measurement artifact
The intensity in nearly all trials was set from a single graded exercise test, which overestimates FATmax in obesity: steady-state fat utilisation and energy expenditure run «approximately 46% and 36% lower, respectively, than the values predicted by the GXT» (Jiang et al., 2026). The moderator analysis then shows the artifact’s fingerprint — the trials using Short-Step protocols (<=4 min) and the Highest-observed method, i.e. those most likely to have overestimated the intensity (pushing participants harder than true FATmax), reported «significantly larger reductions in body fat percentage (QM = 22.31, p < 0.01)» (Jiang et al., 2026). So some of the fat-loss signal may be an intensity-measurement artifact rather than a property of training at the true crossover.
The optimal 2446 min is not a target
The restricted-cubic-spline dose-response is U-shaped, with an apex the authors carry into their Practical Applications as an «optimal moderate-volume dose of 60 min per session, 5 times per week, for at least 8 weeks (totaling ~2446 min)» (Jiang et al., 2026). Do not read 2446 min as a prescription target. The authors demote it themselves: «while we identified a U-shaped dose–response relationship, the extremes of the dos ing curve were driven by a limited number of research groups; therefore, this observational finding requires independent repli cation to confirm the optimal training volume» (Jiang et al., 2026).
This is a textbook The Underivable Optimum case: a point optimum read off an observational curve whose arms are driven by a few research groups, with no confidence interval on the apex and no causal design behind the descending arm. The descending (high-volume) arm is explained by energy compensation and hormonal/overtraining responses — a mechanism consistent with Exercise Energy Compensation, but a mechanism, not a measured causal fall. Treat the apparent harm of too much with The U-Shaped Association Artifact discipline: the upper arm is a hypothesis to replicate, not a dosing ceiling to obey. (inferred from Jiang et al., 2026)
Decision relevance
- For an inactive adult with overweight or obesity, the big rock is doing regular aerobic activity at all -> Layer 1 - Ranking Interventions for a Stratum. FATmax (individualised low-to-moderate aerobic) is a reasonable, well-tolerated way to prescribe that starting dose, and it improves fitness and glycaemic markers vs control (the fitness gain larger in under-18s, ~8/19 trials; publication bias flagged for VO2max and fasting glucose — the adult magnitude is likely smaller). But its marginal rank over ordinary moderate aerobic exercise is not established by this evidence — there is no head-to-head arm.
- Do not sell FATmax as a superior fat-burning method. The popular framing (“the fat-burning zone melts fat fastest”) outruns the evidence: the MA shows FATmax beats doing nothing, which any aerobic exercise does, and part of the fat-loss signal tracks intensity-measurement error. Whether the crossover-intensity individualisation adds anything is an open gap.
- Weight loss still needs a dietary deficit. The -3.66 kg vs control may partly reflect dietary co-intervention arms (where controls did not diet) and the measurement-artifact moderator above; the source tested dietary restriction as a moderator and reported it as significant for VO2max, not body composition, so this is a hypothesis, not a finding (corrected 2026-10-06: is inflated by diet co-intervention and higher volumes -> hypothesis; self-critique). It does not overturn the held picture that exercise alone barely moves the scale -> Exercise Modality for Body Composition in Obesity, Does Weight Loss Reduce Cardiovascular Events. Body-composition and fitness, not weight, are where aerobic training earns its place.
- If measuring FATmax, validate it. A single GXT overestimates the target; a 40-60 min steady-state verification trial and extended (5-10 min) test stages are needed to prescribe the true intensity -> Measuring and Raising Cardiorespiratory Fitness.
Limits
- Single gold source,
confidence: low. One meta-analysis (N=644) carries the effect menu; overall GRADE certainty is very-low-to-moderate (the per-outcome GRADE table is SI-bound) and the prediction intervals cross zero for several outcomes. - No active comparator. FATmax-vs-control and FATmax-vs-diet only — the headline effective cannot be read as better than ordinary moderate aerobic.
- Measurement artifact in the exposure. GXT-set intensity is overestimated in obesity, and the trials most likely to overestimate reported the largest fat loss, confounding the body-composition result.
- Heterogeneous, non-generalisable sample. Concentrated in Chinese/Tunisian/French cohorts, mixing adolescents and metabolically-ill strata; dietary co-intervention varied; surrogate outcomes only; short interventions (8-20 weeks; corrected 2026-10-06 from 8-16, Table 1 chunk 02). Publication bias flagged for VO2max, fasting glucose and DBP.
Self-critique [run 2026-10-04, before commit]
- Overclaim guard: the headline is held to what the evidence supports (FATmax beats no-exercise; superiority over moderate aerobic is an explicit gap, not a finding); the effective strategy conclusion is quoted with its own PI-crosses-zero qualifier.
- Fake-tension guard: the FATmax -3.66 kg vs O’Donoghue ~1 kg difference is filed as a distinction (different comparator, dose, population, and object) via the parameter table and the not-joined check, not laminated as a tension; the held exercise alone barely moves weight claim is preserved.
- Underivable-optimum / U-artifact guard: the 2446-min apex is explicitly not treated as a target; the U-curve’s descending arm is marked observational/mechanistic and routed to the artifact discipline.
- Surrogate-as-outcome guard: all outcomes marked surrogates; no mortality/event claimed.
- Attribution guard: every effect size is chunk-pinned and carries its CI (and PI where reported); the distinction rests on individually-quoted, matched parameters.