Where FATmax-intensity training fits (not a seventh modality)
Train at FATmax, in the fat-burning zone names an intensity-prescription method, not a modality: individually-measured maximal-fat-oxidation intensity, applied within aerobic exercise. In adults with overweight/obesity it is prescribed at ~42-49% VO2max (34.5-46% observed), i.e. at or below the floor of the AE-M (45-65%) band ranked above — so it is individualised light-to-low-moderate aerobic, folded into this page’s aerobic category rather than ranked beside it -> FatMax Training for Body Composition and Cardiometabolic Health (Jiang et al., 2026) (observed range: chunk 02).
The FATmax meta-analysis (Jiang 2026) reports a much larger body-weight reduction (MD -3.66 kg vs non-exercise/diet-only control) than this NMA’s ~1 kg across modalities. That is a distinction, not a contradiction — the comparator (vs control, not within a modality network), the training volume, the inclusion of diet co-intervention in several trials, and the population (BMI>=25 incl. adolescents and metabolically-ill, vs BMI>=30 adults here) all differ. It does not overturn the finding below that exercise alone is not a weight-loss lever; the FATmax number is a confounded against-control effect, and its own prediction intervals for BMI/body-fat/waist cross zero (Jiang et al., 2026). (inferred from O’Donoghue et al., 2020)
The decision. An adult with obesity decides to exercise. Does it matter which kind — aerobic, resistance, or both — for body composition and fitness? O’Donoghue 2021 is a gold-tier network meta-analysis (45 RCTs, 3566 adults with BMI >=30) that ranks six modalities head-to-head against each other and against no-exercise control, across four anthropometric outcomes plus CRF. It is the integrating comparison the component exercise pages do not hold: Measuring and Raising Cardiorespiratory Fitness has HIIT-vs-MICT, Exercise vs Caloric Restriction for Visceral Fat has exercise-vs-diet, but neither ranks aerobic-vs-resistance-vs-combined for body composition in this stratum. (inferred from O’Donoghue et al., 2020)
The six modalities (F.I.T.T.-defined): AE-V aerobic vigorous (>65% VO2max), AE-M aerobic moderate (45-65%), R-HI resistance high load (>75% 1RM), R-LM resistance low-moderate (50-75% 1RM), COM-HI combined high-intensity, COM-LM combined low-moderate (O’Donoghue et al., 2020).
The robust finding: combined on top, low-load resistance-only on the bottom
«Our main findings indicate that in subjects with a BMI >=30 kg m−2, a combined exercise intervention consisting of aerobic and resistance training (COM-HI or COM-LM) is the most promising for reducing WC and %BF, as well as increasing CRF, despite no substantial weight loss. Out of the six interventions investigated, low-to-moderate load resistance training was deemed to be the least effective.» (O’Donoghue et al., 2020) The top three overall were COM-HI, COM-LM and AE-M.
What survives as a decision is coarse: aerobic-containing and combined programmes rank above low-moderate-load resistance-only, and combining aerobic with resistance ranks at or near the top of every anthropometric outcome. For BMI the source states the aerobic-component arms «were superior to resistance only interventions» (near-identical wording for waist circumference; not stated for weight) (O’Donoghue et al., 2020). High-load resistance (R-HI) is not separated from aerobic: its Table 3 point estimates vs control are similar (weight −0.63 vs AE-V −0.65 kg; %BF −1.81 vs AE-M −1.70 / AE-V −1.27) with wider CIs that cross zero (as AE-V’s %BF CI also does), and it outranks both aerobic arms on %BF (P-score .58) (O’Donoghue et al., 2020). The fine ordering among the leaders does not survive (next section). (corrected 2026-10-06: aerobic beats resistance-only, consistent across outcomes -> scoped to low-moderate-load resistance; self-critique)
P-score ranking (best -> worst), by outcome
| Rank | Weight | BMI | Waist circ. | %Body fat | CRF (VO2max) |
|---|---|---|---|---|---|
| 1 | COM-HI .83 | COM-HI .85 | COM-LM .82 | COM-HI .80 | COM-HI .73 |
| 2 | COM-LM .66 | AE-M .78 | AE-M .69 | COM-LM .68 | AE-M .64 |
| 3 | AE-M .65 | COM-LM .69 | AE-V .59 | R-HI .58 | AE-V .39 |
| 4 | AE-V .55 | AE-V .45 | R-HI .45 | AE-M .55 | COM-LM .39 |
| 5 | R-HI .50 | R-HI .44 | R-LM .42 | AE-V .48 | R-HI .37 |
| 6 | R-LM .16 | R-LM .21 | — | R-LM .40? | R-LM .18? |
P score is the frequentist SUCRA analogue (0-1, 1 = best with no uncertainty) (O’Donoghue et al., 2020). No study reported COM-HI for waist circumference, so it is absent from that column. Row 6 of Table 4 extracts ambiguously as four R-LM values (.16/.21/.40/.18); CRF used «all six intervention categories», so R-LM is in that network, and the %BF/CRF assignment (.40/.18) is inferred from the P-score sum (k/2) — R-LM is last or near-last either way. (corrected 2026-10-06: %BF .18 / CRF — -> .40? / .18?; self-critique)
Why the fine ranking is a ranking of noise
The P-scores order the modalities, but most of the between-modality differences behind them are not statistically significant — the superiority that is significant is almost always aerobic-containing vs control or aerobic-containing vs low-moderate-load resistance-only, not one leader over another. In the CRF network the problem is total: no modality reached significance at all. «Although all interventions resulted in an increase in absolute VO2max, there were no statistically significant improvements in fitness found; in accordance with the P score, COM-HI was the intervention … most likely to increase VO2max.» (O’Donoghue et al., 2020) So the CRF ranking (COM-HI .73 on top) orders a set of null results.
The authors supply the governing caveat themselves: «such scores should be interpreted cautiously unless there are actual clinically meaningful differences between interventions.» (O’Donoghue et al., 2020) Read the ranking at the resolution the evidence supports: combined and aerobic beat low-moderate-load resistance-only; aerobic vs high-load resistance is not separated; the precise order among the top arms is not established. (inferred from O’Donoghue et al., 2020)
A different stratum is silent on between-mode CRF differences — Khalafi (post-menopausal women; CRF analysis «35 intervention arms from 25 studies» of a 129-study review) finds every mode raises CRF vs control (aerobic SMD 1.21, resistance 1.26, combined 1.47, all significant) but runs no head-to-head test between modes (Khalafi et al., 2023), so it neither confirms nor refutes this NMA’s non-separation. Versus control the two disagree: Khalafi finds significant CRF gains for every mode, this NMA finds none; O’Donoghue’s use of absolute (L/min) VO2max is a candidate explanation, not yet checked . (corrected 2026-10-06: 129 RCTs … converging -> 25-study CRF analysis, silent on between-mode order; self-critique) The cross-stratum synthesis (matched-parameter table, the low-stakes modality-for-CRF decision) is worked on Measuring and Raising Cardiorespiratory Fitness.
Exercise alone is not a weight-loss lever — regardless of modality
The single most decision-relevant number is how small the weight effect is. «Overall, weight loss was minimal, with mean values ranging from −0.05 to −1.01 kg.» (O’Donoghue et al., 2020) across every modality. The authors find this similar to earlier meta-analyses of aerobic and combined exercise, which they report as concluding «for weight loss, exercise alone is not an effective therapy, a hypocaloric balance is necessary» (O’Donoghue et al., 2020) — that sentence is O’Donoghue citing prior work (their refs 18-19), not a conclusion pooled in this NMA; the NMA’s own finding is the ~1 kg ceiling above (corrected 2026-10-06: authors conclude -> authors report earlier MAs’ conclusion, ODonoghue chunk 02).
This is the modality-resolved version of the vault’s held picture on Exercise Energy Compensation (energy expended is partly offset by compensatory intake) and Does Weight Loss Reduce Cardiovascular Events (exercise disappoints as a scale lever). O’Donoghue adds that no choice of modality rescues exercise as a weight-loss tool — the ~1 kg ceiling holds across aerobic, resistance and combined. The body-composition effects (%BF, WC) are the better-responding outcomes, and %BF is argued to be more responsive than body weight.
Magnitudes and the surrogate boundary
The anthropometric effects are small in absolute terms: best-arm reductions of roughly 1 kg (weight), ~2.8 cm waist (COM-LM), ~2.8% body fat (COM-HI; -2.82, CI -5.50 to -0.14 (O’Donoghue et al., 2020)). CRF rose 7-15% but non-significantly. The BMI estimates (COM-HI −2.79, CI −5.95 to −0.36; AE-M −1.76) are incompatible with the same arms’ ~1 kg weight changes at the sample’s mean (90.5 kg, BMI 32.3: −2.79 kg/m2 is ~7.8 kg), and the COM-HI CIs are asymmetric (O’Donoghue et al., 2020); treat the BMI magnitudes as unreliable — the weight figure is the coherent one. (corrected 2026-10-06: ~2.8 BMI units dropped from the small list; self-critique) Every outcome here is a surrogate — no mortality or CV event is measured -> Surrogate Outcomes.
The paper re-expresses its surrogates as risk reductions, but those figures are borrowed coefficients, not measured outcomes. «an increase of 1 cm in WC resulted in a 2% increase in CVD risk» (a separate meta-regression) is applied «in reverse» to claim COM-LM’s WC loss buys a 5.5% CVD-risk decrease; and «an improvement in fitness of between 8% and 10% is associated with a 12% decrease in mortality» is imported to argue the CRF gains matter (O’Donoghue et al., 2020). These multiply this NMA’s surrogate effects by effect sizes from other studies — treat them as the source’s cited context, never as evidence that a modality prevents disease here. (inferred from O’Donoghue et al., 2020)
Decision relevance
- If an adult with obesity is going to exercise for body composition, combined (aerobic + resistance) is the best-supported choice and low-moderate resistance alone the weakest. The best-supported part is don’t choose low-load resistance alone; aerobic vs high-load resistance is not separated, and the fine order among leading arms is not established.
- Do not prescribe exercise as a weight-loss intervention. At ~1 kg across all modalities, the scale barely moves; weight loss needs a dietary energy deficit, with exercise combined in for its body-composition, fitness and muscle-sparing edge -> Exercise vs Caloric Restriction for Visceral Fat, Does Weight Loss Reduce Cardiovascular Events.
- Rank this within Layer 1. For an inactive adult with obesity the big rock is doing regular activity at all; modality choice is a second-order refinement once that rock is being pulled -> Layer 1 - Ranking Interventions for a Stratum. The one place modality choice earns weight is where low-moderate-load resistance-only would be chosen instead of an aerobic-containing programme — the NMA ranks that the weakest option for body composition.
- Out of scope (appraise, not prescribe): the specific sets/reps/intensity to program, and managing exercise around comorbidity. For resistance-training dose, see Resistance Training Prescription - Load Sets and Frequency; for the muscle/strength->function->mortality case for including resistance, Muscle-Strengthening Activity and Mortality.
Limits
- Single gold source,
confidence: low. One NMA carries the ranking; the cross-links to compensation, visceral fat and surrogate honesty are the wiki’s weave, not replication of the ranking. - High risk of bias. «more than 70% (n = 33) of the included trials were judged as being at moderate to high ROB … the results of this present NMA should be interpreted in a conservative manner»; only 27% used intention-to-treat (O’Donoghue et al., 2020).
- Sparse data for the winner. Aerobic dominated; resistance and combined each contributed only ~10-18% per network, and no study reported COM-HI for waist circumference — so the combined-training advantage rests on few head-to-head comparisons and the authors warn these are less reliable.
- Surrogate outcomes only; small absolute effects; short interventions (mean 20 weeks, mostly 3x45 min/week) — the ranking cannot be extrapolated beyond that studied dose or to non-obese adults.
- The CRF ranking orders non-significant results and must not be read as establishing a fitness-modality order.
Self-critique [run 2026-10-04, before commit]
- Overclaim guard: the headline is held at the resolution the evidence supports (aerobic-containing beats low-moderate-load resistance-only; combined ranks top) with the fine ordering explicitly demoted as non-significant; the CRF ranking is flagged as ordering null results, not asserted as a fitness order.
- Surrogate-as-outcome guard: all five outcomes are marked surrogates; the CVD/mortality re-expressions are flagged as imported coefficients the NMA did not measure, with the source’s own «in reverse» wording quoted.
- Fake-tension guard: the exercise-alone-minimal-weight finding is composed with Exercise Energy Compensation / Does Weight Loss Reduce Cardiovascular Events as a consistent refinement (modality-resolved version of a held claim), not filed as a tension — they answer the same question and agree.
- Homelessness check: vsearch + title/alias scan found no existing page holding a head-to-head modality ranking for body composition; the RT-prescription page governs load/sets/frequency within resistance training, a different question — so minting, not extending, is correct.
- Attribution guard: every effect size and the ranking are chunk-pinned (corrected 2026-10-06: the Jiang and Khalafi figures were not, now tagged); magnitudes carry their CIs; the least effective and most promising verdicts are quoted (in the main-finding span) with their stratum (BMI >=30).