REF-720 · Kobayashi et al., 2026 · Frontiers in Sports and Active Living, 2026;7:1710224 · Paper · PubMed
Pooled across 14 trials the effect is ~2-3%, not 4%
And the plate alone cannot be credited for it
Key Findings
Systematic review and meta-analysis pooling crossover trials in healthy adults that pitted plated against non-plated running shoes (MEDLINE, Scopus, LILACS, Embase, searched September 2025; random-effects mean differences; PROSPERO-registered). 14 studies met inclusion criteria; the abstract does not give the pooled participant count.
| Outcome | Pooled effect |
|---|---|
| Running economy | −5.34 mL·kg⁻¹·km⁻¹ (MD; 95% CI −8.48 to −2.20) |
| Metabolic cost | −0.38 W·kg⁻¹ (MD; 95% CI −0.59 to −0.16) |
| Oxygen consumption | −1.23 mL·kg⁻¹·min⁻¹ (MD; 95% CI −1.82 to −0.63) |
| Energetic cost of transport | SMD −0.37 (95% CI −0.71 to −0.03) |
| Overall, as a percentage across outcomes | ≈2-3% lower metabolic demand; mean −2.75%, range −0.99% to −4.47% |
Certainty: moderate for running economy, metabolic cost and oxygen consumption; low for energetic cost of transport (downgraded for indirectness and imprecision).
Authors' caveat, and the important one: concurrent advanced-footwear features (compliant, resilient foams) likely contribute, so causality cannot be attributed to the plate alone; the authors call for trials that vary plate and foam separately while matching mass, stack and outsole.
Practical Application
Across 14 crossover trials in healthy adults, carbon-plated shoes lowered metabolic demand during submaximal running by about 2-3% versus non-plated shoes (mean −2.75%, range −0.99% to −4.47%; moderate certainty), and the plate alone cannot be credited. Quote 2-3%, not 4% — the famous figure (REF-719) sits near the top of this range.
Assume a distribution, not a guarantee. Inference, not tested: the reported range means some shoe-and-runner combinations will gain much less than the mean.
Do not credit the carbon plate specifically. If an athlete asks whether a plate is the thing to look for, the honest answer is that most trials did not separate plate and foam, but several prototype studies held midsole material constant, and when midsole material and geometry were matched, plate insertion generally favoured lower metabolic demand (corrected 2026-10-12, full text); the authors still do not attribute causality to the plate alone.
Inference, not tested: even at the pooled figure this is a sizeable effect on metabolic cost. The point is calibration, not dismissal.
Limitations
Metabolic outcomes only — no performance, no race times, no injury.
Certainty moderate at best; downgraded for indirectness, meaning the trials do not directly measure what runners care about.
The abstract does not give per-trial sample sizes or participants' training status beyond "healthy adults".
Inference, not tested: shoe models from different generations of technology are likely pooled into one estimate, and they are not the same intervention.
Published 2026 — no post-publication scrutiny yet.
How this entry was checked and changed
Content correction 2026-09-27
Removed per-outcome study counts (4, 9, "13 contributed data") and "typically 10-20 participants each, mostly trained male runners" — not in the abstract.
Added oxygen-consumption and ECOT results, the plated-vs-non-plated crossover design and "healthy adults".
"Dwarfs most legal ergogenic aids" and "marketing label" removed; range-interpretation and shoe-generation points labelled as inference.
Full text (checked 2026-10-12)
The 14 crossover studies (8 randomized, 6 not randomized) provided data of 271 runners aged between 18 and 45 years; participants were predominantly male, only three studies included female runners and one analyzed a female-only sample (full text).
Participants ranged from recreational to high-caliber runners, tested mostly on a treadmill at 10.8 to 16 km·h−1 (full text, Table 2).
Per-outcome study counts were k = 4 for running economy, k = 9 for metabolic cost, k = 8 for oxygen consumption and k = 5 for energetic cost of transport; 13 of the 14 studies contributed data to meta-analyses (full text).
Percentage changes were −2.88% for running economy, −2.64% for metabolic cost, −2.84% for oxygen consumption and −2.62% for energetic cost of transport; the −0.99% to −4.47% range is the mean of the 95% CI bounds, not a range of individual results (full text, Table 5).
Thirteen studies were classified as low risk of bias and one as moderate risk (average score 15.7 on a 0–20 scale) (full text).
Correction: the entry says "plate and foam were not separated in this evidence", but the full text states that "several prototype studies held midsole material constant", and that when midsole material and geometry were matched, plate insertion generally favored lower metabolic demand (full text).
The authors note shoe mass modestly elevates energy cost (∼1% per +100 g), so part of the benefit may be confounded when the plated model is lighter, though studies that tightly matched shoe mass still showed a plate-related advantage (full text).